ClawBench / app.py

Commit History

UI: default to V2 Hermes (single harness) + add Reset button to mirror website headline view
74fbf87
verified

AgPerry commited on

leaderboard: render Cost / task column (avg per-task USD)
7ead7b7
verified

AgPerry commited on

docs: re-apply TABLE_INTRO grammar fix (lost in concurrent revert)
ba965ff

AgPerry commited on

fix: sort leaderboard by Reward (lenient) DESC (was: preserve CSV order)
d921761
verified

AgPerry commited on

fix: preserve CSV row order in leaderboard (matches website ranking)
e29739e
verified

AgPerry commited on

docs: fix TABLE_INTRO grammar after lenient/strict split; mark Reward (lenient) as default sort key
bbb0385

AgPerry commited on

feat: split Reward into lenient/strict columns + render both
78eb1d2
verified

AgPerry commited on

feat: default-sort leaderboard by Reward DESC
3835b02

Yuxuan Zhang commited on

fix: auto-refresh leaderboard from dataset on every page load (no module-level cache)
72e98ec

Yuxuan Zhang commited on

V2: switch to raw Intercepted DESC sort + Hermes-only filter (visible-column order)
a036d16
verified

AgPerry commited on

Sync Space INTRO+TABLE_INTRO+citation: Intercepted is sort key, full author list in BibTeX
a478c75
verified

AgPerry commited on

Switch sort key to corpus-interception (Stage 1) with corpus-reward as tiebreak
e77a483
verified

AgPerry commited on

Sort by corpus reward (passed/full_corpus_size) so partial batches don't outrank complete ones
78cfa0c
verified

AgPerry commited on

Upload folder using huggingface_hub
41e181d
verified

AgPerry commited on