Update batch_0602 typeI: refresh tasks (Jun 8 edits)
Browse files- baseball_pythagorean_winpct_lahman__win_pct/metadata.yaml +20 -19
- binary_pulsar_post_keplerian_atnf__PBDOT/metadata.yaml +28 -27
- bird_flight_speed_alerstam__Ue/metadata.yaml +26 -25
- bns_postmerger_waveform_core__f2/metadata.yaml +20 -19
- budyko_partitioning_mach__et_over_p/metadata.yaml +14 -13
- catchment_mean_runoff_camels__q_mean/metadata.yaml +20 -19
- cepheid_period_luminosity__M_W/metadata.yaml +14 -13
- commuting_gravity_lodes__flow_Tij/metadata.yaml +26 -25
- dna_melting_temperature_khandelwal__Tm/metadata.yaml +26 -25
- drosophila_isolation_by_distance_dest__M_stat/metadata.yaml +14 -13
- eclipsing_binary_mass_luminosity_debcat__log_L_Lsun/data/test.csv +2 -2
- eclipsing_binary_mass_luminosity_debcat__log_L_Lsun/data/train.csv +2 -2
- eclipsing_binary_mass_luminosity_debcat__log_L_Lsun/metadata.yaml +31 -26
- exfor_neutron_capture_resonance_gold__sigma_E/metadata.yaml +28 -25
- exoplanet_mass_radius__R/metadata.yaml +20 -19
- galaxy_mass_metallicity_sdss__oh/metadata.yaml +30 -27
- galaxy_stellar_mass_sdss__log_M_star/metadata.yaml +26 -25
- geomagnetic_storm_dst__dDst_dt/metadata.yaml +36 -35
- gravity_wgs84_somigliana__g0/metadata.yaml +32 -31
- gutenberg_richter_b_value_usgs__gr_freq_mag/metadata.yaml +8 -7
- hea_hardness_lattice_distortion_couzinie__HV/metadata.yaml +39 -38
- her_hor_pt_acid_kucernak__current_density/metadata.yaml +32 -31
- life_expectancy_preston__e0_vs_gdppc/metadata.yaml +29 -27
- liion_capacity_fade_nasa_pcoe__capacity_Ah/metadata.yaml +20 -19
- liquid_dispersion_glycols__refractive_index/metadata.yaml +14 -13
- lpbf_meltpool_rosenthal_inconel__depth/metadata.yaml +26 -25
- magnet_3c90_core_loss__log_P_vol/metadata.yaml +20 -19
- mars_crater_size_frequency__N_cum/metadata.yaml +27 -25
- materials_project_elastic_birch_murnaghan__K_VRH/metadata.yaml +28 -25
- mauna_loa_co2_keeling_curve_noaa__co2_ppm/metadata.yaml +14 -13
- metabolic_scaling_anage__log_bmr_W/metadata.yaml +22 -19
- mincer_earnings_acs_pums__log_wage/metadata.yaml +26 -25
- neo_size_frequency_distribution__N_cum_H/metadata.yaml +21 -19
- nuclear_binding_energy_ame2020__BE_per_A/metadata.yaml +33 -32
- ocean_mesoscale_eddy_closure__Sx/metadata.yaml +26 -25
- parker_spiral_imf__B_phi/metadata.yaml +14 -13
- pfas_soil_sorption__log_kd/metadata.yaml +14 -13
- pp_total_cross_section_donnachie_landshoff__sigma_tot/metadata.yaml +20 -19
- pv_module_temperature_efficiency_pvpmc__P_max/metadata.yaml +14 -13
- qg_turbulence_subgrid_pyqg__Sq/metadata.yaml +14 -13
- red_giant_asteroseismology__delta_nu/metadata.yaml +20 -19
- red_giant_asteroseismology__nu_max/metadata.yaml +20 -19
- rock_core_permeability_norweg__log_k/metadata.yaml +20 -19
- running_endurance_iaaf__velocity/metadata.yaml +20 -19
- sne_ia_distance_modulus_pantheonplus__mu/metadata.yaml +20 -19
- solar_flat_plate_collector_graz_efficiency__eta_thermal/metadata.yaml +28 -25
- traffic_flow_density_ngsim__q_flow/metadata.yaml +32 -31
- tree_biomass_allometry_chave__agb/metadata.yaml +20 -19
- volcanic_column_mer_ivespa__H_top/metadata.yaml +29 -25
- wind_turbine_power_curve_engie__power_kW/metadata.yaml +20 -19
baseball_pythagorean_winpct_lahman__win_pct/metadata.yaml
CHANGED
|
@@ -73,25 +73,26 @@ n_test: 938
|
|
| 73 |
# Raised caps.max_law_constants 1→2 for Pythagenport. pythagenpat is the new best
|
| 74 |
# baseline (rmse 0.02527). G does not leak win_pct (games total, not W/L split).
|
| 75 |
# Final priors: 3 distractors, 0 candidates.
|
| 76 |
-
priors:
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
|
|
|
| 95 |
|
| 96 |
references:
|
| 97 |
- id: james_1980
|
|
|
|
| 73 |
# Raised caps.max_law_constants 1→2 for Pythagenport. pythagenpat is the new best
|
| 74 |
# baseline (rmse 0.02527). G does not leak win_pct (games total, not W/L split).
|
| 75 |
# Final priors: 3 distractors, 0 candidates.
|
| 76 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 77 |
+
# priors:
|
| 78 |
+
# - name: games_per_season_modern
|
| 79 |
+
# value: 162.0
|
| 80 |
+
# unit: games
|
| 81 |
+
# description: Number of regular-season games per team in the modern MLB schedule (since 1961 AL / 1962 NL).
|
| 82 |
+
# source: Historical MLB schedule.
|
| 83 |
+
# _role: distractor
|
| 84 |
+
# - name: innings_per_game
|
| 85 |
+
# value: 9.0
|
| 86 |
+
# unit: innings
|
| 87 |
+
# description: Standard number of innings in a baseball game. A structural game-rule constant unrelated to the run-ratio power law.
|
| 88 |
+
# source: Official Baseball Rules, Rule 9.03(a); MLB Official Rules (any edition).
|
| 89 |
+
# _role: distractor
|
| 90 |
+
# - name: batting_avg_lg_typical
|
| 91 |
+
# value: 0.260
|
| 92 |
+
# unit: ""
|
| 93 |
+
# description: Approximate long-run MLB league batting average (hits per at-bat, ~0.260).
|
| 94 |
+
# source: Thorn & Palmer (1984), The Hidden Game of Baseball, p. 64; historical MLB statistics (Retrosheet).
|
| 95 |
+
# _role: distractor
|
| 96 |
|
| 97 |
references:
|
| 98 |
- id: james_1980
|
binary_pulsar_post_keplerian_atnf__PBDOT/metadata.yaml
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
task_id: binary_pulsar_post_keplerian_atnf__PBDOT
|
| 2 |
-
metric: rmse
|
| 3 |
domain: astronomy / pulsar_timing
|
| 4 |
type: typeI
|
| 5 |
has_group_id: false
|
|
@@ -130,31 +130,32 @@ n_test: 10
|
|
| 130 |
# either below measurement noise on the test set or require
|
| 131 |
# free parameters not measured for these systems. See WAVE17_ATTEMPT.md.
|
| 132 |
|
| 133 |
-
priors:
|
| 134 |
-
|
| 135 |
-
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
|
| 141 |
-
|
| 142 |
-
|
| 143 |
-
|
| 144 |
-
|
| 145 |
-
|
| 146 |
-
|
| 147 |
-
|
| 148 |
-
|
| 149 |
-
|
| 150 |
-
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
|
| 154 |
-
|
| 155 |
-
|
| 156 |
-
|
| 157 |
-
|
|
|
|
| 158 |
|
| 159 |
# Reference-baseline bank
|
| 160 |
# Wave-17 (2026-05-27): single live baseline (SB tag accepted).
|
|
@@ -197,4 +198,4 @@ validity_rubrics:
|
|
| 197 |
- "predicted PBDOT is strictly negative for every bound binary with Pb > 0 and 0 <= e < 1"
|
| 198 |
- "predicted PBDOT remains defined throughout the physical input domain Pb > 0 and 0 <= e < 1"
|
| 199 |
|
| 200 |
-
best_baseline:
|
|
|
|
| 1 |
task_id: binary_pulsar_post_keplerian_atnf__PBDOT
|
| 2 |
+
metric: smape # 2026-06-08 rmse->smape: PBDOT ~1e-12, so abs rmse <<1e-4 always trips the harness non-discriminative guard (numeric_score=None); smape is scale-free (ref peters_1963 smape=0.117, discriminative)
|
| 3 |
domain: astronomy / pulsar_timing
|
| 4 |
type: typeI
|
| 5 |
has_group_id: false
|
|
|
|
| 130 |
# either below measurement noise on the test set or require
|
| 131 |
# free parameters not measured for these systems. See WAVE17_ATTEMPT.md.
|
| 132 |
|
| 133 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 134 |
+
# priors:
|
| 135 |
+
# - name: T_sun
|
| 136 |
+
# value: 4.925490947e-6
|
| 137 |
+
# unit: s
|
| 138 |
+
# description: Solar mass in geometric time units (G M_sun / c^3). Universal physics constant.
|
| 139 |
+
# source: Kramer et al. 2006, PDF p. 9
|
| 140 |
+
# _role: physics_anchor
|
| 141 |
+
# - name: M_c_canonical
|
| 142 |
+
# value: 1.20
|
| 143 |
+
# unit: M_sun
|
| 144 |
+
# description: Canonical chirp mass for galactic DNS binaries (population mean).
|
| 145 |
+
# source: Antoniadis 2013 ApJ 778 / Bagchi 2013 MNRAS 428
|
| 146 |
+
# _role: physics_anchor
|
| 147 |
+
# - name: coeff_73_24
|
| 148 |
+
# value: 3.0416667
|
| 149 |
+
# unit: ""
|
| 150 |
+
# description: Dimensionless rational coefficient 73/24.
|
| 151 |
+
# source: null
|
| 152 |
+
# _role: distractor
|
| 153 |
+
# - name: coeff_37_96
|
| 154 |
+
# value: 0.3854167
|
| 155 |
+
# unit: ""
|
| 156 |
+
# description: Dimensionless rational coefficient 37/96.
|
| 157 |
+
# source: null
|
| 158 |
+
# _role: distractor
|
| 159 |
|
| 160 |
# Reference-baseline bank
|
| 161 |
# Wave-17 (2026-05-27): single live baseline (SB tag accepted).
|
|
|
|
| 198 |
- "predicted PBDOT is strictly negative for every bound binary with Pb > 0 and 0 <= e < 1"
|
| 199 |
- "predicted PBDOT remains defined throughout the physical input domain Pb > 0 and 0 <= e < 1"
|
| 200 |
|
| 201 |
+
best_baseline: 0.117242 # peters_1963 — smape (lower is better); metric changed rmse->smape 2026-06-08
|
bird_flight_speed_alerstam__Ue/metadata.yaml
CHANGED
|
@@ -73,31 +73,32 @@ n_test: 62
|
|
| 73 |
# discover the functional form Ue = sqrt(2*m*g/(rho_0*S*C_L))).
|
| 74 |
|
| 75 |
# Optional priors — grounded from reference papers and aerodynamic literature
|
| 76 |
-
priors:
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
|
|
|
| 101 |
|
| 102 |
# Reference-baseline bank
|
| 103 |
references:
|
|
|
|
| 73 |
# discover the functional form Ue = sqrt(2*m*g/(rho_0*S*C_L))).
|
| 74 |
|
| 75 |
# Optional priors — grounded from reference papers and aerodynamic literature
|
| 76 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 77 |
+
# priors:
|
| 78 |
+
# - name: g_gravity
|
| 79 |
+
# value: 9.81
|
| 80 |
+
# unit: m/s^2
|
| 81 |
+
# description: Standard gravitational acceleration — BIPM/ISO 80000 definition (exact). A universal physical constant.
|
| 82 |
+
# source: CODATA/BIPM; ISO 80000-3.
|
| 83 |
+
# _role: candidate
|
| 84 |
+
# - name: rho_0_sea_level
|
| 85 |
+
# value: 1.225
|
| 86 |
+
# unit: kg/m^3
|
| 87 |
+
# description: "ISA sea-level air density — ICAO Standard Atmosphere (ISO 2533:1975) reference value. Defines the equivalent airspeed (Ue): true airspeed corrected to sea-level air density. A universal atmospheric reference constant."
|
| 88 |
+
# source: ICAO Standard Atmosphere ISO 2533:1975.
|
| 89 |
+
# _role: candidate
|
| 90 |
+
# - name: exponent_Q_theoretical
|
| 91 |
+
# value: 0.5
|
| 92 |
+
# unit: dimensionless
|
| 93 |
+
# description: A candidate exponent value for the airspeed vs wing-loading scaling under dynamical similarity.
|
| 94 |
+
# source: null
|
| 95 |
+
# _role: distractor
|
| 96 |
+
# - name: exponent_mass_theoretical
|
| 97 |
+
# value: 0.1667
|
| 98 |
+
# unit: dimensionless
|
| 99 |
+
# description: A candidate exponent value for the airspeed vs body-mass scaling under geometrical similarity.
|
| 100 |
+
# source: null
|
| 101 |
+
# _role: distractor
|
| 102 |
|
| 103 |
# Reference-baseline bank
|
| 104 |
references:
|
bns_postmerger_waveform_core__f2/metadata.yaml
CHANGED
|
@@ -110,25 +110,26 @@ references:
|
|
| 110 |
# both of which ARE in USED_INPUTS).
|
| 111 |
# The OOD signal IS anchored through the m1/m2 → q transform.
|
| 112 |
# Borderline, no redesign needed; documented for transparency per FM G1 v0.7.
|
| 113 |
-
priors:
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
|
| 129 |
-
|
| 130 |
-
|
| 131 |
-
|
|
|
|
| 132 |
|
| 133 |
caps:
|
| 134 |
max_law_constants: 10
|
|
|
|
| 110 |
# both of which ARE in USED_INPUTS).
|
| 111 |
# The OOD signal IS anchored through the m1/m2 → q transform.
|
| 112 |
# Borderline, no redesign needed; documented for transparency per FM G1 v0.7.
|
| 113 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 114 |
+
# priors:
|
| 115 |
+
# - name: GM_sun_c3
|
| 116 |
+
# value: 4.925490947e-6
|
| 117 |
+
# unit: s
|
| 118 |
+
# description: IAU/CODATA derived constant G*M_sun/c^3 — the geometric-units solar mass in seconds; exact value follows from the IAU 2012 nominal solar mass parameter GM_sun = 1.32712440018e20 m^3/s^2 and the exact CODATA speed of light c = 299792458 m/s. This is NOT a fit coefficient; it is a universal conversion constant fixed by IAU+CODATA, independent of any NR simulation dataset.
|
| 119 |
+
# source: IAU 2012 Division I Working Group (TDB-compatible GM_sun); CODATA 2018 c.
|
| 120 |
+
# _role: candidate
|
| 121 |
+
# - name: f2_typical_equal_mass
|
| 122 |
+
# value: 2.5
|
| 123 |
+
# unit: kHz
|
| 124 |
+
# description: Typical post-merger peak frequency for equal-mass BNS with total mass ~2.7 M_sun and moderate EOS.
|
| 125 |
+
# source: null
|
| 126 |
+
# _role: distractor
|
| 127 |
+
# - name: nuclear_saturation_density
|
| 128 |
+
# value: 2.3e14
|
| 129 |
+
# unit: g/cm^3
|
| 130 |
+
# description: Nuclear saturation density rho_0 — density at which neutron star matter reaches equilibrium.
|
| 131 |
+
# source: Standard nuclear-physics value.
|
| 132 |
+
# _role: distractor
|
| 133 |
|
| 134 |
caps:
|
| 135 |
max_law_constants: 10
|
budyko_partitioning_mach__et_over_p/metadata.yaml
CHANGED
|
@@ -150,19 +150,20 @@ references:
|
|
| 150 |
# would leak the answer (FM C12 / FM-H1). The block is therefore distractors-only:
|
| 151 |
# two genuine hydrology reference values that name no LAW/OTHER constant of this
|
| 152 |
# task's baselines. The SR system must discover the shape parameter itself.
|
| 153 |
-
priors:
|
| 154 |
-
|
| 155 |
-
|
| 156 |
-
|
| 157 |
-
|
| 158 |
-
|
| 159 |
-
|
| 160 |
-
|
| 161 |
-
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
|
|
|
|
| 166 |
|
| 167 |
# === v2 anti-dump caps ===
|
| 168 |
caps:
|
|
|
|
| 150 |
# would leak the answer (FM C12 / FM-H1). The block is therefore distractors-only:
|
| 151 |
# two genuine hydrology reference values that name no LAW/OTHER constant of this
|
| 152 |
# task's baselines. The SR system must discover the shape parameter itself.
|
| 153 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 154 |
+
# priors:
|
| 155 |
+
# - name: Budyko_asymptote_humid
|
| 156 |
+
# value: 1.0
|
| 157 |
+
# unit: ""
|
| 158 |
+
# description: Theoretical upper asymptote of the Budyko curve in the energy-limited (humid) regime — ET/P approaches 1 as PET/P → ∞. This is a structural boundary condition common to ALL Budyko-framework formulas, not a fitted parameter.
|
| 159 |
+
# source: Budyko (1974) Climate and Life; Milly (1994) Water Resources Research 30:2065-2077.
|
| 160 |
+
# _role: distractor
|
| 161 |
+
# - name: omega_Zhang_2001_global
|
| 162 |
+
# value: 0.9
|
| 163 |
+
# unit: ""
|
| 164 |
+
# description: Zhang et al. (2001) plant-available water coefficient omega for mixed vegetation (forests + grasslands), a shape parameter in the Zhang-Dawes-Walker Budyko variant.
|
| 165 |
+
# source: "Zhang L et al. (2001) Wat. Resour. Res. 37:701-708, Table 1 — omega=0.9 for mixed (forest+grass) catchments."
|
| 166 |
+
# _role: distractor
|
| 167 |
|
| 168 |
# === v2 anti-dump caps ===
|
| 169 |
caps:
|
catchment_mean_runoff_camels__q_mean/metadata.yaml
CHANGED
|
@@ -137,25 +137,26 @@ n_train: 469
|
|
| 137 |
n_test: 201
|
| 138 |
|
| 139 |
# Candidate prior constants — the `priors` prompt slot.
|
| 140 |
-
priors:
|
| 141 |
-
|
| 142 |
-
|
| 143 |
-
|
| 144 |
-
|
| 145 |
-
|
| 146 |
-
|
| 147 |
-
|
| 148 |
-
|
| 149 |
-
|
| 150 |
-
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
|
| 154 |
-
|
| 155 |
-
|
| 156 |
-
|
| 157 |
-
|
| 158 |
-
|
|
|
|
| 159 |
|
| 160 |
# === Reference baselines ===
|
| 161 |
references:
|
|
|
|
| 137 |
n_test: 201
|
| 138 |
|
| 139 |
# Candidate prior constants — the `priors` prompt slot.
|
| 140 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 141 |
+
# priors:
|
| 142 |
+
# - name: budyko_fu_exponent
|
| 143 |
+
# value: 2.6
|
| 144 |
+
# unit: dimensionless
|
| 145 |
+
# description: Typical Budyko/Fu/Choudhury shape exponent for mid-latitude catchments; a candidate value for the one-parameter Budyko curve's exponent.
|
| 146 |
+
# source: Reported across field sites and large basins in the range α ≈ 1.5-3.0. A candidate for the shape exponent.
|
| 147 |
+
# _role: candidate
|
| 148 |
+
# - name: priestley_taylor_alpha
|
| 149 |
+
# value: 1.26
|
| 150 |
+
# unit: dimensionless
|
| 151 |
+
# description: Priestley-Taylor coefficient relating equilibrium to potential evapotranspiration.
|
| 152 |
+
# source: Priestley & Taylor 1972 Mon. Weather Rev. 100:81. Used in computing pet_mean.
|
| 153 |
+
# _role: distractor
|
| 154 |
+
# - name: runoff_ratio_global_mean
|
| 155 |
+
# value: 0.39
|
| 156 |
+
# unit: dimensionless
|
| 157 |
+
# description: Approximate global-mean runoff ratio Q/P for continental land surfaces.
|
| 158 |
+
# source: Budyko 1974 "Climate and Life". A rough scale for Q/P; not a closed-form constant (the actual ratio varies from ~0 in deserts to ~0.9 in wet maritime catchments).
|
| 159 |
+
# _role: candidate
|
| 160 |
|
| 161 |
# === Reference baselines ===
|
| 162 |
references:
|
cepheid_period_luminosity__M_W/metadata.yaml
CHANGED
|
@@ -63,19 +63,20 @@ n_test: 7
|
|
| 63 |
# and neither value is consumed by any shipped W_G baseline. Final priors: 2 distractors
|
| 64 |
# anchored to disk PDFs (gamma_WH from riess_2022 / cruzreyes_2023 Eq. 15; R_WG from
|
| 65 |
# cruzreyes_2023 Eq. 6).
|
| 66 |
-
priors:
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
|
|
|
| 79 |
|
| 80 |
references:
|
| 81 |
- id: breuval_2022
|
|
|
|
| 63 |
# and neither value is consumed by any shipped W_G baseline. Final priors: 2 distractors
|
| 64 |
# anchored to disk PDFs (gamma_WH from riess_2022 / cruzreyes_2023 Eq. 15; R_WG from
|
| 65 |
# cruzreyes_2023 Eq. 6).
|
| 66 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 67 |
+
# priors:
|
| 68 |
+
# - name: gamma_WH
|
| 69 |
+
# value: -0.217
|
| 70 |
+
# unit: mag/dex
|
| 71 |
+
# description: Metallicity coefficient for the NIR W_H Wesenheit band, from Riess 2022; adopted as fixed in Cruz Reyes 2023 W_H fit. Not applicable to the Gaia W_G band used in this task.
|
| 72 |
+
# source: riess_2022; Cruz Reyes 2023 Eq. 15, PDF p. 14
|
| 73 |
+
# _role: distractor
|
| 74 |
+
# - name: R_WG
|
| 75 |
+
# value: 1.921
|
| 76 |
+
# unit: ""
|
| 77 |
+
# description: Reddening ratio for the Gaia W_G Wesenheit definition (G − 1.921*(Bp − Rp)). A photometric construction constant, not a PLZ slope or zero-point.
|
| 78 |
+
# source: cruzreyes_2023, Eq. 6, PDF p. 12
|
| 79 |
+
# _role: distractor
|
| 80 |
|
| 81 |
references:
|
| 82 |
- id: breuval_2022
|
commuting_gravity_lodes__flow_Tij/metadata.yaml
CHANGED
|
@@ -61,31 +61,32 @@ n_test: 5284
|
|
| 61 |
# derived scaling-law values at a specific US county mean area (S=2596.8 km^2);
|
| 62 |
# not universal physical constants. Removed (strategy B).
|
| 63 |
# Two existing clean distractors retained; two more added.
|
| 64 |
-
priors:
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
|
|
|
| 89 |
references:
|
| 90 |
- id: simini_2012
|
| 91 |
label: "Simini 2012 \u2014 parameter-free radiation model (Eq. 2)"
|
|
|
|
| 61 |
# derived scaling-law values at a specific US county mean area (S=2596.8 km^2);
|
| 62 |
# not universal physical constants. Removed (strategy B).
|
| 63 |
# Two existing clean distractors retained; two more added.
|
| 64 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 65 |
+
# priors:
|
| 66 |
+
# - name: commuter_fraction_US2000
|
| 67 |
+
# value: 0.12
|
| 68 |
+
# unit: dimensionless
|
| 69 |
+
# description: Ratio of total commuters to total population N_c/N for US Census 2000. An empirical census figure for the US population.
|
| 70 |
+
# source: "US Census 2000 (34,116,820 commuters / 281M total population \u2248 0.121)."
|
| 71 |
+
# _role: distractor
|
| 72 |
+
# - name: zipf_slope_empirical
|
| 73 |
+
# value: 0.8472
|
| 74 |
+
# unit: dimensionless
|
| 75 |
+
# description: Empirical slope b of a log-log fit on historical Railway Express data (78 city-pairs, May 1939). A historical gravity-model calibration for 1930s rail traffic.
|
| 76 |
+
# source: Zipf 1946.
|
| 77 |
+
# _role: distractor
|
| 78 |
+
# - name: gravity_exponent_classic
|
| 79 |
+
# value: 2.0
|
| 80 |
+
# unit: dimensionless
|
| 81 |
+
# description: Classic gravity-model distance exponent by analogy with Newtonian gravity. A theoretical reference value.
|
| 82 |
+
# source: Ravenstein 1885 / Newton analogy.
|
| 83 |
+
# _role: distractor
|
| 84 |
+
# - name: beta_pow_EU_municipalities
|
| 85 |
+
# value: 1.74
|
| 86 |
+
# unit: dimensionless
|
| 87 |
+
# description: Power-law distance exponent for European municipality-level commuting at mean area S=71.4 km^2.
|
| 88 |
+
# source: null
|
| 89 |
+
# _role: distractor
|
| 90 |
references:
|
| 91 |
- id: simini_2012
|
| 92 |
label: "Simini 2012 \u2014 parameter-free radiation model (Eq. 2)"
|
dna_melting_temperature_khandelwal__Tm/metadata.yaml
CHANGED
|
@@ -75,31 +75,32 @@ n_test: 100
|
|
| 75 |
# gold philosophy that an empty candidate slot is preferable to leaking fit answers).
|
| 76 |
# Reference: audit/AUDIT_CHECKLIST.md §C12 (v0.6);
|
| 77 |
# audit/proposed_FMs/2026-05-26_priors_leakage_GM.md §6.
|
| 78 |
-
priors:
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
|
|
|
| 103 |
references:
|
| 104 |
- id: khandelwal_2010
|
| 105 |
label: Khandelwal & Bhyravabhotla 2010 — phenomenological linear regression model (Eq. 1)
|
|
|
|
| 75 |
# gold philosophy that an empty candidate slot is preferable to leaking fit answers).
|
| 76 |
# Reference: audit/AUDIT_CHECKLIST.md §C12 (v0.6);
|
| 77 |
# audit/proposed_FMs/2026-05-26_priors_leakage_GM.md §6.
|
| 78 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 79 |
+
# priors:
|
| 80 |
+
# - name: pKa_phosphate_backbone
|
| 81 |
+
# value: 2.0
|
| 82 |
+
# unit: dimensionless
|
| 83 |
+
# description: Approximate pKa of the DNA phosphate backbone. The backbone phosphodiester groups carry a negative charge at physiological pH (pKa ~2), contributing to electrostatic interactions screened by [Na+].
|
| 84 |
+
# source: Blackburn & Gait (eds.) "Nucleic Acids in Chemistry and Biology" 3rd ed. (2006); standard biochemistry reference.
|
| 85 |
+
# _role: distractor
|
| 86 |
+
# - name: Boltzmann_J_per_K
|
| 87 |
+
# value: 1.380649e-23
|
| 88 |
+
# unit: J/K
|
| 89 |
+
# description: Boltzmann constant k_B (CODATA 2018 exact definition). Appears in thermodynamic treatments of DNA helix-coil transitions (e.g. nearest-neighbour entropy terms).
|
| 90 |
+
# source: CODATA 2018 exact value (SI redefinition).
|
| 91 |
+
# _role: distractor
|
| 92 |
+
# - name: R_gas_J_per_mol_K
|
| 93 |
+
# value: 8.314462618
|
| 94 |
+
# unit: J/(mol*K)
|
| 95 |
+
# description: Molar gas constant R (CODATA 2018). Appears in van't Hoff / nearest-neighbour thermodynamic melting models (ΔH / ΔS framework).
|
| 96 |
+
# source: CODATA 2018 (derived from k_B × N_A).
|
| 97 |
+
# _role: distractor
|
| 98 |
+
# - name: dna_persistence_length_bp
|
| 99 |
+
# value: 50.0
|
| 100 |
+
# unit: bp
|
| 101 |
+
# description: Canonical persistence length of double-stranded B-form DNA (~50 bp ≈ 17 nm). Governs elastic properties and loop formation probabilities.
|
| 102 |
+
# source: Bustamante et al. (1994) Science 265:1599; Hagerman (1988) Annu Rev Biophys Biophys Chem 17:265 — standard biophysics value.
|
| 103 |
+
# _role: distractor
|
| 104 |
references:
|
| 105 |
- id: khandelwal_2010
|
| 106 |
label: Khandelwal & Bhyravabhotla 2010 — phenomenological linear regression model (Eq. 1)
|
drosophila_isolation_by_distance_dest__M_stat/metadata.yaml
CHANGED
|
@@ -49,19 +49,20 @@ n_train: 828
|
|
| 49 |
n_test: 550
|
| 50 |
|
| 51 |
# Candidate prior constants — the `priors` prompt slot.
|
| 52 |
-
priors:
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
|
|
|
| 65 |
|
| 66 |
# === Reference baselines ===
|
| 67 |
references:
|
|
|
|
| 49 |
n_test: 550
|
| 50 |
|
| 51 |
# Candidate prior constants — the `priors` prompt slot.
|
| 52 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 53 |
+
# priors:
|
| 54 |
+
# - name: ibd_2d_slope_mainland_subset
|
| 55 |
+
# value: 0.0091
|
| 56 |
+
# unit: 1 / ln(km)
|
| 57 |
+
# description: Drosophila-specific b = 1/(4πD·σ²) self-fit on a mainland Europe subset only.
|
| 58 |
+
# source: Sanity-check OLS regression on mainland Europe pairs. Provided as a candidate ballpark for the 2-D isolation-by-distance slope.
|
| 59 |
+
# _role: candidate
|
| 60 |
+
# - name: chromosome_translocation_rate_typical
|
| 61 |
+
# value: 1.0e-3
|
| 62 |
+
# unit: per generation
|
| 63 |
+
# description: A typical Drosophila chromosomal translocation rate.
|
| 64 |
+
# source: Drosophila-adjacent quantity; population differentiation in this task is driven by pair-wise geographic distance, not chromosomal rearrangement.
|
| 65 |
+
# _role: distractor
|
| 66 |
|
| 67 |
# === Reference baselines ===
|
| 68 |
references:
|
eclipsing_binary_mass_luminosity_debcat__log_L_Lsun/data/test.csv
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b9bf0dd438ccbb09a0b188bc5f412b088595e244750e1df623b9e93ed1165507
|
| 3 |
+
size 7330
|
eclipsing_binary_mass_luminosity_debcat__log_L_Lsun/data/train.csv
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:644980babb3aff2c181d0d5c62b959210eb2b36bf0bff40bf1f8240b1ed8327c
|
| 3 |
+
size 4929
|
eclipsing_binary_mass_luminosity_debcat__log_L_Lsun/metadata.yaml
CHANGED
|
@@ -29,7 +29,7 @@ inputs:
|
|
| 29 |
- name: metallicity
|
| 30 |
symbol: "[M/H]"
|
| 31 |
unit: dex
|
| 32 |
-
description: Logarithmic stellar metallicity relative to the Sun.
|
| 33 |
range:
|
| 34 |
train: [-1.86, 0.55]
|
| 35 |
test: [-1.47, 0.51]
|
|
@@ -64,31 +64,36 @@ n_train: 238
|
|
| 64 |
n_test: 355
|
| 65 |
|
| 66 |
# Candidate prior constants — the `priors` prompt slot.
|
| 67 |
-
priors:
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 92 |
|
| 93 |
# === Reference baselines ===
|
| 94 |
references:
|
|
|
|
| 29 |
- name: metallicity
|
| 30 |
symbol: "[M/H]"
|
| 31 |
unit: dex
|
| 32 |
+
description: Logarithmic stellar metallicity relative to the Sun. A secondary/distractor input — the reference M-L laws use stellar mass only. Rows without a published spectroscopic determination (63% of train / 33% of test) are filled with the train-set observed median (-0.11 dex) as a neutral placeholder (2026-06-08; the harness requires finite inputs, and median fill avoids the solar-0.0 metal-rich bias).
|
| 33 |
range:
|
| 34 |
train: [-1.86, 0.55]
|
| 35 |
test: [-1.47, 0.51]
|
|
|
|
| 64 |
n_test: 355
|
| 65 |
|
| 66 |
# Candidate prior constants — the `priors` prompt slot.
|
| 67 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 68 |
+
# priors:
|
| 69 |
+
# [LEAK-REVIEW 2026-06-08] value 3.5 = textbook M-L exponent ≈ fitted answer (eker_2018.A5, LAW≈20%);
|
| 70 |
+
# flagged for the later language-revision pass (also rename key off "kuiper_1938").
|
| 71 |
+
# - name: kuiper_1938_slope_classical
|
| 72 |
+
# value: 3.5
|
| 73 |
+
# unit: dimensionless
|
| 74 |
+
# description: A historically-reported mass-luminosity power-law slope.
|
| 75 |
+
# source: Early empirical fit on an available stellar sample. A candidate value for the power-law slope; modern compilations with more high-mass coverage tend toward a steeper value.
|
| 76 |
+
# _role: candidate
|
| 77 |
+
# [LEAK-REVIEW 2026-06-08] value 3.0 ≈ fitted M-L slope (eker_2018.A6=2.865, LAW≈5%);
|
| 78 |
+
# flagged for the later language-revision pass (defer relax/removal).
|
| 79 |
+
# - name: eddington_classical_ml_slope
|
| 80 |
+
# value: 3.0
|
| 81 |
+
# unit: dimensionless
|
| 82 |
+
# description: A theoretical mass-luminosity slope for uniform-composition stars dominated by electron-scattering opacity.
|
| 83 |
+
# source: Classical stellar-structure theory. A candidate value for the power-law slope; the idealized theoretical slope is shallower than empirical fits because real main-sequence stars have varying opacity and composition.
|
| 84 |
+
# _role: candidate
|
| 85 |
+
# - name: eddington_luminosity_solar
|
| 86 |
+
# value: 1.26e31
|
| 87 |
+
# unit: W
|
| 88 |
+
# description: Eddington luminosity for a 1 M_sun star with electron-scattering opacity.
|
| 89 |
+
# source: Classical stellar-structure theory. The Eddington-limit luminosity above which radiation pressure exceeds gravity; only relevant for super-massive (M > 50 M_sun) or accreting objects.
|
| 90 |
+
# _role: distractor
|
| 91 |
+
# - name: chandrasekhar_mass_limit
|
| 92 |
+
# value: 1.4
|
| 93 |
+
# unit: M_sun
|
| 94 |
+
# description: Maximum mass of a stable white dwarf.
|
| 95 |
+
# source: Chandrasekhar 1931, ApJ 74:81. Relevant to white-dwarf physics, not to main-sequence mass-luminosity scaling.
|
| 96 |
+
# _role: distractor
|
| 97 |
|
| 98 |
# === Reference baselines ===
|
| 99 |
references:
|
exfor_neutron_capture_resonance_gold__sigma_E/metadata.yaml
CHANGED
|
@@ -58,31 +58,34 @@ n_train: 1732
|
|
| 58 |
n_test: 433
|
| 59 |
|
| 60 |
# Candidate prior constants — the `priors` prompt slot.
|
| 61 |
-
priors:
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
|
|
|
|
|
|
|
|
|
| 86 |
|
| 87 |
# === Reference baselines ===
|
| 88 |
references:
|
|
|
|
| 58 |
n_test: 433
|
| 59 |
|
| 60 |
# Candidate prior constants — the `priors` prompt slot.
|
| 61 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 62 |
+
# priors:
|
| 63 |
+
# [LEAK-REMOVE 2026-06-08] value = fitted answer constant; commented out (NOT deleted).
|
| 64 |
+
# - name: au197_resonance_energy
|
| 65 |
+
# value: 4.906
|
| 66 |
+
# unit: eV
|
| 67 |
+
# description: Energy of the dominant low-energy s-wave resonance of Au-197; a candidate for the resonance peak energy.
|
| 68 |
+
# source: Mughabghab 'Atlas of Neutron Resonances' 6th ed. 2018 (reproduced in ENDF/B-VIII.0, JEFF-3.3, JENDL-5 within ~0.5%).
|
| 69 |
+
# _role: candidate
|
| 70 |
+
# [LEAK-REMOVE 2026-06-08] value = fitted answer constant; commented out (NOT deleted).
|
| 71 |
+
# - name: au197_resonance_width
|
| 72 |
+
# value: 0.1372
|
| 73 |
+
# unit: eV
|
| 74 |
+
# description: Total width Gamma_total = Gamma_n + Gamma_gamma of the 4.906 eV resonance; a candidate for the resonance width.
|
| 75 |
+
# source: Mughabghab 2018 (Gamma_n = 15.2 meV, Gamma_gamma = 122 meV).
|
| 76 |
+
# _role: candidate
|
| 77 |
+
# - name: one_over_v_exponent
|
| 78 |
+
# value: -0.5
|
| 79 |
+
# unit: dimensionless
|
| 80 |
+
# description: The energy exponent in the s-wave low-energy limit of compound-nucleus capture theory.
|
| 81 |
+
# source: Compound-nucleus reaction theory. A candidate for the background power-law exponent.
|
| 82 |
+
# _role: candidate
|
| 83 |
+
# - name: au197_thermal_sigma
|
| 84 |
+
# value: 98.66
|
| 85 |
+
# unit: barn
|
| 86 |
+
# description: Thermal (0.0253 eV) capture cross section of Au-197 (a neutron-flux standard).
|
| 87 |
+
# source: Mughabghab 2018. Sets the smooth background scale; a candidate for the cross-section normalisation.
|
| 88 |
+
# _role: candidate
|
| 89 |
|
| 90 |
# === Reference baselines ===
|
| 91 |
references:
|
exoplanet_mass_radius__R/metadata.yaml
CHANGED
|
@@ -89,25 +89,26 @@ caps:
|
|
| 89 |
# the formula stores log10 of these values — no numerical match — so they remained CLEAN.
|
| 90 |
# (Audit 2026-05-29: T1/T3 are NOT in the live priors block below — it is three
|
| 91 |
# distractors (AU, R_jupiter, solar_metallicity), none matching a chen/bashi LAW value.)
|
| 92 |
-
priors:
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
|
|
|
| 111 |
|
| 112 |
metric: smape
|
| 113 |
# === Validity rubric ===
|
|
|
|
| 89 |
# the formula stores log10 of these values — no numerical match — so they remained CLEAN.
|
| 90 |
# (Audit 2026-05-29: T1/T3 are NOT in the live priors block below — it is three
|
| 91 |
# distractors (AU, R_jupiter, solar_metallicity), none matching a chen/bashi LAW value.)
|
| 92 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 93 |
+
# priors:
|
| 94 |
+
# - name: AU_m
|
| 95 |
+
# value: 149597870700.0
|
| 96 |
+
# unit: m
|
| 97 |
+
# description: IAU 2012 exact definition of the astronomical unit — an orbital distance scale.
|
| 98 |
+
# source: "IAU 2012 Resolution B2 (exact: 1 au = 149597870700 m)."
|
| 99 |
+
# _role: distractor
|
| 100 |
+
# - name: R_jupiter_in_R_earth
|
| 101 |
+
# value: 11.21
|
| 102 |
+
# unit: R_earth
|
| 103 |
+
# description: Radius of Jupiter in Earth radii — a planetary radius scale.
|
| 104 |
+
# source: IAU 2015 nominal value; standard planetary science reference.
|
| 105 |
+
# _role: distractor
|
| 106 |
+
# - name: solar_metallicity_Z
|
| 107 |
+
# value: 0.0134
|
| 108 |
+
# unit: dimensionless
|
| 109 |
+
# description: Present-day solar bulk metallicity (heavy-element mass fraction) — a stellar abundance reference used to contextualise host-star properties.
|
| 110 |
+
# source: Asplund et al. 2009, Annual Reviews of Astronomy and Astrophysics 47, 481.
|
| 111 |
+
# _role: distractor
|
| 112 |
|
| 113 |
metric: smape
|
| 114 |
# === Validity rubric ===
|
galaxy_mass_metallicity_sdss__oh/metadata.yaml
CHANGED
|
@@ -62,7 +62,8 @@ n_test: 38981
|
|
| 62 |
# ~0.28 dex C9 calibration offset and stay negative (mixed bank, §9.14). Metric now
|
| 63 |
# discriminates (R2 from +0.58 down to -0.51).
|
| 64 |
|
| 65 |
-
priors:
|
|
|
|
| 66 |
# FM C12 fix (2026-05-26): removed andrews_Z_asm=8.798 (Andrews 2013 LAW_CONSTANTS
|
| 67 |
# Z_asm) and curti_Z0=8.793 (Curti 2020 LAW_CONSTANTS Z0) — both are paper-specific
|
| 68 |
# empirical fit parameters (asymptotic plateau values from regression fits to SDSS
|
|
@@ -74,32 +75,34 @@ priors:
|
|
| 74 |
# scale is a physical parameter constrained by baryon physics, not a free fit number).
|
| 75 |
# Retained MZR_intrinsic_scatter and mannucci_FMR_scatter as distractors (astrophysical
|
| 76 |
# scatter values — real but not consumed by any of the formula's LAW/OTHER constants).
|
| 77 |
-
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
|
|
|
|
|
|
| 103 |
|
| 104 |
# Sanity-gate note (post RE-SPLIT 2026-05-29): with the representative split this is a
|
| 105 |
# proper mixed bank. 4 SDSS-strong-line baselines beat the naive mean —
|
|
|
|
| 62 |
# ~0.28 dex C9 calibration offset and stay negative (mixed bank, §9.14). Metric now
|
| 63 |
# discriminates (R2 from +0.58 down to -0.51).
|
| 64 |
|
| 65 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 66 |
+
# priors:
|
| 67 |
# FM C12 fix (2026-05-26): removed andrews_Z_asm=8.798 (Andrews 2013 LAW_CONSTANTS
|
| 68 |
# Z_asm) and curti_Z0=8.793 (Curti 2020 LAW_CONSTANTS Z0) — both are paper-specific
|
| 69 |
# empirical fit parameters (asymptotic plateau values from regression fits to SDSS
|
|
|
|
| 75 |
# scale is a physical parameter constrained by baryon physics, not a free fit number).
|
| 76 |
# Retained MZR_intrinsic_scatter and mannucci_FMR_scatter as distractors (astrophysical
|
| 77 |
# scatter values — real but not consumed by any of the formula's LAW/OTHER constants).
|
| 78 |
+
# [LEAK-REVIEW 2026-06-08] value 10.5 ≈ fitted MZR turnover (curti_2020.log_M0=10.02, LAW≈5%);
|
| 79 |
+
# flagged for the later language-revision pass (defer relax/removal).
|
| 80 |
+
# - name: MZR_turnover_logM_tremonti
|
| 81 |
+
# value: 10.5
|
| 82 |
+
# unit: log(M_sun)
|
| 83 |
+
# description: >-
|
| 84 |
+
# Approximate stellar mass at which the gas-phase mass-metallicity relation flattens
|
| 85 |
+
# (upper calibrated range, log M_* ~ 10.5 M_sun). Sets the physical scale for the
|
| 86 |
+
# saturation knee — a structurally important candidate for the turnover mass scale.
|
| 87 |
+
# source: "Empirical fit range upper bound from SDSS star-forming galaxy samples."
|
| 88 |
+
# _role: candidate
|
| 89 |
+
# - name: MZR_intrinsic_scatter
|
| 90 |
+
# value: 0.1
|
| 91 |
+
# unit: dex
|
| 92 |
+
# description: >-
|
| 93 |
+
# Typical 1-sigma intrinsic scatter around the SDSS mass-metallicity relation
|
| 94 |
+
# (reflects the irreducible astrophysical scatter from star-formation history
|
| 95 |
+
# variation at fixed stellar mass).
|
| 96 |
+
# source: "Standard value reported for SDSS star-forming galaxy samples."
|
| 97 |
+
# _role: distractor
|
| 98 |
+
# - name: mannucci_FMR_scatter
|
| 99 |
+
# value: 0.05
|
| 100 |
+
# unit: dex
|
| 101 |
+
# description: >-
|
| 102 |
+
# Residual 1-sigma scatter around the star-formation-rate-dependent fundamental
|
| 103 |
+
# metallicity relation. Approximately half the mass-metallicity-relation scatter.
|
| 104 |
+
# source: "Standard value reported for SDSS star-forming galaxy samples."
|
| 105 |
+
# _role: distractor
|
| 106 |
|
| 107 |
# Sanity-gate note (post RE-SPLIT 2026-05-29): with the representative split this is a
|
| 108 |
# proper mixed bank. 4 SDSS-strong-line baselines beat the naive mean —
|
galaxy_stellar_mass_sdss__log_M_star/metadata.yaml
CHANGED
|
@@ -168,7 +168,8 @@ best_baseline: 0.315019 # taylor_2011 — rmse (lower is better); regenerated
|
|
| 168 |
# Prior constants — a small grounded set from reference papers.
|
| 169 |
# Candidates are consumed by the baselines; distractors appear in the papers
|
| 170 |
# but are not used by the shipped formulas.
|
| 171 |
-
priors:
|
|
|
|
| 172 |
# FM C12 fix (2026-05-26): removed all 5 leaking priors:
|
| 173 |
# - log_ML_intercept_bell_i_gr=-0.222 (Bell 2003 LAW_CONSTANTS a_i): EMPIRICAL_FIT
|
| 174 |
# - log_ML_slope_bell_i_gr=0.864 (Bell 2003 LAW_CONSTANTS b_i): EMPIRICAL_FIT
|
|
@@ -180,27 +181,27 @@ priors:
|
|
| 180 |
# Added solar_g_AB=5.11 and solar_r_AB=4.65 (Willmer 2018; not in any formula constant)
|
| 181 |
# as distractors that hint the solar magnitude scale without directly giving M_sun_i.
|
| 182 |
# Retained imf_offset_salpeter_to_kroupa=0.15 as a distractor (Bell 2003).
|
| 183 |
-
|
| 184 |
-
|
| 185 |
-
|
| 186 |
-
|
| 187 |
-
|
| 188 |
-
|
| 189 |
-
|
| 190 |
-
|
| 191 |
-
|
| 192 |
-
|
| 193 |
-
|
| 194 |
-
|
| 195 |
-
|
| 196 |
-
|
| 197 |
-
|
| 198 |
-
|
| 199 |
-
|
| 200 |
-
|
| 201 |
-
|
| 202 |
-
|
| 203 |
-
|
| 204 |
-
|
| 205 |
-
|
| 206 |
-
|
|
|
|
| 168 |
# Prior constants — a small grounded set from reference papers.
|
| 169 |
# Candidates are consumed by the baselines; distractors appear in the papers
|
| 170 |
# but are not used by the shipped formulas.
|
| 171 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 172 |
+
# priors:
|
| 173 |
# FM C12 fix (2026-05-26): removed all 5 leaking priors:
|
| 174 |
# - log_ML_intercept_bell_i_gr=-0.222 (Bell 2003 LAW_CONSTANTS a_i): EMPIRICAL_FIT
|
| 175 |
# - log_ML_slope_bell_i_gr=0.864 (Bell 2003 LAW_CONSTANTS b_i): EMPIRICAL_FIT
|
|
|
|
| 181 |
# Added solar_g_AB=5.11 and solar_r_AB=4.65 (Willmer 2018; not in any formula constant)
|
| 182 |
# as distractors that hint the solar magnitude scale without directly giving M_sun_i.
|
| 183 |
# Retained imf_offset_salpeter_to_kroupa=0.15 as a distractor (Bell 2003).
|
| 184 |
+
# - name: solar_g_AB
|
| 185 |
+
# value: 5.11
|
| 186 |
+
# unit: mag
|
| 187 |
+
# description: >-
|
| 188 |
+
# g-band AB solar absolute magnitude (Willmer 2018 Table 3). Used in g-band or
|
| 189 |
+
# g-r colour based mass-to-light estimates.
|
| 190 |
+
# source: "Willmer 2018, ApJS 236 47, Table 3."
|
| 191 |
+
# _role: distractor
|
| 192 |
+
# - name: solar_r_AB
|
| 193 |
+
# value: 4.65
|
| 194 |
+
# unit: mag
|
| 195 |
+
# description: >-
|
| 196 |
+
# r-band AB solar absolute magnitude (Willmer 2018 Table 3). Relevant for r-band
|
| 197 |
+
# mass-to-light calibrations.
|
| 198 |
+
# source: "Willmer 2018, ApJS 236 47, Table 3."
|
| 199 |
+
# _role: distractor
|
| 200 |
+
# - name: imf_offset_salpeter_to_kroupa
|
| 201 |
+
# value: 0.15
|
| 202 |
+
# unit: dex
|
| 203 |
+
# description: >-
|
| 204 |
+
# Logarithmic stellar-mass offset to convert diet-Salpeter IMF masses to
|
| 205 |
+
# Kroupa/Kennicutt IMF masses (subtract ~0.15 dex for a Kennicutt or Kroupa IMF).
|
| 206 |
+
# source: "Standard initial-mass-function conversion offset."
|
| 207 |
+
# _role: distractor
|
geomagnetic_storm_dst__dDst_dt/metadata.yaml
CHANGED
|
@@ -132,41 +132,42 @@ n_test: 3672
|
|
| 132 |
# so the all-distractor block is intentional, not an omission. See VERDICT.md.
|
| 133 |
# Leak re-check is against LAW_CONSTANTS only (priors<->OTHER_CONSTANTS overlap
|
| 134 |
# is normal design, not a leak); all three formulas have OTHER_CONSTANTS = {}.
|
| 135 |
-
priors:
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
|
| 141 |
-
|
| 142 |
-
|
| 143 |
-
|
| 144 |
-
|
| 145 |
-
|
| 146 |
-
|
| 147 |
-
|
| 148 |
-
|
| 149 |
-
|
| 150 |
-
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
|
| 154 |
-
|
| 155 |
-
|
| 156 |
-
|
| 157 |
-
|
| 158 |
-
|
| 159 |
-
|
| 160 |
-
|
| 161 |
-
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
|
| 166 |
-
|
| 167 |
-
|
| 168 |
-
|
| 169 |
-
|
|
|
|
| 170 |
|
| 171 |
references:
|
| 172 |
- id: obrien_2000
|
|
|
|
| 132 |
# so the all-distractor block is intentional, not an omission. See VERDICT.md.
|
| 133 |
# Leak re-check is against LAW_CONSTANTS only (priors<->OTHER_CONSTANTS overlap
|
| 134 |
# is normal design, not a leak); all three formulas have OTHER_CONSTANTS = {}.
|
| 135 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 136 |
+
# priors:
|
| 137 |
+
# - name: solar_wind_speed_typical_km_s
|
| 138 |
+
# value: 400.0
|
| 139 |
+
# unit: km/s
|
| 140 |
+
# description: >-
|
| 141 |
+
# Typical quiet-time solar wind bulk speed at 1 AU (~400 km/s).
|
| 142 |
+
# Characterises the solar wind environment.
|
| 143 |
+
# source: "Schwenn 2006 Living Rev. Sol. Phys. 3:2 (community reference value)"
|
| 144 |
+
# _role: distractor
|
| 145 |
+
# - name: Earth_magnetic_dipole_moment_A_m2
|
| 146 |
+
# value: 8.0e22
|
| 147 |
+
# unit: A*m^2
|
| 148 |
+
# description: >-
|
| 149 |
+
# Earth's geomagnetic dipole moment (~8e22 A·m²). Provides the geomagnetic
|
| 150 |
+
# context (ring-current injection scales with the magnetopause standoff
|
| 151 |
+
# distance which depends on the dipole moment).
|
| 152 |
+
# source: "Finlay et al. 2010, Geophys. J. Int. 183:1216 (IGRF-11 reference value)"
|
| 153 |
+
# _role: distractor
|
| 154 |
+
# - name: proton_charge_C
|
| 155 |
+
# value: 1.602176634e-19
|
| 156 |
+
# unit: C
|
| 157 |
+
# description: >-
|
| 158 |
+
# CODATA 2018 exact value of the elementary (proton) charge. Enters plasma
|
| 159 |
+
# physics formulas for solar-wind pressure and particle energy.
|
| 160 |
+
# source: "CODATA 2018 (SI defining constant, exact)"
|
| 161 |
+
# _role: distractor
|
| 162 |
+
# - name: kp_index_quiet
|
| 163 |
+
# value: 1.0
|
| 164 |
+
# unit: dimensionless
|
| 165 |
+
# description: >-
|
| 166 |
+
# Reference Kp value for geomagnetically quiet conditions (Kp ≤ 1).
|
| 167 |
+
# Kp is a different geomagnetic activity index (planetary K-index,
|
| 168 |
+
# 3-hourly, 0-9 scale).
|
| 169 |
+
# source: "Menvielle & Berthelier 1991 Rev. Geophys. 29:415 (Kp index definition)"
|
| 170 |
+
# _role: distractor
|
| 171 |
|
| 172 |
references:
|
| 173 |
- id: obrien_2000
|
gravity_wgs84_somigliana__g0/metadata.yaml
CHANGED
|
@@ -53,37 +53,38 @@ n_test: 4337
|
|
| 53 |
# Candidate prior constants — the `priors` prompt slot.
|
| 54 |
# Distractors are additional WGS84 constants that appear in the same tables
|
| 55 |
# but are NOT consumed by Eq. (4-1).
|
| 56 |
-
priors:
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
|
|
|
| 87 |
|
| 88 |
# Reference-baseline bank
|
| 89 |
references:
|
|
|
|
| 53 |
# Candidate prior constants — the `priors` prompt slot.
|
| 54 |
# Distractors are additional WGS84 constants that appear in the same tables
|
| 55 |
# but are NOT consumed by Eq. (4-1).
|
| 56 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 57 |
+
# priors:
|
| 58 |
+
# - name: g_standard
|
| 59 |
+
# value: 9.80665
|
| 60 |
+
# unit: m/s^2
|
| 61 |
+
# description: Standard acceleration of gravity (conventional standard adopted by BIPM/CGPM). Represents the average global sea-level gravity. A candidate value for a gravitational-acceleration scale constant.
|
| 62 |
+
# source: BIPM / CGPM Resolution (1901); NIST Reference on Constants, CODATA 2018 (physics.nist.gov/cuu/Constants/); numerical value defined exactly as 9.80665 m/s^2.
|
| 63 |
+
# _role: candidate
|
| 64 |
+
# - name: omega_earth_WGS84
|
| 65 |
+
# value: 7.292115e-5
|
| 66 |
+
# unit: rad/s
|
| 67 |
+
# description: Angular velocity of the Earth's rotation, one of the four WGS84 defining parameters.
|
| 68 |
+
# source: WGS84 defining-parameter set (value 7.292115 × 10^-5 rad/s).
|
| 69 |
+
# _role: candidate
|
| 70 |
+
# - name: a_WGS84
|
| 71 |
+
# value: 6378137.0
|
| 72 |
+
# unit: m
|
| 73 |
+
# description: Semi-major axis of the WGS84 reference ellipsoid — the equatorial radius. One of the four WGS84 defining parameters.
|
| 74 |
+
# source: WGS84 defining-parameter set.
|
| 75 |
+
# _role: candidate
|
| 76 |
+
# - name: gamma_p
|
| 77 |
+
# value: 9.8321849378
|
| 78 |
+
# unit: m/s^2
|
| 79 |
+
# description: Theoretical normal gravity at the WGS84 poles on the ellipsoid surface.
|
| 80 |
+
# source: WGS84 derived-constant set.
|
| 81 |
+
# _role: distractor
|
| 82 |
+
# - name: m_somigliana
|
| 83 |
+
# value: 0.00344978650684
|
| 84 |
+
# unit: dimensionless
|
| 85 |
+
# description: WGS84 constant m = omega^2 * a^2 * b / GM, appearing in geodetic derivations.
|
| 86 |
+
# source: WGS84 derived-constant set.
|
| 87 |
+
# _role: distractor
|
| 88 |
|
| 89 |
# Reference-baseline bank
|
| 90 |
references:
|
gutenberg_richter_b_value_usgs__gr_freq_mag/metadata.yaml
CHANGED
|
@@ -52,13 +52,14 @@ n_test: 17
|
|
| 52 |
# hanks_kanamori_1979 LAW b=0.75 (both are G-R slope b; empirical vs theoretical point
|
| 53 |
# estimates), FM-H1 semantic leak. log10_e=0.4343 retained as candidate (math constant
|
| 54 |
# in Aki MLE derivation, different physical quantity from b/m_t).
|
| 55 |
-
priors:
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
|
|
|
| 62 |
|
| 63 |
# Reference-baseline bank
|
| 64 |
references:
|
|
|
|
| 52 |
# hanks_kanamori_1979 LAW b=0.75 (both are G-R slope b; empirical vs theoretical point
|
| 53 |
# estimates), FM-H1 semantic leak. log10_e=0.4343 retained as candidate (math constant
|
| 54 |
# in Aki MLE derivation, different physical quantity from b/m_t).
|
| 55 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 56 |
+
# priors:
|
| 57 |
+
# - name: log10_e
|
| 58 |
+
# value: 0.4343
|
| 59 |
+
# unit: ""
|
| 60 |
+
# description: log10(e) — the base-10 logarithm of Euler's number.
|
| 61 |
+
# source: Mathematical constant.
|
| 62 |
+
# _role: candidate
|
| 63 |
|
| 64 |
# Reference-baseline bank
|
| 65 |
references:
|
hea_hardness_lattice_distortion_couzinie__HV/metadata.yaml
CHANGED
|
@@ -70,44 +70,45 @@ n_test: 14
|
|
| 70 |
# toda_caraballo baseline (missing PDF + non-published constants). Temesi 2023
|
| 71 |
# now beats naive (R² ~0.44).
|
| 72 |
|
| 73 |
-
priors:
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
|
|
|
| 111 |
|
| 112 |
references:
|
| 113 |
- id: temesi_2023_hardness_bonding
|
|
|
|
| 70 |
# toda_caraballo baseline (missing PDF + non-published constants). Temesi 2023
|
| 71 |
# now beats naive (R² ~0.44).
|
| 72 |
|
| 73 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 74 |
+
# priors:
|
| 75 |
+
# - name: "K_SSH"
|
| 76 |
+
# value: 0.7
|
| 77 |
+
# unit: ""
|
| 78 |
+
# description: >-
|
| 79 |
+
# A candidate proportionality constant for a solid-solution hardening
|
| 80 |
+
# contribution, fitted empirically over a high-entropy-alloy dataset when
|
| 81 |
+
# shear modulus is in GPa and hardness in kgf/mm².
|
| 82 |
+
# source: "Solid-solution hardening literature for high-entropy alloys."
|
| 83 |
+
# _role: distractor
|
| 84 |
+
# - name: "HV_base"
|
| 85 |
+
# value: 60.0
|
| 86 |
+
# unit: "kgf/mm²"
|
| 87 |
+
# description: >-
|
| 88 |
+
# A candidate baseline hardness term representing a rule-of-mixture
|
| 89 |
+
# constituent hardness floor, empirically 60 ± 20 kgf/mm² for a
|
| 90 |
+
# high-entropy-alloy dataset.
|
| 91 |
+
# source: "Solid-solution hardening literature for high-entropy alloys."
|
| 92 |
+
# _role: distractor
|
| 93 |
+
# - name: "VEC_BCC_threshold"
|
| 94 |
+
# value: 6.87
|
| 95 |
+
# unit: ""
|
| 96 |
+
# description: >-
|
| 97 |
+
# Critical VEC below which BCC structure is preferred in high-entropy alloys,
|
| 98 |
+
# empirically established from the Guo et al. (2011) structural-stability
|
| 99 |
+
# criterion for multi-component alloys. VEC < 6.87 favours BCC; VEC > 8 favours
|
| 100 |
+
# FCC.
|
| 101 |
+
# source: "Guo S et al. (2011) J. Appl. Phys. 109:103505, Table 2 — BCC/FCC boundary criterion."
|
| 102 |
+
# _role: distractor
|
| 103 |
+
# - name: "n_SSH_Labusch"
|
| 104 |
+
# value: 0.667
|
| 105 |
+
# unit: ""
|
| 106 |
+
# description: >-
|
| 107 |
+
# Labusch-model solid-solution hardening exponent (2/3) for the concentration
|
| 108 |
+
# dependence of yield stress in dilute alloys: Δσ ∝ c^(2/3). In HEA literature
|
| 109 |
+
# this exponent is sometimes replaced by 0.5 (Fleischer model).
|
| 110 |
+
# source: "Labusch R (1970) Phys. Status Solidi 41:659-669; see also Varvenne et al. (2016) Acta Mater. 124:660."
|
| 111 |
+
# _role: distractor
|
| 112 |
|
| 113 |
references:
|
| 114 |
- id: temesi_2023_hardness_bonding
|
her_hor_pt_acid_kucernak__current_density/metadata.yaml
CHANGED
|
@@ -74,37 +74,38 @@ n_test: 334
|
|
| 74 |
# formula consumes, plus distractors. Audit 2026-06-02: a priors block was added
|
| 75 |
# (was missing, H3). LAW ∩ priors = ∅ (the intrinsic kinetic constants are never
|
| 76 |
# exposed); OTHER ⊆ priors.
|
| 77 |
-
priors:
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
|
|
|
| 108 |
|
| 109 |
# === Reference baselines ===
|
| 110 |
# Two formulas from Kucernak & Zalitis (2016): the full Heyrovsky-Volmer (HV)
|
|
|
|
| 74 |
# formula consumes, plus distractors. Audit 2026-06-02: a priors block was added
|
| 75 |
# (was missing, H3). LAW ∩ priors = ∅ (the intrinsic kinetic constants are never
|
| 76 |
# exposed); OTHER ⊆ priors.
|
| 77 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 78 |
+
# priors:
|
| 79 |
+
# - name: faraday_constant
|
| 80 |
+
# value: 96485.0
|
| 81 |
+
# unit: C mol-1
|
| 82 |
+
# description: Faraday constant F — charge per mole of electrons.
|
| 83 |
+
# source: NIST CODATA 2018.
|
| 84 |
+
# _role: candidate
|
| 85 |
+
# - name: gas_constant
|
| 86 |
+
# value: 8.314
|
| 87 |
+
# unit: J mol-1 K-1
|
| 88 |
+
# description: Molar gas constant R.
|
| 89 |
+
# source: NIST CODATA 2018.
|
| 90 |
+
# _role: candidate
|
| 91 |
+
# - name: ref_temperature
|
| 92 |
+
# value: 298.15
|
| 93 |
+
# unit: K
|
| 94 |
+
# description: Reference temperature T_ref; 25 °C standard.
|
| 95 |
+
# source: Standard reference temperature (25 °C).
|
| 96 |
+
# _role: candidate
|
| 97 |
+
# - name: elementary_charge
|
| 98 |
+
# value: 1.602176634e-19
|
| 99 |
+
# unit: C
|
| 100 |
+
# description: Elementary charge e — a fundamental constant (F = N_A·e).
|
| 101 |
+
# source: NIST CODATA 2018.
|
| 102 |
+
# _role: distractor
|
| 103 |
+
# - name: avogadro_constant
|
| 104 |
+
# value: 6.02214076e23
|
| 105 |
+
# unit: mol-1
|
| 106 |
+
# description: Avogadro constant N_A — fundamental constant (F = N_A·e).
|
| 107 |
+
# source: NIST CODATA 2018.
|
| 108 |
+
# _role: distractor
|
| 109 |
|
| 110 |
# === Reference baselines ===
|
| 111 |
# Two formulas from Kucernak & Zalitis (2016): the full Heyrovsky-Volmer (HV)
|
life_expectancy_preston__e0_vs_gdppc/metadata.yaml
CHANGED
|
@@ -55,33 +55,35 @@ n_test: 19
|
|
| 55 |
# is the fitted kappa=72.67 != 71.5), so there is no C12/H1 LAW-in-priors leak.
|
| 56 |
# kappa is a globally-fitted defining coefficient (Type I, no per-group fit), not a
|
| 57 |
# LOCAL_FITTABLE parameter.
|
| 58 |
-
priors:
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
|
|
|
|
|
|
| 85 |
|
| 86 |
references:
|
| 87 |
- id: preston_1975_logistic
|
|
|
|
| 55 |
# is the fitted kappa=72.67 != 71.5), so there is no C12/H1 LAW-in-priors leak.
|
| 56 |
# kappa is a globally-fitted defining coefficient (Type I, no per-group fit), not a
|
| 57 |
# LOCAL_FITTABLE parameter.
|
| 58 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 59 |
+
# priors:
|
| 60 |
+
# [LEAK-REMOVE 2026-06-08] value = fitted answer constant; commented out (NOT deleted).
|
| 61 |
+
# - name: kappa_1960s
|
| 62 |
+
# value: 71.5
|
| 63 |
+
# unit: years
|
| 64 |
+
# description: >-
|
| 65 |
+
# Practical upper asymptote of life expectancy for the 1960s cross-section
|
| 66 |
+
# (the biologically achievable ceiling of e0 through income gains alone in
|
| 67 |
+
# that decade). A candidate value for an upper-ceiling role.
|
| 68 |
+
# source: Published demographic literature for the 1960s cross-section.
|
| 69 |
+
# _role: candidate
|
| 70 |
+
# - name: kappa_1930s
|
| 71 |
+
# value: 66.8
|
| 72 |
+
# unit: years
|
| 73 |
+
# description: >-
|
| 74 |
+
# Practical upper asymptote of life expectancy for the 1930s cross-section.
|
| 75 |
+
# A real published value but for a different decade than this 1960s
|
| 76 |
+
# cross-section.
|
| 77 |
+
# source: Published demographic literature for the 1930s cross-section.
|
| 78 |
+
# _role: distractor
|
| 79 |
+
# - name: r_log_linear_1960s
|
| 80 |
+
# value: 0.880
|
| 81 |
+
# unit: ""
|
| 82 |
+
# description: >-
|
| 83 |
+
# A published goodness-of-fit summary statistic (Pearson correlation
|
| 84 |
+
# coefficient) for the 1960s cross-section.
|
| 85 |
+
# source: Published demographic literature for the 1960s cross-section.
|
| 86 |
+
# _role: distractor
|
| 87 |
|
| 88 |
references:
|
| 89 |
- id: preston_1975_logistic
|
liion_capacity_fade_nasa_pcoe__capacity_Ah/metadata.yaml
CHANGED
|
@@ -43,25 +43,26 @@ n_train: 382
|
|
| 43 |
n_test: 254
|
| 44 |
|
| 45 |
# Candidate prior constants — the `priors` prompt slot.
|
| 46 |
-
priors:
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
|
|
|
| 65 |
|
| 66 |
# === Reference baselines ===
|
| 67 |
references:
|
|
|
|
| 43 |
n_test: 254
|
| 44 |
|
| 45 |
# Candidate prior constants — the `priors` prompt slot.
|
| 46 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 47 |
+
# priors:
|
| 48 |
+
# - name: parabolic_growth_exponent
|
| 49 |
+
# value: 0.5
|
| 50 |
+
# unit: dimensionless
|
| 51 |
+
# description: Parabolic growth exponent for diffusion-limited film thickness.
|
| 52 |
+
# source: Standard kinetic-theory result (Tammann 1920, Wagner 1933) — diffusion-limited growth of an oxide / SEI layer scales as sqrt(t) because the rate-limiting step is diffusion through an increasingly thick layer. A candidate value for a cycle-fade exponent.
|
| 53 |
+
# _role: candidate
|
| 54 |
+
# - name: standard_li_18650_nominal_capacity
|
| 55 |
+
# value: 2.0
|
| 56 |
+
# unit: Ah
|
| 57 |
+
# description: Nominal capacity of a typical Li-ion 18650 cell.
|
| 58 |
+
# source: NASA PCoE B0005-B0007 cell datasheets; the manufacturer-rated capacity for these specific 18650 cells. A candidate value for an initial-capacity constant.
|
| 59 |
+
# _role: candidate
|
| 60 |
+
# - name: lead_acid_capacity_fade_rate_typical
|
| 61 |
+
# value: 0.005
|
| 62 |
+
# unit: Ah/cycle
|
| 63 |
+
# description: A typical linear capacity-fade rate for lead-acid batteries.
|
| 64 |
+
# source: A different battery chemistry — lead-acid batteries fade approximately linearly with cycles.
|
| 65 |
+
# _role: distractor
|
| 66 |
|
| 67 |
# === Reference baselines ===
|
| 68 |
references:
|
liquid_dispersion_glycols__refractive_index/metadata.yaml
CHANGED
|
@@ -70,19 +70,20 @@ references:
|
|
| 70 |
# jakubczyk_2023.OTHER_CONSTANTS (was a hidden `_T_REF` module global; n_other_constants 0->1). It is a
|
| 71 |
# structural protocol value, not a discoverable law, so it is exposed below as the
|
| 72 |
# `reference_temperature_sellmeier` candidate prior (OTHER_CONSTANTS->priors; not a leak vs LAW_CONSTANTS).
|
| 73 |
-
priors:
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
|
|
|
| 86 |
|
| 87 |
# === v2 anti-dump caps ===
|
| 88 |
caps:
|
|
|
|
| 70 |
# jakubczyk_2023.OTHER_CONSTANTS (was a hidden `_T_REF` module global; n_other_constants 0->1). It is a
|
| 71 |
# structural protocol value, not a discoverable law, so it is exposed below as the
|
| 72 |
# `reference_temperature_sellmeier` candidate prior (OTHER_CONSTANTS->priors; not a leak vs LAW_CONSTANTS).
|
| 73 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 74 |
+
# priors:
|
| 75 |
+
# - name: reference_temperature_sellmeier
|
| 76 |
+
# value: 20.0
|
| 77 |
+
# unit: degC
|
| 78 |
+
# description: Reference temperature at which the dispersion fit is defined.
|
| 79 |
+
# source: Jakubczyk et al. 2023, experimental protocol description, PDF p. 3.
|
| 80 |
+
# _role: candidate
|
| 81 |
+
# - name: nm_per_um_conversion
|
| 82 |
+
# value: 1000.0
|
| 83 |
+
# unit: nm/um
|
| 84 |
+
# description: Unit conversion between nanometres and micrometres.
|
| 85 |
+
# source: SI unit definition.
|
| 86 |
+
# _role: distractor
|
| 87 |
|
| 88 |
# === v2 anti-dump caps ===
|
| 89 |
caps:
|
lpbf_meltpool_rosenthal_inconel__depth/metadata.yaml
CHANGED
|
@@ -92,31 +92,32 @@ references:
|
|
| 92 |
# capacity in the Rosenthal Eq.6 denominator, i.e. a semantic leak of the LAW combination.
|
| 93 |
# Live priors are now 4: 1 candidate (euler_number) + 3 distractors (Tm_iron_K,
|
| 94 |
# thermal_diff_water_m2_s, stefan_boltzmann_W_m2_K4).
|
| 95 |
-
priors:
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
|
|
|
| 120 |
|
| 121 |
caps:
|
| 122 |
max_law_constants: 6
|
|
|
|
| 92 |
# capacity in the Rosenthal Eq.6 denominator, i.e. a semantic leak of the LAW combination.
|
| 93 |
# Live priors are now 4: 1 candidate (euler_number) + 3 distractors (Tm_iron_K,
|
| 94 |
# thermal_diff_water_m2_s, stefan_boltzmann_W_m2_K4).
|
| 95 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 96 |
+
# priors:
|
| 97 |
+
# - name: euler_number
|
| 98 |
+
# value: 2.71828182845904523536
|
| 99 |
+
# unit: dimensionless
|
| 100 |
+
# description: Euler's number e — universal mathematical constant. This is NOT an alloy-specific fit coefficient; it is the base of the natural logarithm, fixed by mathematics.
|
| 101 |
+
# source: Mathematical constant; standard heat-conduction point-source theory.
|
| 102 |
+
# _role: candidate
|
| 103 |
+
# - name: Tm_iron_K
|
| 104 |
+
# value: 1811.0
|
| 105 |
+
# unit: K
|
| 106 |
+
# description: Melting point of pure iron — a reference scale for iron-based alloys.
|
| 107 |
+
# source: NIST WebBook / CODATA standard value for pure Fe.
|
| 108 |
+
# _role: distractor
|
| 109 |
+
# - name: thermal_diff_water_m2_s
|
| 110 |
+
# value: 1.43e-7
|
| 111 |
+
# unit: m^2/s
|
| 112 |
+
# description: Thermal diffusivity of water at room temperature — reference value for comparison; several orders of magnitude lower than metal alloys.
|
| 113 |
+
# source: Standard thermophysical data; CRC Handbook of Chemistry and Physics.
|
| 114 |
+
# _role: distractor
|
| 115 |
+
# - name: stefan_boltzmann_W_m2_K4
|
| 116 |
+
# value: 5.670374419e-8
|
| 117 |
+
# unit: W/(m^2 K^4)
|
| 118 |
+
# description: Stefan-Boltzmann radiation constant sigma — relevant to radiative heat loss at high temperatures.
|
| 119 |
+
# source: CODATA 2018; exact value in SI.
|
| 120 |
+
# _role: distractor
|
| 121 |
|
| 122 |
caps:
|
| 123 |
max_law_constants: 6
|
magnet_3c90_core_loss__log_P_vol/metadata.yaml
CHANGED
|
@@ -55,25 +55,26 @@ n_train: 3444
|
|
| 55 |
n_test: 769
|
| 56 |
|
| 57 |
# Candidate prior constants — the `priors` prompt slot.
|
| 58 |
-
priors:
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
|
|
|
| 77 |
|
| 78 |
# === Reference baselines ===
|
| 79 |
references:
|
|
|
|
| 55 |
n_test: 769
|
| 56 |
|
| 57 |
# Candidate prior constants — the `priors` prompt slot.
|
| 58 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 59 |
+
# priors:
|
| 60 |
+
# - name: bertotti_hysteresis_exponent
|
| 61 |
+
# value: 2.0
|
| 62 |
+
# unit: dimensionless
|
| 63 |
+
# description: Geometric B-exponent expected from a fixed-shape hysteresis loop.
|
| 64 |
+
# source: Derived from the geometry of B-H loop area integration assuming a fixed loop shape. A candidate value for the BETA exponent.
|
| 65 |
+
# _role: candidate
|
| 66 |
+
# - name: steinmetz_frequency_exponent_typical_ferrite
|
| 67 |
+
# value: 1.6
|
| 68 |
+
# unit: dimensionless
|
| 69 |
+
# description: A typical frequency exponent for a soft NiZn or MnZn ferrite in the kHz-MHz band.
|
| 70 |
+
# source: Power-electronics textbook value. A candidate value for the ALPHA exponent.
|
| 71 |
+
# _role: candidate
|
| 72 |
+
# - name: copper_resistivity_room_temperature
|
| 73 |
+
# value: 1.68e-08
|
| 74 |
+
# unit: ohm·m
|
| 75 |
+
# description: Electrical resistivity of pure copper at 25°C.
|
| 76 |
+
# source: NIST handbook. Domain-adjacent (relevant to winding losses, not core losses).
|
| 77 |
+
# _role: distractor
|
| 78 |
|
| 79 |
# === Reference baselines ===
|
| 80 |
references:
|
mars_crater_size_frequency__N_cum/metadata.yaml
CHANGED
|
@@ -73,31 +73,33 @@ n_train: 56
|
|
| 73 |
n_test: 18
|
| 74 |
|
| 75 |
# Candidate prior constants — the `priors` prompt slot.
|
| 76 |
-
priors:
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
|
|
|
|
|
|
| 101 |
|
| 102 |
# === Reference baselines ===
|
| 103 |
references:
|
|
|
|
| 73 |
n_test: 18
|
| 74 |
|
| 75 |
# Candidate prior constants — the `priors` prompt slot.
|
| 76 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 77 |
+
# priors:
|
| 78 |
+
# - name: neukum_npf_slope_at_1km
|
| 79 |
+
# value: -1.8
|
| 80 |
+
# unit: dimensionless
|
| 81 |
+
# description: Approximate cumulative-SFD slope near D = 1 km; a candidate value for the small-crater power-law exponent.
|
| 82 |
+
# source: Published planetary-science value — the cumulative-SFD slope near D = 1 km is roughly -1.8 to -2.0. A candidate exponent for the small-crater regime.
|
| 83 |
+
# _role: candidate
|
| 84 |
+
# [LEAK-REMOVE 2026-06-08] value = fitted answer constant; commented out (NOT deleted).
|
| 85 |
+
# - name: large_crater_branch_slope
|
| 86 |
+
# value: -2.6
|
| 87 |
+
# unit: dimensionless
|
| 88 |
+
# description: Approximate cumulative-SFD slope in the large-crater / basin regime (D > 30 km); a candidate value for the steep large-crater exponent.
|
| 89 |
+
# source: Published planetary-science value — the SFD steepens to slope ~ -2.5 to -3 in the large-crater regime. A candidate exponent for the large-crater regime.
|
| 90 |
+
# _role: candidate
|
| 91 |
+
# - name: simple_to_complex_transition_diameter
|
| 92 |
+
# value: 7.0
|
| 93 |
+
# unit: km
|
| 94 |
+
# description: Diameter of the simple-to-complex crater morphology transition on Mars; a candidate diameter scale.
|
| 95 |
+
# source: Published planetary-science value — Mars simple-to-complex transition D* ~ 7 km. A candidate diameter scale.
|
| 96 |
+
# _role: candidate
|
| 97 |
+
# - name: mars_surface_area_km2
|
| 98 |
+
# value: 144798500.0
|
| 99 |
+
# unit: km^2
|
| 100 |
+
# description: Total surface area of Mars, used to convert crater counts to areal densities.
|
| 101 |
+
# source: NASA Mars Fact Sheet. Used in data preparation (N_cum = count / area).
|
| 102 |
+
# _role: distractor
|
| 103 |
|
| 104 |
# === Reference baselines ===
|
| 105 |
references:
|
materials_project_elastic_birch_murnaghan__K_VRH/metadata.yaml
CHANGED
|
@@ -73,31 +73,34 @@ n_test: 177
|
|
| 73 |
# (3) The hf_new_source dejong_2015 "K_VRH = (K_Voigt + K_Reuss)/2" formula was correctly
|
| 74 |
# NOT adopted -- it is the VRH definitional identity (a tautology).
|
| 75 |
|
| 76 |
-
priors:
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
|
|
|
|
|
|
|
|
|
| 101 |
|
| 102 |
references:
|
| 103 |
- id: cohen_1985
|
|
|
|
| 73 |
# (3) The hf_new_source dejong_2015 "K_VRH = (K_Voigt + K_Reuss)/2" formula was correctly
|
| 74 |
# NOT adopted -- it is the VRH definitional identity (a tautology).
|
| 75 |
|
| 76 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 77 |
+
# priors:
|
| 78 |
+
# [LEAK-REMOVE 2026-06-08] value = fitted answer constant; commented out (NOT deleted).
|
| 79 |
+
# - name: cohen_bulk_exponent
|
| 80 |
+
# value: -1.1667
|
| 81 |
+
# unit: dimensionless
|
| 82 |
+
# description: A candidate volume exponent for a bulk-modulus-versus-atomic-volume scaling of covalent solids.
|
| 83 |
+
# source: Published solid-state-physics value for a covalent-solid bulk-modulus scaling.
|
| 84 |
+
# _role: candidate
|
| 85 |
+
# [LEAK-REMOVE 2026-06-08] value = fitted answer constant; commented out (NOT deleted).
|
| 86 |
+
# - name: anderson_nafe_exponent
|
| 87 |
+
# value: -1.3333
|
| 88 |
+
# unit: dimensionless
|
| 89 |
+
# description: A candidate slope for a log-linear bulk-modulus-versus-atomic-volume relation of oxide isoelectronic families.
|
| 90 |
+
# source: Published geophysics value for an oxide bulk-modulus log-volume relation.
|
| 91 |
+
# _role: candidate
|
| 92 |
+
# - name: birch_murnaghan_exponent_73
|
| 93 |
+
# value: 2.3333
|
| 94 |
+
# unit: dimensionless
|
| 95 |
+
# description: 7/3 exponent appearing in a third-order finite-strain equation of state.
|
| 96 |
+
# source: Standard finite-strain equation-of-state theory.
|
| 97 |
+
# _role: distractor
|
| 98 |
+
# - name: murnaghan_k0prime_typical
|
| 99 |
+
# value: 4.0
|
| 100 |
+
# unit: dimensionless
|
| 101 |
+
# description: Typical value of K0-prime (pressure derivative of bulk modulus at zero pressure) for many ionic and covalent solids, used in equation-of-state fits.
|
| 102 |
+
# source: Standard equation-of-state literature value.
|
| 103 |
+
# _role: distractor
|
| 104 |
|
| 105 |
references:
|
| 106 |
- id: cohen_1985
|
mauna_loa_co2_keeling_curve_noaa__co2_ppm/metadata.yaml
CHANGED
|
@@ -76,19 +76,20 @@ n_train: 742
|
|
| 76 |
n_test: 76
|
| 77 |
|
| 78 |
# Candidate prior constants — the `priors` prompt slot.
|
| 79 |
-
priors:
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
|
|
|
| 92 |
|
| 93 |
# === Reference baselines ===
|
| 94 |
references:
|
|
|
|
| 76 |
n_test: 76
|
| 77 |
|
| 78 |
# Candidate prior constants — the `priors` prompt slot.
|
| 79 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 80 |
+
# priors:
|
| 81 |
+
# - name: keeling_seasonal_amplitude_first_harmonic
|
| 82 |
+
# value: 3.0
|
| 83 |
+
# unit: ppm
|
| 84 |
+
# description: Mauna Loa first-harmonic seasonal-cycle peak-to-peak amplitude.
|
| 85 |
+
# source: Published empirical first-harmonic seasonal-cycle amplitude at Mauna Loa. A candidate value for the seasonal-cycle amplitude.
|
| 86 |
+
# _role: candidate
|
| 87 |
+
# - name: ipcc_ar6_co2_radiative_efficiency
|
| 88 |
+
# value: 1.37e-5
|
| 89 |
+
# unit: W/m^2/ppm
|
| 90 |
+
# description: IPCC AR6 stratospherically-adjusted radiative forcing efficiency of CO2.
|
| 91 |
+
# source: IPCC AR6 WG1 Table 7.8 (Forster et al. 2021). Domain-adjacent (climate sensitivity); operates on radiative forcing rather than ppm directly.
|
| 92 |
+
# _role: distractor
|
| 93 |
|
| 94 |
# === Reference baselines ===
|
| 95 |
references:
|
metabolic_scaling_anage__log_bmr_W/metadata.yaml
CHANGED
|
@@ -43,25 +43,28 @@ n_train: 238
|
|
| 43 |
n_test: 158
|
| 44 |
|
| 45 |
# Candidate prior constants — the `priors` prompt slot.
|
| 46 |
-
priors:
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
|
|
|
|
|
|
|
|
|
| 65 |
|
| 66 |
# === Reference baselines ===
|
| 67 |
references:
|
|
|
|
| 43 |
n_test: 158
|
| 44 |
|
| 45 |
# Candidate prior constants — the `priors` prompt slot.
|
| 46 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 47 |
+
# priors:
|
| 48 |
+
# [LEAK-REMOVE 2026-06-08] value = fitted answer constant; commented out (NOT deleted).
|
| 49 |
+
# - name: kleiber_exponent_three_quarters
|
| 50 |
+
# value: 0.75
|
| 51 |
+
# unit: dimensionless
|
| 52 |
+
# description: A 3/4-power metabolic scaling exponent.
|
| 53 |
+
# source: Classic empirical fit across mammalian species groups; subsequently derived from fractal-network theory. A candidate value for a metabolic-mass scaling exponent.
|
| 54 |
+
# _role: candidate
|
| 55 |
+
# [LEAK-REMOVE 2026-06-08] value = fitted answer constant; commented out (NOT deleted).
|
| 56 |
+
# - name: rubner_surface_exponent_two_thirds
|
| 57 |
+
# value: 0.6667
|
| 58 |
+
# unit: dimensionless
|
| 59 |
+
# description: A 2/3-power surface-area scaling exponent.
|
| 60 |
+
# source: Classic surface-area scaling argument for geometrically similar animals. A candidate value for a metabolic-mass scaling exponent.
|
| 61 |
+
# _role: candidate
|
| 62 |
+
# - name: stefan_boltzmann_radiation_constant
|
| 63 |
+
# value: 5.67e-8
|
| 64 |
+
# unit: W / (m^2 K^4)
|
| 65 |
+
# description: Stefan-Boltzmann constant for blackbody radiation.
|
| 66 |
+
# source: Universal physics constant; radiation-based heat transfer relates to body surface but the relevant scaling here is metabolic generation, not blackbody emission.
|
| 67 |
+
# _role: distractor
|
| 68 |
|
| 69 |
# === Reference baselines ===
|
| 70 |
references:
|
mincer_earnings_acs_pums__log_wage/metadata.yaml
CHANGED
|
@@ -55,31 +55,32 @@ n_test: 26713
|
|
| 55 |
# superseded). The four priors are 1990/1960 Census Table-2 (PDF p.156) cohort
|
| 56 |
# values — DISTRACTORS that differ from every train-fit LAW value (LAW ∩ priors = ∅,
|
| 57 |
# verified 2026-05-30), so no leak. Distractor role unchanged.
|
| 58 |
-
priors:
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
|
|
|
| 83 |
|
| 84 |
references:
|
| 85 |
- id: heckman_2005
|
|
|
|
| 55 |
# superseded). The four priors are 1990/1960 Census Table-2 (PDF p.156) cohort
|
| 56 |
# values — DISTRACTORS that differ from every train-fit LAW value (LAW ∩ priors = ∅,
|
| 57 |
# verified 2026-05-30), so no leak. Distractor role unchanged.
|
| 58 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 59 |
+
# priors:
|
| 60 |
+
# - name: schooling_return_1990
|
| 61 |
+
# value: 0.1292
|
| 62 |
+
# unit: per year
|
| 63 |
+
# description: Schooling coefficient (rho_s) from OLS on 1990 Census white men.
|
| 64 |
+
# source: "Heckman, Lochner & Todd (2005), Table 2, PDF p. 156."
|
| 65 |
+
# _role: distractor
|
| 66 |
+
# - name: experience_linear_1990
|
| 67 |
+
# value: 0.1301
|
| 68 |
+
# unit: per year
|
| 69 |
+
# description: An experience-related earnings coefficient from OLS on 1990 Census white men.
|
| 70 |
+
# source: "Heckman, Lochner & Todd (2005), Table 2, PDF p. 156."
|
| 71 |
+
# _role: distractor
|
| 72 |
+
# - name: experience_squared_1990
|
| 73 |
+
# value: -0.0023
|
| 74 |
+
# unit: per year squared
|
| 75 |
+
# description: An experience-related earnings coefficient from OLS on 1990 Census white men.
|
| 76 |
+
# source: "Heckman, Lochner & Todd (2005), Table 2, PDF p. 156."
|
| 77 |
+
# _role: distractor
|
| 78 |
+
# - name: schooling_return_1960
|
| 79 |
+
# value: 0.1152
|
| 80 |
+
# unit: per year
|
| 81 |
+
# description: Schooling coefficient from OLS on 1960 Census white men — a lower-premium era.
|
| 82 |
+
# source: "Heckman, Lochner & Todd (2005), Table 2, PDF p. 156."
|
| 83 |
+
# _role: distractor
|
| 84 |
|
| 85 |
references:
|
| 86 |
- id: heckman_2005
|
neo_size_frequency_distribution__N_cum_H/metadata.yaml
CHANGED
|
@@ -52,25 +52,27 @@ n_train: 57
|
|
| 52 |
n_test: 14
|
| 53 |
|
| 54 |
# Candidate prior constants — the `priors` prompt slot.
|
| 55 |
-
priors:
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
|
|
|
|
|
|
| 74 |
|
| 75 |
# === Reference baselines ===
|
| 76 |
references:
|
|
|
|
| 52 |
n_test: 14
|
| 53 |
|
| 54 |
# Candidate prior constants — the `priors` prompt slot.
|
| 55 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 56 |
+
# priors:
|
| 57 |
+
# [LEAK-REMOVE 2026-06-08] value = fitted answer constant; commented out (NOT deleted).
|
| 58 |
+
# - name: dohnanyi_magnitude_slope
|
| 59 |
+
# value: 0.5
|
| 60 |
+
# unit: dimensionless
|
| 61 |
+
# description: Collisional-equilibrium cumulative magnitude slope; a candidate value for the SFD exponent.
|
| 62 |
+
# source: Classic collisional-equilibrium theory of small-body populations. A candidate value for the SFD magnitude slope.
|
| 63 |
+
# _role: candidate
|
| 64 |
+
# - name: bottke_2002_break_diameter
|
| 65 |
+
# value: 100.0
|
| 66 |
+
# unit: m
|
| 67 |
+
# description: Approximate diameter (~H 22) where the debiased NEO SFD shows a slope change / "bump"; a candidate diameter scale.
|
| 68 |
+
# source: Published debiased NEO size-distribution literature. This feature is FAINTER than the H<18 window of this task.
|
| 69 |
+
# _role: distractor
|
| 70 |
+
# - name: spaceguard_completeness_H
|
| 71 |
+
# value: 17.75
|
| 72 |
+
# unit: mag
|
| 73 |
+
# description: Absolute magnitude corresponding to ~1 km diameter, the Spaceguard survey completeness goal.
|
| 74 |
+
# source: NASA Spaceguard mandate. Marks the bright end of high completeness.
|
| 75 |
+
# _role: candidate
|
| 76 |
|
| 77 |
# === Reference baselines ===
|
| 78 |
references:
|
nuclear_binding_energy_ame2020__BE_per_A/metadata.yaml
CHANGED
|
@@ -120,37 +120,38 @@ references:
|
|
| 120 |
# genuine nuclear-physics constants traceable to the reference papers, mixed
|
| 121 |
# with distractors from unrelated physics. The SR system must select which
|
| 122 |
# are relevant. `_role` is audit-only (stripped before the priors are shown).
|
| 123 |
-
priors:
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
|
| 129 |
-
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
|
| 133 |
-
|
| 134 |
-
|
| 135 |
-
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
|
| 141 |
-
|
| 142 |
-
|
| 143 |
-
|
| 144 |
-
|
| 145 |
-
|
| 146 |
-
|
| 147 |
-
|
| 148 |
-
|
| 149 |
-
|
| 150 |
-
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
|
|
|
|
| 154 |
|
| 155 |
# === v2 anti-dump caps ===
|
| 156 |
caps:
|
|
@@ -181,4 +182,4 @@ validity_rubrics:
|
|
| 181 |
|
| 182 |
# === Scoring ===
|
| 183 |
metric: rmse # RMS error on B/A — canonical for nuclear mass models
|
| 184 |
-
best_baseline: 0.0145 # wang_2014 rmse on data/test.csv
|
|
|
|
| 120 |
# genuine nuclear-physics constants traceable to the reference papers, mixed
|
| 121 |
# with distractors from unrelated physics. The SR system must select which
|
| 122 |
# are relevant. `_role` is audit-only (stripped before the priors are shown).
|
| 123 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 124 |
+
# priors:
|
| 125 |
+
# - name: e2_coulomb
|
| 126 |
+
# value: 1.4399764
|
| 127 |
+
# unit: MeV*fm
|
| 128 |
+
# description: Coulomb coupling e^2 — electrostatic energy of two unit charges 1 fm apart.
|
| 129 |
+
# source: CODATA; appears in the electrostatic energy of nuclear mass models.
|
| 130 |
+
# _role: candidate
|
| 131 |
+
# - name: r0_nuclear_radius
|
| 132 |
+
# value: 1.16
|
| 133 |
+
# unit: fm
|
| 134 |
+
# description: Nuclear-radius constant in R = r0 * A^(1/3).
|
| 135 |
+
# source: Standard nuclear-physics value (1.16 fm).
|
| 136 |
+
# _role: candidate
|
| 137 |
+
# - name: magic_number_N126
|
| 138 |
+
# value: 126
|
| 139 |
+
# unit: dimensionless
|
| 140 |
+
# description: A nuclear shell-closure ("magic") nucleon number.
|
| 141 |
+
# source: AME2020 evaluation and the mass-model literature — shell closures at N/Z = 2, 8, 20, 28, 50, 82, 126.
|
| 142 |
+
# _role: distractor
|
| 143 |
+
# - name: compressibility_const
|
| 144 |
+
# value: 60.0
|
| 145 |
+
# unit: MeV
|
| 146 |
+
# description: Pre-exponential constant of the nuclear-compressibility correction term.
|
| 147 |
+
# source: Nuclear mass-model constant table.
|
| 148 |
+
# _role: distractor
|
| 149 |
+
# - name: spin_orbit_strength
|
| 150 |
+
# value: 26.4796
|
| 151 |
+
# unit: dimensionless
|
| 152 |
+
# description: Spin-orbit strength parameter of a Woods-Saxon single-particle potential.
|
| 153 |
+
# source: Woods-Saxon single-particle potential parameter table.
|
| 154 |
+
# _role: distractor
|
| 155 |
|
| 156 |
# === v2 anti-dump caps ===
|
| 157 |
caps:
|
|
|
|
| 182 |
|
| 183 |
# === Scoring ===
|
| 184 |
metric: rmse # RMS error on B/A — canonical for nuclear mass models
|
| 185 |
+
best_baseline: 0.0145 # wang_2014 rmse on data/test.csv
|
ocean_mesoscale_eddy_closure__Sx/metadata.yaml
CHANGED
|
@@ -102,31 +102,32 @@ n_test: 14336
|
|
| 102 |
# purely data-driven ML weights with no upstream physical anchor — Strategy B
|
| 103 |
# (remove). w4_zb2020 was already the sole distractor and stays; three more
|
| 104 |
# domain-relevant distractors added to reach the required 2-4 total.
|
| 105 |
-
priors:
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
|
| 129 |
-
|
|
|
|
| 130 |
|
| 131 |
# Reference-baseline bank
|
| 132 |
references:
|
|
|
|
| 102 |
# purely data-driven ML weights with no upstream physical anchor — Strategy B
|
| 103 |
# (remove). w4_zb2020 was already the sole distractor and stays; three more
|
| 104 |
# domain-relevant distractors added to reach the required 2-4 total.
|
| 105 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 106 |
+
# priors:
|
| 107 |
+
# - name: w4_zb2020
|
| 108 |
+
# value: -6.332e8
|
| 109 |
+
# unit: m^2
|
| 110 |
+
# description: An empirical eddy-closure weight for the meridional S_y momentum component.
|
| 111 |
+
# source: Published eddy-momentum-closure literature.
|
| 112 |
+
# _role: distractor
|
| 113 |
+
# - name: beta_Coriolis
|
| 114 |
+
# value: 2.0e-11
|
| 115 |
+
# unit: m^-1 s^-1
|
| 116 |
+
# description: Meridional gradient of the Coriolis parameter at mid-latitudes (planetary beta-plane approximation). Relevant to geostrophic turbulence scaling.
|
| 117 |
+
# source: Standard oceanography reference value; Vallis 2006, Atmospheric and Oceanic Fluid Dynamics.
|
| 118 |
+
# _role: distractor
|
| 119 |
+
# - name: Rd_baroclinic
|
| 120 |
+
# value: 25000.0
|
| 121 |
+
# unit: m
|
| 122 |
+
# description: First baroclinic Rossby deformation radius — typical mid-latitude ocean value; sets the scale for baroclinic instability.
|
| 123 |
+
# source: Standard oceanography reference; Chelton et al. 1998.
|
| 124 |
+
# _role: distractor
|
| 125 |
+
# - name: rho_seawater_ref
|
| 126 |
+
# value: 1025.0
|
| 127 |
+
# unit: kg m^-3
|
| 128 |
+
# description: Reference seawater density at surface conditions; used in density-normalized pressure formulations.
|
| 129 |
+
# source: UNESCO/TEOS-10 standard reference density.
|
| 130 |
+
# _role: distractor
|
| 131 |
|
| 132 |
# Reference-baseline bank
|
| 133 |
references:
|
parker_spiral_imf__B_phi/metadata.yaml
CHANGED
|
@@ -90,19 +90,20 @@ references:
|
|
| 90 |
# rotation rate Omega is deliberately NOT offered: it is the law's own constant
|
| 91 |
# (LAW in parker_1958_omega_stated), so handing over any value of it — sidereal,
|
| 92 |
# synodic, or stated — would leak the answer (FM H1).
|
| 93 |
-
priors:
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
|
|
|
| 106 |
|
| 107 |
caps:
|
| 108 |
max_law_constants: 1
|
|
|
|
| 90 |
# rotation rate Omega is deliberately NOT offered: it is the law's own constant
|
| 91 |
# (LAW in parker_1958_omega_stated), so handing over any value of it — sidereal,
|
| 92 |
# synodic, or stated — would leak the answer (FM H1).
|
| 93 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 94 |
+
# priors:
|
| 95 |
+
# - name: AU_m
|
| 96 |
+
# value: 1.495978707e11
|
| 97 |
+
# unit: m
|
| 98 |
+
# description: One astronomical unit in metres (IAU 2012 Resolution B2, exact by definition) — the AU->m unit conversion applied to heliocentric distance r. A universal length-scale building block, not a fit coefficient.
|
| 99 |
+
# source: IAU 2012 Resolution B2 (exact definition, 149 597 870 700 m).
|
| 100 |
+
# _role: candidate
|
| 101 |
+
# - name: mu0_vacuum_permeability
|
| 102 |
+
# value: 1.25663706e-6
|
| 103 |
+
# unit: N*A^-2
|
| 104 |
+
# description: Vacuum magnetic permeability — a fundamental electromagnetic constant appearing in magnetic pressure and tension and in MHD wave speeds such as the Alfven speed. It belongs to the dynamic force-balance description of the magnetized plasma.
|
| 105 |
+
# source: CODATA 2018 (mu0 = 1.25663706212e-6 N*A^-2; ~ 4*pi*1e-7).
|
| 106 |
+
# _role: distractor
|
| 107 |
|
| 108 |
caps:
|
| 109 |
max_law_constants: 1
|
pfas_soil_sorption__log_kd/metadata.yaml
CHANGED
|
@@ -150,19 +150,20 @@ n_test: 540
|
|
| 150 |
# (PAH/HOC vs PFAS). Cross-class same-paper leakage: brackets the LAW slope into
|
| 151 |
# [0.5, 1.2] and reveals the log_KOC = slope * log_KOW concept.
|
| 152 |
# Ref: audit/AUDIT_CHECKLIST.md FM-H1 + proposed_FMs/2026-05-26_wave10_v1.0_systematic_sweep.md HIGH-2
|
| 153 |
-
priors:
|
| 154 |
-
|
| 155 |
-
|
| 156 |
-
|
| 157 |
-
|
| 158 |
-
|
| 159 |
-
|
| 160 |
-
|
| 161 |
-
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
|
|
|
|
| 166 |
|
| 167 |
references:
|
| 168 |
- id: fabregat_2025
|
|
|
|
| 150 |
# (PAH/HOC vs PFAS). Cross-class same-paper leakage: brackets the LAW slope into
|
| 151 |
# [0.5, 1.2] and reveals the log_KOC = slope * log_KOW concept.
|
| 152 |
# Ref: audit/AUDIT_CHECKLIST.md FM-H1 + proposed_FMs/2026-05-26_wave10_v1.0_systematic_sweep.md HIGH-2
|
| 153 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 154 |
+
# priors:
|
| 155 |
+
# - name: CEC_clay_smectite
|
| 156 |
+
# value: 100.0
|
| 157 |
+
# unit: cmol+/kg
|
| 158 |
+
# description: Typical cation exchange capacity of smectite clay minerals; indicates upper bound of CEC for high-clay soils. Relevant to soil characterization.
|
| 159 |
+
# source: Sparks 2003, Environmental Soil Chemistry (2nd ed.), Table 2.1.
|
| 160 |
+
# _role: distractor
|
| 161 |
+
# - name: pKa_PFOA
|
| 162 |
+
# value: -0.5
|
| 163 |
+
# unit: ""
|
| 164 |
+
# description: Approximate pKa of perfluorooctanoic acid (PFOA); negative value indicates full ionization at environmental pH. A molecular property relevant to PFAS speciation.
|
| 165 |
+
# source: Goss 2008, Environmental Science and Technology 42, 456-458, Table 1.
|
| 166 |
+
# _role: distractor
|
| 167 |
|
| 168 |
references:
|
| 169 |
- id: fabregat_2025
|
pp_total_cross_section_donnachie_landshoff__sigma_tot/metadata.yaml
CHANGED
|
@@ -76,25 +76,26 @@ n_test: 23
|
|
| 76 |
# Post-fix: CLEAN — verified no prior value matches any LAW_CONSTANTS or
|
| 77 |
# OTHER_CONSTANTS value in donnachie_1992.py or block_halzen_2005.py.
|
| 78 |
|
| 79 |
-
priors:
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
|
|
|
| 98 |
|
| 99 |
references:
|
| 100 |
- id: donnachie_1992
|
|
|
|
| 76 |
# Post-fix: CLEAN — verified no prior value matches any LAW_CONSTANTS or
|
| 77 |
# OTHER_CONSTANTS value in donnachie_1992.py or block_halzen_2005.py.
|
| 78 |
|
| 79 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 80 |
+
# priors:
|
| 81 |
+
# - name: alpha_f2_trajectory
|
| 82 |
+
# value: 0.70
|
| 83 |
+
# unit: ""
|
| 84 |
+
# description: f_2/rho Regge-trajectory intercept alpha_f2(0) ≈ 0.70, relevant to secondary Reggeon exchange in hadronic reactions. Universal QCD phenomenology reference.
|
| 85 |
+
# source: "Collins P.D.B. (1977) Introduction to Regge Theory, Cambridge UP, Chapter 4 — Reggeon trajectories table."
|
| 86 |
+
# _role: candidate
|
| 87 |
+
# - name: Y_pbarp
|
| 88 |
+
# value: 98.39
|
| 89 |
+
# unit: mb
|
| 90 |
+
# description: Reggeon coupling coefficient for pbar-p (proton-antiproton) scattering; the pbar-p reaction has a different Reggeon amplitude due to charge-conjugation than the proton-proton reaction studied here.
|
| 91 |
+
# source: Hadronic Regge-trajectory phenomenology; f2/rho secondary-Reggeon intercept.
|
| 92 |
+
# _role: distractor
|
| 93 |
+
# - name: sigma_tot_pp_low_energy
|
| 94 |
+
# value: 39.5
|
| 95 |
+
# unit: mb
|
| 96 |
+
# description: Reference pp total cross-section near sqrt(s)=10 GeV (ISR threshold). A domain-knowledge anchor for the cross-section scale.
|
| 97 |
+
# source: "Particle Data Group (2022) Cross Section Data, Table — pp total cross-section near sqrt(s)=10 GeV."
|
| 98 |
+
# _role: distractor
|
| 99 |
|
| 100 |
references:
|
| 101 |
- id: donnachie_1992
|
pv_module_temperature_efficiency_pvpmc__P_max/metadata.yaml
CHANGED
|
@@ -101,19 +101,20 @@ references:
|
|
| 101 |
# gamma_standard_Si (=gamma -0.0047, a direct leak of the temperature coefficient).
|
| 102 |
# The two fundamental constants below are genuine distractors: no active baseline
|
| 103 |
# uses them (they were SAPM k/q before king_2004 was dropped — see VERDICT.md).
|
| 104 |
-
priors:
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
|
|
|
| 117 |
|
| 118 |
# === v2 anti-dump caps ===
|
| 119 |
# max_law_constants 0 — auto-derived: dobos_2014 has LAW = {} after the 2026-06-02
|
|
|
|
| 101 |
# gamma_standard_Si (=gamma -0.0047, a direct leak of the temperature coefficient).
|
| 102 |
# The two fundamental constants below are genuine distractors: no active baseline
|
| 103 |
# uses them (they were SAPM k/q before king_2004 was dropped — see VERDICT.md).
|
| 104 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 105 |
+
# priors:
|
| 106 |
+
# - name: Boltzmann_k
|
| 107 |
+
# value: 1.38066e-23
|
| 108 |
+
# unit: J/K
|
| 109 |
+
# description: Boltzmann constant.
|
| 110 |
+
# source: "CODATA fundamental constant."
|
| 111 |
+
# _role: distractor
|
| 112 |
+
# - name: elementary_charge_q
|
| 113 |
+
# value: 1.60218e-19
|
| 114 |
+
# unit: C
|
| 115 |
+
# description: Elementary charge.
|
| 116 |
+
# source: "CODATA fundamental constant."
|
| 117 |
+
# _role: distractor
|
| 118 |
|
| 119 |
# === v2 anti-dump caps ===
|
| 120 |
# max_law_constants 0 — auto-derived: dobos_2014 has LAW = {} after the 2026-06-02
|
qg_turbulence_subgrid_pyqg__Sq/metadata.yaml
CHANGED
|
@@ -102,19 +102,20 @@ n_test: 131072
|
|
| 102 |
# LAW C_S=0.15; cited set {0.075, 0.15, 0.3} brackets the LAW value and reveals
|
| 103 |
# both the Smagorinsky-constant concept and order of magnitude).
|
| 104 |
# Ref: audit/AUDIT_CHECKLIST.md FM-H1 + proposed_FMs/2026-05-26_wave10_v1.0_systematic_sweep.md HIGH-1
|
| 105 |
-
priors:
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
|
|
|
| 118 |
|
| 119 |
references:
|
| 120 |
- id: smagorinsky_1963
|
|
|
|
| 102 |
# LAW C_S=0.15; cited set {0.075, 0.15, 0.3} brackets the LAW value and reveals
|
| 103 |
# both the Smagorinsky-constant concept and order of magnitude).
|
| 104 |
# Ref: audit/AUDIT_CHECKLIST.md FM-H1 + proposed_FMs/2026-05-26_wave10_v1.0_systematic_sweep.md HIGH-1
|
| 105 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 106 |
+
# priors:
|
| 107 |
+
# - name: beta_Coriolis_pyqg
|
| 108 |
+
# value: 1.5e-11
|
| 109 |
+
# unit: m^-1 s^-1
|
| 110 |
+
# description: Meridional Coriolis gradient (beta-plane coefficient) in the pyqg two-layer eddy configuration. Controls geostrophic turbulence scale and Rossby wave propagation.
|
| 111 |
+
# source: Ross et al. 2023, Table 1, PDF p. 6 (pyqg eddy configuration parameters).
|
| 112 |
+
# _role: distractor
|
| 113 |
+
# - name: f0_coriolis_pyqg
|
| 114 |
+
# value: 1.03e-4
|
| 115 |
+
# unit: s^-1
|
| 116 |
+
# description: Coriolis parameter f_0 at the reference latitude in the pyqg eddy configuration. Enters the QG PV definition.
|
| 117 |
+
# source: Ross et al. 2023, Table 1, PDF p. 6 (pyqg eddy configuration parameters).
|
| 118 |
+
# _role: distractor
|
| 119 |
|
| 120 |
references:
|
| 121 |
- id: smagorinsky_1963
|
red_giant_asteroseismology__delta_nu/metadata.yaml
CHANGED
|
@@ -111,25 +111,26 @@ n_test: 3219
|
|
| 111 |
# (it hands the SR system brown_1991's exponent). Per H1 (delete, don't replace):
|
| 112 |
# beta_theoretical REMOVED. The 3 remaining solar-reference priors only overlap
|
| 113 |
# OTHER_CONSTANTS of the active baselines (allowed by design), not any LAW.
|
| 114 |
-
priors:
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
|
| 129 |
-
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
|
|
|
|
| 133 |
|
| 134 |
# Reference-baseline bank
|
| 135 |
# Wave-16 (2026-05-27): brown_1991 added to bring runnable baseline count
|
|
|
|
| 111 |
# (it hands the SR system brown_1991's exponent). Per H1 (delete, don't replace):
|
| 112 |
# beta_theoretical REMOVED. The 3 remaining solar-reference priors only overlap
|
| 113 |
# OTHER_CONSTANTS of the active baselines (allowed by design), not any LAW.
|
| 114 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 115 |
+
# priors:
|
| 116 |
+
# - name: delta_nu_sun
|
| 117 |
+
# value: 135.1
|
| 118 |
+
# unit: uHz
|
| 119 |
+
# description: Solar large-frequency-separation reference adopted in the seismic scaling relation (Yu+2018 Eq. 3). A solar reference constant.
|
| 120 |
+
# source: Yu et al. 2018, Eq. 3, PDF p. 5
|
| 121 |
+
# _role: distractor
|
| 122 |
+
# - name: nu_max_sun
|
| 123 |
+
# value: 3090.0
|
| 124 |
+
# unit: uHz
|
| 125 |
+
# description: Solar reference frequency of maximum oscillation power (Yu+2018 Eq. 2). Solar anchor value.
|
| 126 |
+
# source: Yu et al. 2018, Eq. 2, PDF p. 5
|
| 127 |
+
# _role: distractor
|
| 128 |
+
# - name: Teff_sun
|
| 129 |
+
# value: 5777.0
|
| 130 |
+
# unit: K
|
| 131 |
+
# description: Solar effective temperature used as the reference in the ν_max scaling relation (Yu+2018 Eq. 2). Standard solar value.
|
| 132 |
+
# source: Yu et al. 2018, §3.1, PDF p. 5
|
| 133 |
+
# _role: distractor
|
| 134 |
|
| 135 |
# Reference-baseline bank
|
| 136 |
# Wave-16 (2026-05-27): brown_1991 added to bring runnable baseline count
|
red_giant_asteroseismology__nu_max/metadata.yaml
CHANGED
|
@@ -132,25 +132,26 @@ n_test: 3219
|
|
| 132 |
# correct: the Yu+2018 law is purely empirical, so any candidate would expose a fit
|
| 133 |
# coefficient (per §C12 guidance, prefer no candidate over a leaking one).
|
| 134 |
|
| 135 |
-
priors:
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
|
| 141 |
-
|
| 142 |
-
|
| 143 |
-
|
| 144 |
-
|
| 145 |
-
|
| 146 |
-
|
| 147 |
-
|
| 148 |
-
|
| 149 |
-
|
| 150 |
-
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
|
|
|
|
| 154 |
|
| 155 |
# Reference-baseline bank
|
| 156 |
references:
|
|
|
|
| 132 |
# correct: the Yu+2018 law is purely empirical, so any candidate would expose a fit
|
| 133 |
# coefficient (per §C12 guidance, prefer no candidate over a leaking one).
|
| 134 |
|
| 135 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 136 |
+
# priors:
|
| 137 |
+
# - name: delta_nu_sun
|
| 138 |
+
# value: 135.1
|
| 139 |
+
# unit: μHz
|
| 140 |
+
# description: Solar large-frequency-separation reference adopted in the seismic scaling relation (Yu+2018 Eq. 3). A solar reference constant.
|
| 141 |
+
# source: Yu et al. 2018, Eq. 3, PDF p. 5
|
| 142 |
+
# _role: distractor
|
| 143 |
+
# - name: nu_max_sun
|
| 144 |
+
# value: 3090.0
|
| 145 |
+
# unit: μHz
|
| 146 |
+
# description: Solar reference frequency of maximum oscillation power (Yu+2018 Eq. 2). Solar anchor value. Enters the ν_max scaling relation as a normalisation anchor for the Sun.
|
| 147 |
+
# source: Yu et al. 2018, Eq. 2, PDF p. 5
|
| 148 |
+
# _role: distractor
|
| 149 |
+
# - name: log_g_sun_cgs
|
| 150 |
+
# value: 4.438
|
| 151 |
+
# unit: "cm/s^2 (log10)"
|
| 152 |
+
# description: Solar surface gravity log10(g_sun) ≈ 4.438 cm/s² (cgs), a solar reference value used in seismic scaling-relation normalisation. A domain-relevant solar constant from asteroseismic conventions.
|
| 153 |
+
# source: Chaplin WJ & Miglio A (2013) ARA&A 51:353-392, §2; IAU 2015 B3 nominal solar values.
|
| 154 |
+
# _role: distractor
|
| 155 |
|
| 156 |
# Reference-baseline bank
|
| 157 |
references:
|
rock_core_permeability_norweg__log_k/metadata.yaml
CHANGED
|
@@ -49,25 +49,26 @@ n_test: 22246
|
|
| 49 |
# (slichter_1898 N_EXPONENT / terzaghi_1925 N_THRESHOLD) → priors leak. The KC
|
| 50 |
# numerator exponent 3.0 is retained: it is hardcoded structurally in
|
| 51 |
# kozeny_carman_1937.predict(), not declared in any LAW_CONSTANTS dict.
|
| 52 |
-
priors:
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
|
|
|
| 71 |
|
| 72 |
# === Reference baselines ===
|
| 73 |
# All three formulas demoted to Type I (LOCAL_FITTABLE = {}, no fit()).
|
|
|
|
| 49 |
# (slichter_1898 N_EXPONENT / terzaghi_1925 N_THRESHOLD) → priors leak. The KC
|
| 50 |
# numerator exponent 3.0 is retained: it is hardcoded structurally in
|
| 51 |
# kozeny_carman_1937.predict(), not declared in any LAW_CONSTANTS dict.
|
| 52 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 53 |
+
# priors:
|
| 54 |
+
# - name: kc_pre_factor
|
| 55 |
+
# value: 0.0083
|
| 56 |
+
# unit: ""
|
| 57 |
+
# description: A candidate pre-factor coefficient (8.3e-3) from grain-size permeability correlations.
|
| 58 |
+
# source: Cabalar and Akbulut 2016 Table 4 (PDF p.10, SpringerPlus 5:820).
|
| 59 |
+
# _role: candidate
|
| 60 |
+
# - name: kc_canonical_exponent
|
| 61 |
+
# value: 3.0
|
| 62 |
+
# unit: ""
|
| 63 |
+
# description: A candidate porosity-exponent value from grain-size permeability correlations.
|
| 64 |
+
# source: Classical pore-network permeability theory value.
|
| 65 |
+
# _role: candidate
|
| 66 |
+
# - name: hazen_pre_factor
|
| 67 |
+
# value: 0.0006
|
| 68 |
+
# unit: ""
|
| 69 |
+
# description: A candidate pre-factor value (6e-4) from grain-size permeability correlations.
|
| 70 |
+
# source: Cabalar and Akbulut 2016 Table 4 (PDF p.10).
|
| 71 |
+
# _role: distractor
|
| 72 |
|
| 73 |
# === Reference baselines ===
|
| 74 |
# All three formulas demoted to Type I (LOCAL_FITTABLE = {}, no fit()).
|
running_endurance_iaaf__velocity/metadata.yaml
CHANGED
|
@@ -119,22 +119,23 @@ best_baseline: 0.26277 # emig_2020 — rmse (lower is better); regenerated 20
|
|
| 119 |
# time; distance_m is already a released input, so marathon_distance stays).
|
| 120 |
# Not padded: 4 distractors -> 3. POST-FIX priors: 0 candidate + 3 distractor.
|
| 121 |
# Reference: audit/AUDIT_CHECKLIST.md §B7 (target-leak via the priors channel).
|
| 122 |
-
priors:
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
|
| 129 |
-
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
|
| 133 |
-
|
| 134 |
-
|
| 135 |
-
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
|
|
|
|
|
|
| 119 |
# time; distance_m is already a released input, so marathon_distance stays).
|
| 120 |
# Not padded: 4 distractors -> 3. POST-FIX priors: 0 candidate + 3 distractor.
|
| 121 |
# Reference: audit/AUDIT_CHECKLIST.md §B7 (target-leak via the priors channel).
|
| 122 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 123 |
+
# priors:
|
| 124 |
+
# - name: marathon_distance
|
| 125 |
+
# value: 42195.0
|
| 126 |
+
# unit: m
|
| 127 |
+
# description: Official marathon race distance (IAAF standard).
|
| 128 |
+
# source: IAAF world athletics standards.
|
| 129 |
+
# _role: distractor
|
| 130 |
+
# - name: km_per_mile
|
| 131 |
+
# value: 1.609344
|
| 132 |
+
# unit: km/mile
|
| 133 |
+
# description: Exact kilometres per international mile (IAU/IUPAC definition). A unit conversion factor.
|
| 134 |
+
# source: International Bureau of Weights and Measures (BIPM); exact definition.
|
| 135 |
+
# _role: distractor
|
| 136 |
+
# - name: vo2max_elite_threshold_mlkgmin
|
| 137 |
+
# value: 70.0
|
| 138 |
+
# unit: mL/kg/min
|
| 139 |
+
# description: Approximate VO2max threshold for elite male endurance runners (exercise physiology reference).
|
| 140 |
+
# source: Exercise physiology literature (e.g. Joyner & Coyle 2008, J. Physiol.); approximate canonical threshold.
|
| 141 |
+
# _role: distractor
|
sne_ia_distance_modulus_pantheonplus__mu/metadata.yaml
CHANGED
|
@@ -83,25 +83,26 @@ n_test: 308
|
|
| 83 |
# Post-fix: CLEAN — verified no prior value matches any LAW_CONSTANTS or
|
| 84 |
# OTHER_CONSTANTS value in linder_2003.py.
|
| 85 |
|
| 86 |
-
priors:
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
|
|
|
| 105 |
|
| 106 |
references:
|
| 107 |
- id: linder_2003
|
|
|
|
| 83 |
# Post-fix: CLEAN — verified no prior value matches any LAW_CONSTANTS or
|
| 84 |
# OTHER_CONSTANTS value in linder_2003.py.
|
| 85 |
|
| 86 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 87 |
+
# priors:
|
| 88 |
+
# - name: c_light_km_s
|
| 89 |
+
# value: 299792.458
|
| 90 |
+
# unit: km/s
|
| 91 |
+
# description: Speed of light — CGPM exact SI definition (exact since 17th CGPM 1983; BIPM SI Brochure 9th ed.). A universal physical constant that enters the comoving distance integral as c/H0 [Mpc]. This is a building-block physical standard, not an empirical cosmological fit result.
|
| 92 |
+
# source: BIPM SI Brochure 9th ed. (2019).
|
| 93 |
+
# _role: candidate
|
| 94 |
+
# - name: alpha_tripp_brout
|
| 95 |
+
# value: 0.148
|
| 96 |
+
# unit: ""
|
| 97 |
+
# description: Light-curve stretch coefficient used in the empirical light-curve standardisation of the apparent magnitudes.
|
| 98 |
+
# source: Empirical light-curve-standardisation fit to the SN Ia sample.
|
| 99 |
+
# _role: distractor
|
| 100 |
+
# - name: MB_riess2022
|
| 101 |
+
# value: -19.253
|
| 102 |
+
# unit: mag
|
| 103 |
+
# description: Fiducial SN Ia absolute magnitude from a distance-ladder calibration. The light-curve standardisation is already absorbed into the observed mu values.
|
| 104 |
+
# source: Distance-ladder absolute-magnitude calibration of SNe Ia.
|
| 105 |
+
# _role: distractor
|
| 106 |
|
| 107 |
references:
|
| 108 |
- id: linder_2003
|
solar_flat_plate_collector_graz_efficiency__eta_thermal/metadata.yaml
CHANGED
|
@@ -75,31 +75,34 @@ n_train: 252
|
|
| 75 |
n_test: 63
|
| 76 |
|
| 77 |
# Candidate prior constants — the `priors` prompt slot.
|
| 78 |
-
priors:
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
|
|
|
|
|
|
|
|
|
| 103 |
|
| 104 |
# === Reference baselines ===
|
| 105 |
references:
|
|
|
|
| 75 |
n_test: 63
|
| 76 |
|
| 77 |
# Candidate prior constants — the `priors` prompt slot.
|
| 78 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 79 |
+
# priors:
|
| 80 |
+
# [LEAK-REVIEW 2026-06-08] a0/a1/a2 are certified-datasheet coefficients that ≈ the fitted
|
| 81 |
+
# answer (HWB A0/A1/A2). Commented out (NOT deleted) pending decision; uncomment to restore.
|
| 82 |
+
# - name: optical_efficiency_a0
|
| 83 |
+
# value: 0.745
|
| 84 |
+
# unit: dimensionless
|
| 85 |
+
# description: Certified optical (zero-loss) efficiency eta0 of the Arcon-Sunmark HTHEATstore-35/10 collector (mean-fluid SP-test form); a candidate value for the zero-loss optical-efficiency term.
|
| 86 |
+
# source: Manufacturer SP-certified collector datasheet (mean-fluid form).
|
| 87 |
+
# _role: candidate
|
| 88 |
+
# - name: linear_loss_a1
|
| 89 |
+
# value: 2.067
|
| 90 |
+
# unit: W m^-2 K^-1
|
| 91 |
+
# description: Certified linear heat-loss coefficient (mean-fluid SP-test form); a candidate value for a linear heat-loss term.
|
| 92 |
+
# source: Manufacturer SP-certified collector datasheet.
|
| 93 |
+
# _role: candidate
|
| 94 |
+
# - name: quadratic_loss_a2
|
| 95 |
+
# value: 0.009
|
| 96 |
+
# unit: W m^-2 K^-2
|
| 97 |
+
# description: Certified quadratic (re-radiation) heat-loss coefficient; a candidate value for a second-order heat-loss term.
|
| 98 |
+
# source: Manufacturer SP-certified collector datasheet. Small; within the noise floor on this dataset's narrow operating range.
|
| 99 |
+
# _role: candidate
|
| 100 |
+
# - name: water_cp
|
| 101 |
+
# value: 4186.0
|
| 102 |
+
# unit: J kg^-1 K^-1
|
| 103 |
+
# description: Specific heat capacity of the water/glycol heat-transfer fluid.
|
| 104 |
+
# source: Standard. Used to compute the useful heat (and hence eta) upstream; the efficiency target is provided as given.
|
| 105 |
+
# _role: distractor
|
| 106 |
|
| 107 |
# === Reference baselines ===
|
| 108 |
references:
|
traffic_flow_density_ngsim__q_flow/metadata.yaml
CHANGED
|
@@ -96,37 +96,38 @@ n_train: 2286
|
|
| 96 |
n_test: 572
|
| 97 |
|
| 98 |
# Candidate prior constants — the `priors` prompt slot.
|
| 99 |
-
priors:
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
|
| 129 |
-
|
|
|
|
| 130 |
|
| 131 |
# === Reference baselines ===
|
| 132 |
references:
|
|
|
|
| 96 |
n_test: 572
|
| 97 |
|
| 98 |
# Candidate prior constants — the `priors` prompt slot.
|
| 99 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 100 |
+
# priors:
|
| 101 |
+
# - name: us_freeway_typical_free_flow_speed
|
| 102 |
+
# value: 105.0
|
| 103 |
+
# unit: km/h
|
| 104 |
+
# description: Typical free-flow (uncongested) speed on a US Interstate-class freeway under a 65 mph speed limit; a candidate value for a free-flow-speed term.
|
| 105 |
+
# source: AASHTO Green Book 2018 §3.2 design-speed table; speed-limit data from the California DOT 2005 corridor logs that accompany NGSIM.
|
| 106 |
+
# _role: candidate
|
| 107 |
+
# - name: us_freeway_typical_capacity_per_lane
|
| 108 |
+
# value: 2200.0
|
| 109 |
+
# unit: veh/h/lane
|
| 110 |
+
# description: Practical per-lane capacity at level-of-service E for a US freeway with 12-ft lanes and good shoulders; a candidate value for a maximum-flow / capacity term.
|
| 111 |
+
# source: HCM (Highway Capacity Manual) 6th edition, Exhibit 12-6. Empirically NGSIM cells approach but rarely exceed 2,500 veh/h/lane; the v2 test reaches 3,499 in one outlier cell.
|
| 112 |
+
# _role: candidate
|
| 113 |
+
# - name: us_freeway_typical_jam_density_per_lane
|
| 114 |
+
# value: 180.0
|
| 115 |
+
# unit: veh/km/lane
|
| 116 |
+
# description: Jam density (bumper-to-bumper) for typical US passenger-car traffic with ~6 m vehicle length + minimum gap; a candidate value for K_JAM.
|
| 117 |
+
# source: Transportation Research Board, NCHRP Report 387 (1997).
|
| 118 |
+
# _role: candidate
|
| 119 |
+
# - name: us_freeway_backward_wave_speed
|
| 120 |
+
# value: 19.0
|
| 121 |
+
# unit: km/h
|
| 122 |
+
# description: Magnitude of the LWR backward kinematic-wave speed for US freeways; a candidate value for the speed of the congested branch.
|
| 123 |
+
# source: Cassidy & Bertini 1999 TR-B 33:25 §4, Eisenhower Expressway empirical estimate. Daganzo 1994 §3.1 gives "around 20 km/h" for typical conditions.
|
| 124 |
+
# _role: candidate
|
| 125 |
+
# - name: pedestrian_jam_density
|
| 126 |
+
# value: 5.4
|
| 127 |
+
# unit: ped/m²
|
| 128 |
+
# description: Maximum sustainable pedestrian density (Fruin's level-of-service F).
|
| 129 |
+
# source: Fruin 1971 "Pedestrian Planning and Design", §3. Pedestrian-flow constant, for pedestrian rather than vehicular traffic.
|
| 130 |
+
# _role: distractor
|
| 131 |
|
| 132 |
# === Reference baselines ===
|
| 133 |
references:
|
tree_biomass_allometry_chave__agb/metadata.yaml
CHANGED
|
@@ -102,25 +102,26 @@ n_test: 1001
|
|
| 102 |
# §C12 corrected 2026-05-29; a prior↔OTHER_CONSTANTS overlap is not a leak. Moot
|
| 103 |
# here regardless — both formulas have OTHER_CONSTANTS={}.)
|
| 104 |
|
| 105 |
-
priors:
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
|
|
|
| 124 |
|
| 125 |
# Reference-baseline bank
|
| 126 |
references:
|
|
|
|
| 102 |
# §C12 corrected 2026-05-29; a prior↔OTHER_CONSTANTS overlap is not a leak. Moot
|
| 103 |
# here regardless — both formulas have OTHER_CONSTANTS={}.)
|
| 104 |
|
| 105 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 106 |
+
# priors:
|
| 107 |
+
# - name: WBE_metabolic_exponent
|
| 108 |
+
# value: 0.75
|
| 109 |
+
# unit: dimensionless
|
| 110 |
+
# description: West-Brown-Enquist (1997, 1999) metabolic scaling exponent — metabolic rate B ∝ M^(3/4), derived from fractal vascular network theory. This applies to metabolic scaling, which has a different target variable and input structure from the aboveground-biomass allometry in this task.
|
| 111 |
+
# source: "West GB, Brown JH, Enquist BJ (1997) Science 276:122-126; West GB et al. (1999) Science 284:1677-1679."
|
| 112 |
+
# _role: distractor
|
| 113 |
+
# - name: Jenkins_hardwood_exponent
|
| 114 |
+
# value: 2.4800
|
| 115 |
+
# unit: dimensionless
|
| 116 |
+
# description: Power-law exponent for hardwood component biomass from Jenkins et al. (2003) US national biomass equation for hardwoods — biomass components ∝ D^exponent. Different species group and country calibration from the pantropical setting of this task.
|
| 117 |
+
# source: "Jenkins JC et al. (2003) Ecological Monographs 73:443-463, Table 4 (hardwood, total aboveground biomass)."
|
| 118 |
+
# _role: distractor
|
| 119 |
+
# - name: wood_density_tropical_mean
|
| 120 |
+
# value: 0.57
|
| 121 |
+
# unit: g/cm^3
|
| 122 |
+
# description: Global mean tropical angiosperm wood density from the Zanne et al. (2009) global wood-density database, used in pan-tropical biomass studies as a reference value.
|
| 123 |
+
# source: "Zanne AE et al. (2009) Global wood density database (Dryad). Reported in Chave J et al. (2009) Ecol. Lett. 12:351-366, Table 1."
|
| 124 |
+
# _role: distractor
|
| 125 |
|
| 126 |
# Reference-baseline bank
|
| 127 |
references:
|
volcanic_column_mer_ivespa__H_top/metadata.yaml
CHANGED
|
@@ -58,31 +58,35 @@ n_train: 98
|
|
| 58 |
n_test: 32
|
| 59 |
|
| 60 |
# Candidate prior constants — the `priors` prompt slot.
|
| 61 |
-
priors:
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 86 |
|
| 87 |
# === Reference baselines ===
|
| 88 |
references:
|
|
|
|
| 58 |
n_test: 32
|
| 59 |
|
| 60 |
# Candidate prior constants — the `priors` prompt slot.
|
| 61 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 62 |
+
# priors:
|
| 63 |
+
# [LEAK-REMOVE 2026-06-08] value = fitted answer constant; commented out (NOT deleted).
|
| 64 |
+
# - name: mtt_plume_exponent
|
| 65 |
+
# value: 0.25
|
| 66 |
+
# unit: dimensionless
|
| 67 |
+
# description: First-principles buoyant-plume scaling exponent for a point buoyancy source rising in uniform stratification. A candidate value for a scaling exponent.
|
| 68 |
+
# source: Classical buoyant-plume theory for a point buoyancy source in uniform stratification.
|
| 69 |
+
# _role: candidate
|
| 70 |
+
# [LEAK-REMOVE 2026-06-08] value = fitted answer constant; commented out (NOT deleted).
|
| 71 |
+
# - name: aubry_2023_exponent
|
| 72 |
+
# value: 0.226
|
| 73 |
+
# unit: dimensionless
|
| 74 |
+
# description: Empirical scaling exponent fit on the full 130-event corpus.
|
| 75 |
+
# source: Empirical full-corpus fit. A candidate exponent; note the full-corpus fit "saw" the high-MER events, so on the v2 OOD train alone the fitted exponent is shallower (~0.18).
|
| 76 |
+
# _role: candidate
|
| 77 |
+
# [LEAK-REMOVE 2026-06-08] value = fitted answer constant; commented out (NOT deleted).
|
| 78 |
+
# - name: aubry_2023_prefactor
|
| 79 |
+
# value: 0.345
|
| 80 |
+
# unit: km (kg/s)^-0.226
|
| 81 |
+
# description: Empirical scaling prefactor (full-corpus fit).
|
| 82 |
+
# source: Empirical full-corpus fit. Candidate prefactor.
|
| 83 |
+
# _role: candidate
|
| 84 |
+
# - name: tropopause_height_midlat
|
| 85 |
+
# value: 11.0
|
| 86 |
+
# unit: km
|
| 87 |
+
# description: Typical mid-latitude tropopause height.
|
| 88 |
+
# source: Standard atmosphere. Volcanic columns that overshoot the tropopause spread laterally; the column-height scaling has no clear atmospheric dependence.
|
| 89 |
+
# _role: distractor
|
| 90 |
|
| 91 |
# === Reference baselines ===
|
| 92 |
references:
|
wind_turbine_power_curve_engie__power_kW/metadata.yaml
CHANGED
|
@@ -42,25 +42,26 @@ n_train: 204288
|
|
| 42 |
n_test: 136192
|
| 43 |
|
| 44 |
# Candidate prior constants — the `priors` prompt slot.
|
| 45 |
-
priors:
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
|
|
|
| 64 |
|
| 65 |
# === Reference baselines ===
|
| 66 |
references:
|
|
|
|
| 42 |
n_test: 136192
|
| 43 |
|
| 44 |
# Candidate prior constants — the `priors` prompt slot.
|
| 45 |
+
# priors DISABLED 2026-06-08 — not shown to the agent (test showed they're redundant: models already know the constants / the answer is fittable from data). Kept commented for record.
|
| 46 |
+
# priors:
|
| 47 |
+
# - name: betz_limit
|
| 48 |
+
# value: 0.5926
|
| 49 |
+
# unit: dimensionless
|
| 50 |
+
# description: Betz theoretical maximum power coefficient (16/27).
|
| 51 |
+
# source: Betz 1919, "Das Maximum der theoretisch möglichen Ausnützung des Windes durch Windmotoren" — the universal upper bound on Cp for any wind energy extraction device. A candidate value for a power-coefficient constant; the empirical operating value is somewhat lower.
|
| 52 |
+
# _role: candidate
|
| 53 |
+
# - name: senvion_mm82_rated_power
|
| 54 |
+
# value: 2050.0
|
| 55 |
+
# unit: kW
|
| 56 |
+
# description: Senvion MM82 turbine datasheet rated power.
|
| 57 |
+
# source: Senvion (now Nordex) MM82 turbine product datasheet — the rated electrical power output at v ≥ v_r. A candidate value for the asymptote / rated-power constant.
|
| 58 |
+
# _role: candidate
|
| 59 |
+
# - name: silicon_solar_band_gap
|
| 60 |
+
# value: 1.12
|
| 61 |
+
# unit: eV
|
| 62 |
+
# description: Crystalline-silicon photovoltaic semiconductor band gap.
|
| 63 |
+
# source: A solar-energy domain quantity but irrelevant to wind power generation, which operates via mechanical kinetic-energy extraction rather than photon absorption.
|
| 64 |
+
# _role: distractor
|
| 65 |
|
| 66 |
# === Reference baselines ===
|
| 67 |
references:
|