Add files using upload-large-folder tool

README.md

@@ -34,41 +34,45 @@ configs:
 
 # Spatial MMCoT v1 · `vdrop`
 
-VDrop cross-view (QianYangMILA/vdrop-crossview-8k, [arXiv:2605.27310](https://arxiv.org/abs/2605.27310)), Infinigen Indoors rooms. Two egocentric views of one room with partly overlapping fields of view are the input (`input_image_0` = cam0, `input_image_1` = cam1); the target is upstream's ground-truth `panorama` render, a wide panorama of the room that spans both views. The questions are multiple choice and the trained answer is the option letter. The upstream release ships no reasoning text at all, so every thought here is EMPTY: `output_text_list` is exactly `<think></think><image_start>` and `<image_end><think></think><answer>X</answer>`. The panorama belongs to the scene, not the question: every question about a scene shares that scene's two input views and its one panorama, so the 7,015 rows use 1,158 distinct panoramas. Report validation scores with confidence intervals clustered by scene (for example a bootstrap over `meta.scene_id`), not as independent rows. The scenes were generated with Infinigen Indoors (github.com/princeton-vl/infinigen, whose code is BSD-3-Clause); the renders are distributed by the VDrop authors under CC-BY-4.0. Changes from upstream: the PNG images are re-encoded as JPEG and downscaled (the two input views from 1280x720 to 512x288, the panorama to 1,024 px wide); of the upstream thinking images only the panorama is kept; perspective-taking questions phrased relative to 'your partner' are dropped (as the upstream authors' `build_interleaved_from_visual_only.py` does by default); rows are removed by answer-prior balancing and split into train/validation by scene (both below); the ThinkMorph system prompt is prepended to the unchanged question.
+VDrop cross-view (QianYangMILA/vdrop-crossview-8k, [arXiv:2605.27310](https://arxiv.org/abs/2605.27310)), Infinigen Indoors rooms. Two egocentric views of one room with partly overlapping fields of view are the input (`input_image_0` = cam0, `input_image_1` = cam1); the target is upstream's ground-truth `panorama` render, a wide panorama of the room that spans both views. The questions are multiple choice and the trained answer is the option letter. The upstream release ships no reasoning text at all, so every thought here is EMPTY: `output_text_list` is exactly `<think></think><image_start>` and `<image_end><think></think><answer>X</answer>`. The panorama belongs to the scene, not the question: every question about a scene shares that scene's two input views and its one panorama, so the 6,807 rows use 1,152 distinct panoramas. Report validation scores with confidence intervals clustered by scene (for example a bootstrap over `meta.scene_id`), not as independent rows. The scenes were generated with Infinigen Indoors (github.com/princeton-vl/infinigen, whose code is BSD-3-Clause); the renders are distributed by the VDrop authors under CC-BY-4.0. Changes from upstream: the PNG images are re-encoded as JPEG and downscaled (the two input views from 1280x720 to 512x288, the panorama to 1,024 px wide); of the upstream thinking images only the panorama is kept; perspective-taking questions phrased relative to 'your partner' are dropped (as the upstream authors' `build_interleaved_from_visual_only.py` does by default); rows are removed by answer-prior balancing and split into train/validation by scene (both below); the ThinkMorph system prompt is prepended to the unchanged question.
 
-**Supervision kind** (`supervision_kind` in `meta`): `visual_only` on every row: the thoughts are empty: the rows supervise the image and the answer only.
+## Reasoning text in this release (r4)
 
-Upstream: [`QianYangMILA/vdrop-crossview-8k`](https://huggingface.co/datasets/QianYangMILA/vdrop-crossview-8k). Licence: **cc-by-4.0**.
+The paragraph above describes the upstream data, and where it describes the reasoning text it describes the text of the previous release (r3). In this release every thought was rewritten: Qwen3.8-27B in thinking mode (temperature 1.0, top-p 0.95, top-k 20; prompt v4.1, `scripts/annotate/rationale.py` in our conversion code (the code repository is not public yet)) was shown the row's input images, its target images in order, the question, the options and the answer, and wrote the thought before each target image (what is still unknown, and what the picture should show to settle it) and the read-back after the last one. The images, the question, the options, the answer and the number of steps are r3's, checked row by row (image bytes, instruction, K and the `<answer>` value).
 
-## Known issues
+This is the training subset: 208 rows of r4 are not in it, listed with their reason in `reports/subset_dropped.jsonl`: 208 the rewrite model judged the stored answer doubtful (its note is in the list) (`label_doubt`).
 
-No row of this source is listed in `reports/known_issues/` (its `summary.json` is empty); the measured caveats below still apply.
+Self-check of the rewritten text (measured 2026-10-02 on 22 rows sampled from this source before the removals above, each judged against its full-resolution pictures): 0.86 faithfulness errors per row (a claim the pictures contradict), 55% of rows with none; the plan names the real obstacle on 96%, the read-back reads the picture it follows on 82%, and the answer follows from the text on 82%.
 
-### Measured caveats
+The measured caveats on the r3 card were counted on r3's rows and text and are not repeated here; the per-row known-issue lists below were measured again on these rows and this text.
 
-Measured on this release by the pre-publication review (2026-09-25): problems that cannot be listed row by row (a shortcut in the options, a label convention, an upstream labelling scheme) and what the review found around the lists above. Where a caveat counts listed rows ("listed as `...`"), the count is the table's, read from `reports/known_issues/summary.json`. Its other numbers are the review's own measurements, which no file carries: they hold for exactly these rows and are not re-measured automatically. Items marked **Training-signal defect** are problems in what the rows teach, not only in how they are described; no row was removed for them.
+**Supervision kind** (`supervision_kind` in `meta`): `visual_only` on every row: the thoughts are empty: the rows supervise the image and the answer only.
 
-- Validation is strongly clustered by scene. Its 267 rows come from 43 scenes: one scene supplies 52 of them, and the three largest supply 113 (42.3%). In training, 1,115 scenes give a median of 3 rows per scene and at most 100 (one scene's 100 rows share one panorama). Report validation scores with confidence intervals clustered by scene, not as 267 independent rows.
+Upstream: [`QianYangMILA/vdrop-crossview-8k`](https://huggingface.co/datasets/QianYangMILA/vdrop-crossview-8k). Licence: **cc-by-4.0**.
+
+## Known issues
+
+No row of this source is listed in `reports/known_issues/` (its `summary.json` is empty).
 
 ## Size
 
 | split | rows | target image slots | distinct target images |
 |---|---:|---:|---:|
-| train | 6,748 | 6,748 | 1,115 |
-| validation | 267 | 267 | 43 |
+| train | 6,543 | 6,543 | 1,109 |
+| validation | 264 | 264 | 43 |
 
 A slot is one target position in one row; rows can share a target image (same scene or same intermediate state), and a row can repeat one of its own targets, so there are fewer distinct images (by content hash) than slots.
 
 | task | train | validation |
 |---|---:|---:|
-| anchor | 690 | 21 |
-| counting | 590 | 37 |
-| perspective_taking | 1,245 | 49 |
-| relative_direction | 1,399 | 50 |
-| relative_distance | 2,824 | 110 |
+| anchor | 657 | 20 |
+| counting | 570 | 37 |
+| perspective_taking | 1,218 | 48 |
+| relative_direction | 1,330 | 50 |
+| relative_distance | 2,768 | 109 |
 
 Input images per row: 2. Target images per row (the images the model is trained to generate): 1.
-Image corpus (`source_scene_corpus`): infinigen 7,015.
+Image corpus (`source_scene_corpus`): infinigen 6,807.
 
 ## Row format
 
@@ -85,9 +89,7 @@ output_text_list  list<string>  K+1 elements:
 row_uid           string        join key to `meta` and `preview`
 ```
 
-Every image is a JPEG, and no input image is larger than 512 px on its long edge (measured on this release, 2026-09-25); the size each target was stored at is `target_px` in `meta`.
-
-On every row `<answer>` holds the option key (answer type `mcq_letter` 7,015) that the model is trained to emit (a letter for `mcq_letter`), while `meta.answer_value` (the `answer` column of `preview`) holds that option's text. Map the key through the options listed in the question before comparing the two, and score model output against `<answer>`.
+On every row `<answer>` holds the option key (answer type `mcq_letter` 6,807) that the model is trained to emit (a letter for `mcq_letter`), while `meta.answer_value` (the `answer` column of `preview`) holds that option's text. Map the key through the options listed in the question before comparing the two, and score model output against `<answer>`.
 
 The system prompt is ThinkMorph's `VLM_THINK_SYSTEM_PROMPT` from its `inferencer.py`, verbatim (`GEN_THINK_SYSTEM_PROMPT` there has the same text), including its leading and trailing newline. The markers are plain strings, not tokenizer special tokens; the prompt writes `</image_end>` and the data writes `<image_end>`, exactly as the ThinkMorph-7B checkpoint was trained.
 
@@ -95,14 +97,14 @@ The system prompt is ThinkMorph's `VLM_THINK_SYSTEM_PROMPT` from its `inferencer
 
 `meta` holds the per-row sidecar: task, `scene_id` and `geometry_uid` (the scene and geometry keys; the split key is named in the split paragraph below), `trajectory_id` (a camera-path or sample label, empty where the source has none), `num_steps`, `num_input_images`, `answer_type`, `answer_value`, `majority_class_rate`, `target_image_kind`, `target_px`, `est_tokens`, licence, `split` (`train` / `validation`, the Hub split names), `supervision_kind` (`full_interleaved` / `visual_aux` / `visual_only`) and `filter_flags`. `majority_class_rate` is the share of the task's most frequent `answer_value` among its training rows: it measures answer skew and is not a guessing baseline (where a task mixes question types or each row has its own options it can be far below chance); compare scores with the text-only baselines below.
 
-Per-row `license` in `meta`: cc-by-4.0 7,015.
+Per-row `license` in `meta`: cc-by-4.0 6,807.
 
 Flags on released rows (`filter_flags` in `meta` and `preview`, comma-separated):
 
 | flag | rows | meaning |
 |---|---:|---|
-| `S5.replay_unsupported` | 7,015 | no solver re-derives this task's answer from the trace, so S5 did not replay it |
-| `S14.sampled_qa` | 200 | chosen for the S14 human spot-check (`reports/s14_sample.tsv`) |
+| `S5.replay_unsupported` | 6,807 | no solver re-derives this task's answer from the trace, so S5 did not replay it |
+| `S14.sampled_qa` | 196 | chosen for the S14 human spot-check (`reports/s14_sample.tsv`) |
 
 ### Training with a BAGEL-family loader
 
@@ -141,7 +143,7 @@ json.dump(info, open(os.path.join(root, "vdrop", "parquet_info_abs.json"), "w"))
 
 The Hugging Face cache (`.../snapshots/<hash>/train/`) or a folder named `spatial-mmcot-vdrop` matches no key.
 
-`num_used_data` counts chunk files, not rows: the loader repeats this source's file list up to that number, lists every (file, row group) pair, and deals whole row groups out, floor(R / world_size) to each rank and floor(that / num_workers) to each DataLoader worker. The remainder is never read. This source has 1 training chunk file holding 53 row groups of up to 128 rows, so keep `num_used_data` large, e.g. the 128 of ThinkMorph's `interleaved_reasoning.yaml` (upstream's `example.yaml` asks for more than GPUs x workers); every row group is then read. Set to 1 and alone in its group on 8 GPUs with 4 workers, it reads only 32 of the 53 row groups. In a run that mixes sources, give each source the same multiple of its own training chunk-file count, e.g. 128 per file (128 here): the file list is repeated up to `num_used_data` entries, so a flat 128 for every source would read a two-file source's rows half as often as a one-file source's.
+`num_used_data` counts chunk files, not rows: the loader repeats this source's file list up to that number, lists every (file, row group) pair, and deals whole row groups out, floor(R / world_size) to each rank and floor(that / num_workers) to each DataLoader worker. The remainder is never read. This source has 1 training chunk file holding 52 row groups of up to 128 rows, so keep `num_used_data` large, e.g. the 128 of ThinkMorph's `interleaved_reasoning.yaml` (upstream's `example.yaml` asks for more than GPUs x workers); every row group is then read. Set to 1 and alone in its group on 8 GPUs with 4 workers, it reads only 32 of the 52 row groups. In a run that mixes sources, give each source the same multiple of its own training chunk-file count, e.g. 128 per file (128 here): the file list is repeated up to `num_used_data` entries, so a flat 128 for every source would read a two-file source's rows half as often as a one-file source's.
 
 ## How the rows were chosen
 
@@ -152,7 +154,8 @@ The Hugging Face cache (`.../snapshots/<hash>/train/`) or a folder named `spatia
 | after conversion and per-row filters | 7,570 |
 | removed by S10 (none) | 0 |
 | removed by answer-prior balancing (S13) | 555 |
-| **released** | **7,015** |
+| removed for training after the r4 rewrite (`subset_dropped.jsonl`) | 208 |
+| **released** | **6,807** |
 
 The rows refused before conversion are perspective-taking questions asked relative to "your partner's perspective": most name no image, so the partner's viewpoint cannot be recovered from the two input views, and the rest are dropped too, as the upstream authors' own build script does.
 
@@ -163,6 +166,7 @@ Every removed row has one line, with its reason, in `reports/`:
 | `build/dropped.jsonl` | S0raw | `S3.partner_phrasing` | 158 |
 | `s13_dropped.jsonl` | S13 | `step: rank` | 514 |
 | `s13_dropped.jsonl` | S13 | `step: letter` | 41 |
+| `subset_dropped.jsonl` | r4 | `reason: label_doubt` | 208 |
 
 Every line of `s13_dropped.jsonl` has `reason: prior_downsample`; `step` names the balancing pass that removed it, and `split` is written `train` or `val` (the Hub's `validation`).
 
@@ -184,7 +188,7 @@ Every line of `s13_dropped.jsonl` has `reason: prior_downsample`; `step` names t
 
 </details>
 
-The train/validation split keeps rows sharing a `scene_id` in `meta` on one side, and the assignment is frozen (`splits/` in the summary repository). S12 saw 7,570 rows under 1,251 keys. No validation input image has the content of a training input image, and none is a pixel-level near-copy of one. S12 does not record per source whether that test ran, but it skips it only for a source whose spec sets `split_leak_pixels: false`, and no spec does; over all sources it compared 21,661 candidate pairs (perceptual hash within 6 bits) pixel by pixel and found no near-copy (checked 2026-09-25).
+The train/validation split keeps rows sharing a `scene_id` in `meta` on one side, and the assignment is frozen (`splits/` in the summary repository). S12 saw 7,570 rows under 1,251 keys. No validation input image has the content of a training input image. The S12 run did not record whether its pixel-level near-copy test ran for this source, so near-copies are not ruled out.
 
 ### Answer-prior balancing (S13)
 

meta/train/chunk_00000.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0a1b8f89343ec6f33970279b64be196c65c3c171aca57e094eddbd4d3f6c00d6
-size 424096
+oid sha256:5cca0ba4e32b8108afb2a5cb474815cfdcddff1d894317ff24315113912df6d9
+size 418604

meta/validation/chunk_00000.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0d9d1377fd054d28d8e8fdb11a5ad0a9e9478c4b0da7945a60ba8e65f7cc7a18
-size 22115
+oid sha256:640c8ed5a63a258ee5b6dbd3f74a3cfe2ca80e965556bfeb102b9a8f906b7a0b
+size 22165

parquet_info.json

@@ -1,10 +1,10 @@
 {
  "vdrop/train/chunk_00000.parquet": {
-  "num_row_groups": 53,
-  "num_rows": 6748
+  "num_row_groups": 52,
+  "num_rows": 6543
  },
  "vdrop/validation/chunk_00000.parquet": {
   "num_row_groups": 3,
-  "num_rows": 267
+  "num_rows": 264
  }
 }

preview/train/chunk_00000.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b6539888275754274b0443f4b80ee99294d87a929398c39d5ee54a3a70e45c11
-size 206103898
+oid sha256:1894f4195964829ae8c563865d19affa9d82919e66645b5a7c7bdeb0c147e2d4
+size 207442657

preview/validation/chunk_00000.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:74c1eb9bac4d077c9515bd9bea76f551d030e68e468675ae8f2beda945fc6dea
-size 7834388
+oid sha256:4d958d91493461a02a9d425ff94549c5070b0dc612da25a73064336c8805746f
+size 7956935

reports/export_manifest.json

@@ -24,14 +24,14 @@
  "max_k": 1,
  "preview_max_k": 1,
  "rows": {
-  "train": 6748,
-  "validation": 267
+  "train": 6543,
+  "validation": 264
  },
  "targets": {
-  "train": 6748,
-  "validation": 267
+  "train": 6543,
+  "validation": 264
  },
- "distinct_targets_total": 1158,
+ "distinct_targets_total": 1152,
  "num_input_images": {
   "train": [
    2,
@@ -53,60 +53,60 @@
   ]
  },
  "distinct_targets": {
-  "train": 1115,
+  "train": 1109,
   "validation": 43
  },
  "supervision_kind": {
   "train": {
-   "visual_only": 6748
+   "visual_only": 6543
   },
   "validation": {
-   "visual_only": 267
+   "visual_only": 264
   }
  },
  "source_scene_corpus": {
   "train": {
-   "infinigen": 6748
+   "infinigen": 6543
   },
   "validation": {
-   "infinigen": 267
+   "infinigen": 264
   }
  },
  "filter_flags": {
   "train": {
-   "S5.replay_unsupported": 6748,
-   "S14.sampled_qa": 194
+   "S5.replay_unsupported": 6543,
+   "S14.sampled_qa": 190
   },
   "validation": {
-   "S5.replay_unsupported": 267,
+   "S5.replay_unsupported": 264,
    "S14.sampled_qa": 6
   }
  },
  "answer_type": {
   "train": {
-   "mcq_letter": 6748
+   "mcq_letter": 6543
   },
   "validation": {
-   "mcq_letter": 267
+   "mcq_letter": 264
   }
  },
  "answer_tag": {
   "train": {
    "mcq_letter": {
-    "rows": 6748,
+    "rows": 6543,
     "equals_answer_value": 0
    }
   },
   "validation": {
    "mcq_letter": {
-    "rows": 267,
+    "rows": 264,
     "equals_answer_value": 0
    }
   }
  },
  "options_text_rows": {
-  "train": 6748,
-  "validation": 267
+  "train": 6543,
+  "validation": 264
  },
  "text_checks": {
   "train": {
@@ -130,10 +130,10 @@
  },
  "row_license": {
   "train": {
-   "cc-by-4.0": 6748
+   "cc-by-4.0": 6543
   },
   "validation": {
-   "cc-by-4.0": 267
+   "cc-by-4.0": 264
   }
  },
  "removed": {
@@ -166,43 +166,43 @@
   "shared_target_images": 0
  },
  "row_uid_sha1": {
-  "train": "221680fcf27e83cc879b7b13621134ca04c78084",
-  "validation": "cf79d8808581cbd833e3f30295452b0e41497645"
+  "train": "69db56563a8d8b94ee5a827f193e495bc88d1bfa",
+  "validation": "339d0897f46e326cb103cfb37493903e9a1f7bf2"
  },
  "tasks": {
   "anchor": {
-   "train": 690,
-   "validation": 21
+   "train": 657,
+   "validation": 20
   },
   "counting": {
-   "train": 590,
+   "train": 570,
    "validation": 37
   },
   "perspective_taking": {
-   "train": 1245,
-   "validation": 49
+   "train": 1218,
+   "validation": 48
   },
   "relative_direction": {
-   "train": 1399,
+   "train": 1330,
    "validation": 50
   },
   "relative_distance": {
-   "train": 2824,
-   "validation": 110
+   "train": 2768,
+   "validation": 109
   }
  },
  "answer_counts": {
   "anchor/train": {
-   "window with green curtains": 10,
+   "window with green curtains": 9,
    "blue shelf": 8,
    "window with black curtains": 7,
-   "window with brown shutters": 7,
-   "gray plant container": 7,
-   "green cabinet": 7,
+   "window with brown shutters": 6,
+   "gray plant container": 6,
    "white shelf": 6,
-   "light brown cabinet": 6,
+   "green cabinet": 6,
    "green shelf": 5,
    "red cabinet": 5,
+   "light brown cabinet": 5,
    "brown door": 5,
    "brown plant container": 5,
    "brown shelf": 5,
@@ -224,28 +224,23 @@
    "oven located next to a blue shelf": 3,
    "white desk": 3,
    "black door": 3,
-   "lamp near a computer monitor": 3,
    "beige lamp": 3,
    "window with brown curtains": 3,
    "gray toilet": 3,
    "bed with green bedframe": 3,
    "green sofa": 3,
-   "bed with black bedframe": 3,
    "oven located next to a yellow cabinet": 2,
    "fridge located next to an oven": 2,
    "oven located next to a green plant container": 2,
    "short dark brown shelf": 2,
-   "short blue shelf": 2,
    "white cabinet": 2,
    "window near an oven": 2,
-   "white door located next to a red shelf": 2,
    "cabinet near a window with orange curtains": 2,
    "shelf on which a beige plant container is kept": 2,
    "white shelf located next to a grey plant container": 2,
    "oven located next to a light brown shelf": 2,
    "window with brown curtains near a gray plant container": 2,
    "fridge located next to a dark brown shelf": 2,
-   "window with blue curtains": 2,
    "beige sink located next to a window with brown shutters": 2,
    "cabinet located next to a beige sink": 2,
    "gray toilet next to a green cabinet": 2,
@@ -255,14 +250,13 @@
    "door located next to a dark brown shelf": 2,
    "gold TV near an orange door": 2,
    "window located next to a brown plant container": 2,
-   "red lamp": 2,
+   "lamp near a computer monitor": 2,
    "yellow shelf located next to a grey plant container": 2,
    "white door": 2,
    "window with pink curtains": 2,
    "window near a light brown cabinet": 2,
    "blue sofa located next to a black door": 2,
    "blue sofa near a beige door": 2,
-   "purple sofa near a window with blue curtains": 2,
    "window with blue curtains near a purple sofa": 2,
    "window near a computer monitor": 2,
    "black side table": 2,
@@ -293,7 +287,6 @@
    "plant container near a computer monitor": 2,
    "gray toilet near a gray plant container": 2,
    "grey toilet": 2,
-   "grey toilet located next to a light brown shelf": 2,
    "shelf located near a pink plant container": 2,
    "window located next to a green shelf": 2,
    "oven located next to a yellow shelf": 2,
@@ -302,6 +295,7 @@
    "green sofa located next to a white cabinet": 2,
    "bed with grey bedframe": 2,
    "blue sofa": 2,
+   "bed with black bedframe": 2,
    "window with the brown curtains": 2,
    "black shelf": 2,
    "window with the green curtains": 2,
@@ -311,6 +305,7 @@
    "cabinet located near an oven": 1,
    "tall plant container": 1,
    "oven located near a blue cabinet": 1,
+   "short blue shelf": 1,
    "oven near a red shelf": 1,
    "dark brown cabinet near a black door": 1,
    "window located next to a white shelf": 1,
@@ -319,9 +314,9 @@
    "black door near a white cabinet": 1,
    "window with beige curtains near a blue cabinet": 1,
    "window near a green cabinet": 1,
+   "white door located next to a red shelf": 1,
    "black door near a dark brown cabinet": 1,
    "oven near a black door": 1,
-   "window near a plant container with gray color": 1,
    "oven located next to a white door": 1,
    "red cabinet next to a black door": 1,
    "white shelf near a brown plant container": 1,
@@ -364,14 +359,11 @@
    "window with beige shutters": 1,
    "bed near a purple plant container": 1,
    "bed located near a plant container with purple color": 1,
-   "shelf located next to a window with yellow curtains": 1,
    "window near a sink": 1,
    "wall art located next to a blue sofa": 1,
    "green TV near a black door": 1,
    "door near a beige lamp": 1,
    "yellow sofa located next to a white TV": 1,
-   "beige desk lamp near a brown sofa": 1,
-   "brown sofa near a yellow side table": 1,
    "beige sofa located next to a red cabinet": 1,
    "brown sofa near a beige lamp": 1,
    "yellow desk lamp near a brown door": 1,
@@ -396,10 +388,8 @@
    "black desk lamp near a peach side table": 1,
    "brown sofa located next to a peach side table": 1,
    "brown plant container located next to a wall art": 1,
-   "wall art located next to a brown plant container": 1,
    "blue sofa next to a black door": 1,
    "gray toilet near a white door": 1,
-   "green sink located next to a yellow cabinet": 1,
    "cabinet located next to a gray sink": 1,
    "pink sink located next to a blue shelf": 1,
    "wall art near a yellow cabinet": 1,
@@ -410,6 +400,7 @@
    "window near a purple lamp": 1,
    "desk near a red lamp": 1,
    "computer monitor near a beige lamp": 1,
+   "red lamp": 1,
    "plant container located next to a brown lamp": 1,
    "brown bed with beige bedframe near a window with gray shutters": 1,
    "computer monitor next to a brown plant container": 1,
@@ -443,6 +434,7 @@
    "cyan sofa located next to a grey side table": 1,
    "grey side table located next to a cyan sofa": 1,
    "window with black shutters": 1,
+   "purple sofa near a window with blue curtains": 1,
    "wall art located next to a brown door": 1,
    "beige desk lamp located next to a computer monitor": 1,
    "blue sofa located next to a gray plant container": 1,
@@ -456,7 +448,6 @@
    "gray side table on which a green lamp is kept": 1,
    "purple sofa near a dark brown cabinet": 1,
    "door located next to a white desk": 1,
-   "purple sofa near a black center table": 1,
    "window near a plant container with beige color": 1,
    "light brown shelf located next to a beige plant container": 1,
    "oven located next to a gray plant container": 1,
@@ -503,14 +494,11 @@
    "purple bed near a wall art piece": 1,
    "Cabinet near a window with brown shutters": 1,
    "bed with beige bedframe next to a wall art": 1,
-   "floor lamp located next to a blue bed": 1,
    "bed with blue bedframe next to a beige lamp": 1,
    "bed with a black bedframe near a window with a black curtain": 1,
    "desk next to a bed with brown bedframe": 1,
    "desk located next to a bed with brown bedframe": 1,
    "wall art located next to a beige lamp": 1,
-   "red shelf near a beige lamp": 1,
-   "bed with beige bedframe located next to a brown side table": 1,
    "brown lamp located next to a brown side table": 1,
    "brown side table located next to a brown lamp": 1,
    "desk near a computer monitor": 1,
@@ -532,12 +520,10 @@
    "bed near a pink lamp": 1,
    "window located next to a pink lamp": 1,
    "floor beige lamp near a bed with brown bedframe": 1,
-   "bed with brown bedframe near a beige lamp": 1,
    "bed with olive bedframe located next to a black desk": 1,
    "white shelf near a window with brown curtains": 1,
    "cabinet situated next to a beige lamp": 1,
    "wall art near a beige lamp": 1,
-   "bed with beige bedframe next to a window with beige curtains": 1,
    "floor beige lamp near a white desk": 1,
    "beige floor lamp near a white desk": 1,
    "cabinet located next to a blue bed": 1,
@@ -577,12 +563,10 @@
    "sofa located next to a brown center table": 1,
    "green sofa next to a window with beige curtains": 1,
    "green sofa near a window with beige curtains": 1,
-   "window with yellow shutters": 1,
    "shelf located next to a beige plant container": 1,
    "bed with brown bedframe near a black door": 1,
    "black door near a bed with blue bedframe": 1,
    "floor beige lamp kept next to a dark brown shelf": 1,
-   "computer monitor near a brown plant container": 1,
    "black desk located next to a blue door": 1,
    "bed with green bedframe next to a white desk": 1,
    "white desk near a light brown shelf": 1,
@@ -607,12 +591,12 @@
    "cabinet located next to a gray toilet": 1,
    "black door located near a light brown cabinet": 1,
    "beige sink located next to a black door": 1,
+   "grey toilet located next to a light brown shelf": 1,
    "gray toilet located next to a blue cabinet": 1,
    "bed with yellow bedframe": 1,
    "bed with brown bedframe situated next to a window with beige curtains": 1,
    "bed located next to a white desk": 1,
    "plant container located near a brown door": 1,
-   "shelf located next to a brown door": 1,
    "black door located next to a white shelf": 1,
    "green sink located next to a black plant container": 1,
    "pink door located next to a light brown shelf": 1,
@@ -620,7 +604,6 @@
    "window located near a light brown shelf": 1,
    "wall art located near a gray plant container": 1,
    "beige door located next to a bed with brown bedframe": 1,
-   "red cabinet located next to a window with beige curtains": 1,
    "window with the black curtains located next to a red shelf": 1,
    "black door located next to a white cabinet": 1,
    "oven located next to a dark brown shelf": 1,
@@ -638,7 +621,6 @@
    "black plant container located next to a white shelf": 1,
    "green sofa located near a brown door": 1,
    "purple sofa": 1,
-   "purple sofa located next to a gray centre table": 1,
    "window located near a black desk": 1,
    "bed with black bedframe located next to a beige door": 1,
    "window located next to a white desk": 1,
@@ -653,14 +635,12 @@
    "brown cabinet": 1,
    "bed with beige bedframe": 1,
    "window with the grey curtains": 1,
-   "white desk located next to a bed with white bedframe": 1,
    "short shelf": 1,
    "bed with green bedframe located next to a white shelf": 1,
    "oven located next to a white cabinet": 1,
    "oven located next to a black cabinet": 1,
    "black dishwasher located near a white cabinet": 1,
    "tall cabinet": 1,
-   "white cabinet located next to a window with grey curtains": 1,
    "window with grey curtains": 1,
    "black door located next to an oven": 1,
    "oven located next to a black shelf": 1,
@@ -668,7 +648,7 @@
    "window located next to a green dishwasher": 1,
    "shelf located near a window with beige curtains": 1,
    "oven located near a window with green curtains": 1,
-   "fridge located next to a blue shelf": 1,
+   "window with blue curtains": 1,
    "green door located near a white cabinet": 1,
    "white cabinet located near a green door": 1,
    "window located near an oven": 1,
@@ -681,7 +661,6 @@
   },
   "anchor/validation": {
    "gray plant container located next to a red cabinet": 2,
-   "computer monitor near a beige lamp": 2,
    "cabinet located next to a beige lamp": 2,
    "red cabinet located next to a gray plant container": 1,
    "yellow shelf": 1,
@@ -690,6 +669,7 @@
    "black door located next to a gray lamp": 1,
    "purple lamp": 1,
    "blue lamp": 1,
+   "computer monitor near a beige lamp": 1,
    "door located next to a red cabinet": 1,
    "cabinet near a small wall art": 1,
    "wall art near a red cabinet": 1,
@@ -700,10 +680,10 @@
    "bed located next to a brown shelf": 1
   },
   "counting/train": {
-   "1": 177,
-   "2": 138,
-   "3": 108,
-   "4": 94,
+   "1": 163,
+   "2": 134,
+   "3": 107,
+   "4": 93,
    "5": 58,
    "6": 11,
    "7": 4
@@ -718,33 +698,33 @@
    "7": 1
   },
   "perspective_taking/train": {
-   "left": 330,
-   "right": 303,
-   "front-left": 221,
-   "front-right": 196,
-   "front": 115,
+   "left": 325,
+   "right": 299,
+   "front-left": 215,
+   "front-right": 189,
+   "front": 111,
    "behind-left": 40,
-   "behind-right": 31,
+   "behind-right": 30,
    "behind": 9
   },
   "perspective_taking/validation": {
    "front-right": 13,
    "left": 10,
    "front-left": 9,
-   "right": 9,
+   "right": 8,
    "front": 6,
    "behind-left": 1,
    "behind-right": 1
   },
   "relative_direction/train": {
-   "left": 381,
-   "right": 342,
-   "front-left": 245,
-   "front-right": 223,
-   "front": 116,
-   "behind-right": 42,
-   "behind-left": 38,
-   "behind": 12
+   "left": 365,
+   "right": 329,
+   "front-left": 229,
+   "front-right": 212,
+   "front": 111,
+   "behind-right": 39,
+   "behind-left": 36,
+   "behind": 9
   },
   "relative_direction/validation": {
    "right": 13,
@@ -756,9 +736,9 @@
    "behind-left": 1
   },
   "relative_distance/train": {
-   "brown door": 85,
-   "window with the brown curtains": 73,
-   "window with the beige curtains": 63,
+   "brown door": 84,
+   "window with the brown curtains": 67,
+   "window with the beige curtains": 61,
    "white shelf": 60,
    "black door": 49,
    "white desk": 48,
@@ -766,63 +746,62 @@
    "window with the green curtains": 35,
    "dark brown cabinet": 34,
    "tall shelf": 34,
-   "black plant container": 33,
-   "brown plant container": 33,
+   "black plant container": 32,
+   "brown plant container": 32,
    "window with the black curtains": 32,
-   "window with brown shutters": 32,
    "dark brown shelf": 32,
-   "short shelf": 31,
-   "window with beige curtains": 30,
+   "window with brown shutters": 30,
    "tall cabinet": 30,
    "blue shelf": 30,
-   "window with brown curtains": 28,
-   "green shelf": 28,
+   "short shelf": 29,
+   "window with beige curtains": 28,
    "green cabinet": 27,
+   "green shelf": 27,
+   "window with brown curtains": 26,
    "window with the white curtains": 26,
    "white cabinet": 22,
    "window with the grey curtains": 21,
-   "bed with brown bedframe": 21,
-   "yellow cabinet": 19,
+   "bed with brown bedframe": 20,
    "grey plant container": 19,
    "window with black curtains": 18,
+   "yellow cabinet": 18,
    "brown door located next to a light brown shelf": 17,
-   "window with the purple curtains": 17,
-   "brown cabinet": 17,
-   "short white shelf": 16,
-   "window with green curtains": 15,
+   "window with the purple curtains": 16,
+   "short white shelf": 15,
    "brown door located next to an oven": 15,
    "gray plant container": 14,
    "light brown shelf": 14,
    "door located next to a white shelf": 14,
    "tall green shelf": 14,
    "white door located next to an oven": 14,
-   "grey door": 14,
-   "window located next to a white cabinet": 13,
+   "brown cabinet": 14,
+   "window with green curtains": 13,
    "window with the orange curtains": 13,
    "black desk": 13,
    "window with white shutters": 13,
    "brown shelf": 13,
-   "window with the gray curtains": 12,
-   "yellow shelf": 12,
+   "grey door": 13,
+   "window located next to a white cabinet": 12,
    "short plant container": 12,
    "shelf located next to a wall art": 12,
    "beige door": 12,
    "white cabinet located next to a window with beige curtains": 12,
-   "bed with gray bedframe": 12,
-   "green plant container": 11,
+   "window with the gray curtains": 11,
+   "yellow shelf": 11,
    "bed with green bedframe": 11,
    "wall art located next to a green shelf": 11,
    "black desk located next to a window with grey curtains": 11,
    "black cabinet": 11,
+   "bed with gray bedframe": 11,
    "window located next to a white desk": 10,
-   "brown door located near a white desk": 10,
    "window located near a white desk": 10,
    "short red shelf": 10,
+   "green plant container": 10,
    "window with the yellow curtains": 10,
    "black door located next to a bed with purple bedframe": 10,
    "window located near a gray plant container": 10,
+   "brown door located near a white desk": 9,
    "shelf located next to a window with brown curtains": 9,
-   "window with the blue curtains": 9,
    "grey side table": 9,
    "brown sofa located next to an orange side table": 9,
    "oven located next to a brown plant container": 9,
@@ -830,13 +809,12 @@
    "blue cabinet": 9,
    "oven located next to a brown door": 9,
    "white desk on which a computer monitor is kept": 9,
-   "window with white curtains": 9,
    "computer monitor located next to a black bed": 8,
    "window with gray curtains": 8,
    "white desk located near a brown door": 8,
-   "cabinet located next to a brown door": 8,
    "gray plant container located next to a blue shelf": 8,
    "black door located next to an oven": 8,
+   "window with the blue curtains": 8,
    "gray sofa": 8,
    "white shelf located next to a brown door": 8,
    "blue plant container located next to a bed with orange bedframe": 8,
@@ -846,13 +824,14 @@
    "fridge located next to a black door": 8,
    "computer monitor kept on a white desk": 8,
    "cabinet located near an oven": 8,
+   "window with white curtains": 8,
    "window located next to a gray plant container": 8,
+   "cabinet located next to a brown door": 7,
    "black plant container near a sink": 7,
    "shelf located next to a brown plant container": 7,
    "beige floor lamp located next to a grey plant container": 7,
    "blue plant container located next to a bed with an orange bedframe": 7,
    "shelf located near a window with beige curtains": 7,
-   "window located next to a wall art": 7,
    "shelf located next to an oven": 7,
    "bed with black bedframe": 7,
    "window with brown shutters located next to a bed with brown bedframe": 7,
@@ -863,15 +842,13 @@
    "bed near a brown plant container": 7,
    "grey plant container located next to an oven": 7,
    "black plant container located near a sink": 6,
-   "window with orange curtains": 6,
    "orange side table": 6,
-   "window with beige shutters": 6,
-   "blue shelf located next to a white door": 6,
    "short dark brown shelf": 6,
    "white door located next to a dark brown shelf": 6,
    "fridge located next to a brown plant container": 6,
    "window with blue curtains": 6,
    "oven located next to a white shelf": 6,
+   "window located next to a wall art": 6,
    "cabinet located next to an oven": 6,
    "plant container near a window with blue curtains": 6,
    "short shelf located next to a brown door": 6,
@@ -884,12 +861,13 @@
    "gray door": 6,
    "window with beige curtains located next to a white cabinet": 6,
    "green plant container located next to a white shelf": 6,
-   "window with purple curtains": 5,
    "white desk located next to a window with beige curtains": 5,
    "bed located near a window with green curtains": 5,
    "grey plant container near a dark brown cabinet": 5,
    "window located next to a dark brown cabinet": 5,
    "window with the gray curtains located next to a gray plant container": 5,
+   "window with orange curtains": 5,
+   "window with beige shutters": 5,
    "grey plant container located next to a dark brown shelf": 5,
    "oven located near a window with black curtains": 5,
    "plant container located next to a fridge": 5,
@@ -899,6 +877,7 @@
    "window located near a white shelf": 5,
    "bed with brown bedframe located next to a window with brown shutters": 5,
    "brown floor lamp": 5,
+   "blue shelf located next to a white door": 5,
    "window located near a brown shelf": 5,
    "cabinet near an oven": 5,
    "black shelf located next to a white cabinet": 5,
@@ -908,12 +887,12 @@
    "pink floor lamp near a window with brown curtains": 5,
    "shelf located next to a white desk": 5,
    "beige floor lamp located next to a white cabinet": 5,
-   "brown cabinet located near a window with black curtains": 5,
    "white shelf located next to a window with grey curtains": 5,
    "black door located next to a black desk": 5,
    "black desk located next to a black door": 5,
    "peach cabinet": 5,
    "cabinet located near a white shelf": 5,
+   "window with purple curtains": 4,
    "bed near a window with green curtains": 4,
    "white desk near a brown door": 4,
    "window with black curtains located next to a yellow cabinet": 4,
@@ -950,6 +929,7 @@
    "wall art located next to a window with white curtains": 4,
    "window with white curtains located next to a wall art": 4,
    "window with brown curtains located next to a black plant container": 4,
+   "brown cabinet located near a window with black curtains": 4,
    "oven located next to a grey plant container": 4,
    "black side table": 4,
    "peach dishwasher": 4,
@@ -974,11 +954,9 @@
    "window located near an oven": 3,
    "white shelf located next to a window with brown curtains": 3,
    "blue sofa located next to a brown side table": 3,
-   "blue sofa located next to a beige door": 3,
    "cabinet located next to a beige door": 3,
    "door located next to a green cabinet": 3,
    "beige sofa": 3,
-   "purple plant container located next to a green shelf": 3,
    "window located next to a red shelf": 3,
    "black door located next to a dark brown shelf": 3,
    "window with grey curtains": 3,
@@ -993,16 +971,13 @@
    "oven near a window with black curtains": 3,
    "window with black curtains located near a grey plant container": 3,
    "oven located next to a black cabinet": 3,
-   "desk located near a gray plant container": 3,
    "desk located near a green plant container": 3,
    "pink floor lamp located near a window with brown curtains": 3,
    "desk located next to a window with orange shutters": 3,
    "window near a brown plant container": 3,
-   "wall art located next to a window with the white curtains": 3,
    "brown cabinet located next to a black door": 3,
    "purple plant container located next to a window with brown curtains": 3,
    "black plant container located next to a window with brown curtains": 3,
-   "brown cabinet near a window with black curtains": 3,
    "cabinet located next to a black plant container": 3,
    "pink plant container": 3,
    "black cabinet located next to a window with blue curtains": 3,
@@ -1058,6 +1033,7 @@
    "door located next to a blue sofa": 2,
    "black door located next to a green cabinet": 2,
    "beige floor lamp located next to a bed with a black bedframe": 2,
+   "purple plant container located next to a green shelf": 2,
    "shelf located next to a purple lamp": 2,
    "desk located near a window with beige curtains": 2,
    "window near a white shelf": 2,
@@ -1079,10 +1055,12 @@
    "shelf located next to a fridge": 2,
    "gray plant container located near a black desk": 2,
    "gray plant container near a black desk": 2,
+   "desk located near a gray plant container": 2,
    "brown sofa": 2,
    "green plant container near a window with brown curtains": 2,
    "light brown shelf located next to a window with the brown curtains": 2,
    "grey door near a bed with green bedframe": 2,
+   "wall art located next to a window with the white curtains": 2,
    "black shelf located near a window with white curtains": 2,
    "window with the white curtains located next to a wall art": 2,
    "black door located next to a brown cabinet": 2,
@@ -1094,7 +1072,6 @@
    "oven located next to a black shelf": 2,
    "window located near a brown cabinet": 2,
    "brown cabinet located near a window with the black curtains": 2,
-   "grey door located next to an oven": 2,
    "purple door located next to a white cabinet": 2,
    "window with white curtains near a brown shelf": 2,
    "oven located next to a brown shelf": 2,
@@ -1106,7 +1083,6 @@
    "black desk located next to a black cabinet": 2,
    "bed located next to a white shelf": 2,
    "orange cabinet": 2,
-   "black desk located next to a window with white curtains": 2,
    "black cabinet located next to a window with the blue curtains": 2,
    "gray plant container located next to a window with brown curtains": 2,
    "cabinet near a white shelf": 2,
@@ -1193,6 +1169,7 @@
    "shelf situated next to a window with blue curtains": 1,
    "shelf located next to a window with blue curtains": 1,
    "tall cabinet located next to a beige door": 1,
+   "blue sofa located next to a beige door": 1,
    "red cabinet located next to a brown door": 1,
    "door situated next to a green cabinet": 1,
    "door located next to a green sofa": 1,
@@ -1228,7 +1205,6 @@
    "window with the white curtains located next to a white cabinet": 1,
    "cabinet situated near an oven": 1,
    "oven located next to a window with the brown curtains": 1,
-   "grey door located next to a black shelf": 1,
    "grey plant container located near a window with black curtains": 1,
    "white door located near a white cabinet": 1,
    "brown door situated next to an oven": 1,
@@ -1271,8 +1247,9 @@
    "window with blue curtains located next to a black shelf": 1,
    "white cabinet located next to a green door": 1,
    "window located next to an oven": 1,
-   "brown cabinet positioned near a window with black curtains": 1,
+   "brown cabinet near a window with black curtains": 1,
    "window with the white curtains located next to a brown shelf": 1,
+   "grey door located next to an oven": 1,
    "brown dishwasher": 1,
    "door located near a wall art": 1,
    "brown shelf located next to a grey door": 1,
@@ -1294,6 +1271,7 @@
    "window with yellow curtains": 1,
    "window with white curtains located near a white cabinet": 1,
    "gray plant container located near a window with white curtains": 1,
+   "black desk located next to a window with white curtains": 1,
    "black desk situated next to a window with white curtains": 1,
    "white cabinet near a window with the white curtains": 1,
    "bed with a beige bedframe": 1,
@@ -1340,19 +1318,18 @@
    "white shelf": 2,
    "window near a dark brown shelf": 1,
    "black desk located next to a yellow shelf": 1,
-   "window with grey curtains": 1,
-   "bed with brown bedframe located next to a grey lamp": 1
+   "window with grey curtains": 1
   }
  },
  "majority_class_rate_train": {
-  "anchor": 0.014492753623188406,
-  "relative_direction": 0.272337383845604,
-  "counting": 0.3,
-  "relative_distance": 0.03009915014164306,
-  "perspective_taking": 0.26506024096385544
+  "anchor": 0.0136986301369863,
+  "relative_direction": 0.2744360902255639,
+  "counting": 0.28596491228070176,
+  "relative_distance": 0.030346820809248554,
+  "perspective_taking": 0.26683087027914615
  },
  "s14_sample": {
-  "rows": 200,
+  "rows": 196,
   "target": 200,
   "replay_unsupported": true
  },
@@ -1763,5 +1740,32 @@
   "MIN_CLASS_SHARE": 0.02,
   "PURE_PRIOR": 0.9,
   "BLIND_MARGIN": 0.15
+ },
+ "est_tokens": {
+  "train": [
+   8047,
+   8872,
+   10402
+  ],
+  "validation": [
+   8093,
+   8712,
+   9750
+  ]
+ },
+ "subset": {
+  "parent_export": "v1_r4",
+  "parent_row_uid_sha1": {
+   "train": "221680fcf27e83cc879b7b13621134ca04c78084",
+   "validation": "cf79d8808581cbd833e3f30295452b0e41497645"
+  },
+  "dropped": {
+   "label_doubt": 208
+  },
+  "dropped_rows": 208,
+  "repaired": {
+   "backspace_boxed": 0
+  },
+  "git_sha": "996b875a1a62fb87e9307abfa1ee59d8a5c980a8"
  }
 }

reports/rewrite_summary.json

@@ -0,0 +1,61 @@
+{
+ "missing_policy": "keep",
+ "rewritten": 7015,
+ "missing": 0,
+ "rows_out": 7015,
+ "sources": [
+  "/media/ruy032/lvr/annotate/merge/full_r3_00.slim.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_00b.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_00b2.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_01.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_01b.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_01b2.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_01b4.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_01b5.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_01b6.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_02.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_02b.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_02b3.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_02b4.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_02b5.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_02c.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_03.slim.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_03b.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_04.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_05.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_06.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_06b.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_07.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_07b.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_sweep0.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_sweep1.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_sweep2.jsonl",
+  "/media/ruy032/lvr/annotate/merge/full_r3_sweep3.jsonl"
+ ],
+ "files": {
+  "train": {
+   "chunk_00000.parquet": {
+    "rows_in": 6748,
+    "rewritten": 6748,
+    "missing": 0,
+    "rows_out": 6748
+   }
+  },
+  "validation": {
+   "chunk_00000.parquet": {
+    "rows_in": 267,
+    "rewritten": 267,
+    "missing": 0,
+    "rows_out": 267
+   }
+  },
+  "preview": {
+   "train/chunk_00000.parquet": {
+    "rows_out": 6748
+   },
+   "validation/chunk_00000.parquet": {
+    "rows_out": 267
+   }
+  }
+ }
+}

reports/s14_sample.tsv

@@ -49,7 +49,6 @@ f8b5dbf5e1b2cf54	relative_direction	train
 96e56c5e42e47d04	relative_distance	train
 673ad07982271f9d	counting	train
 58f20df0d6342239	relative_distance	train
-4e134c27d964a029	relative_direction	train
 d8f20cddf86f4baf	relative_direction	train
 69a8bde3af3a9c07	relative_distance	train
 24a6f405a44935dc	anchor	train
@@ -172,7 +171,6 @@ bd089063484f1ed6	relative_direction	train
 40aa7f2cbae27ef5	relative_distance	train
 efc3487673eb661c	relative_direction	train
 ac39c99f8774feaf	relative_direction	train
-a013e61d0f3b31af	relative_distance	train
 041be981091a7ffa	anchor	validation
 914a2c12b09be68b	perspective_taking	train
 3aa1c0a0ca8a39cb	perspective_taking	train
@@ -185,7 +183,6 @@ a137f8fdffe43480	anchor	train
 d8ab73cdb9bd60b6	relative_direction	train
 de9a31a95033ff57	relative_distance	train
 37dd6fa62e64497d	relative_distance	train
-5d83d7ca86802a93	relative_direction	train
 e28775008dfe4d09	anchor	train
 533117830bd6bc28	perspective_taking	train
 91c14f1da760822e	perspective_taking	train
@@ -197,5 +194,4 @@ c530f564a5edcab5	perspective_taking	train
 31a6f72cd9078e41	perspective_taking	train
 f57c31161cd6bdaa	perspective_taking	train
 0dcdd416d652ac9b	anchor	train
-53e55fa98291ce28	anchor	train
 91f5a42de0dbf70b	anchor	train

reports/subset.json

@@ -0,0 +1,37 @@
+{
+ "parent": "/media/ruy032/lvr/export/v1_r4",
+ "sources": [
+  "ipt_mvc",
+  "ipt_pet",
+  "motif",
+  "thinkmorph_jigsaw",
+  "thinkmorph_spatial_nav",
+  "vdrop",
+  "zebra_jigsaw",
+  "zebra_multihop"
+ ],
+ "drop_flags": [
+  "R.no_rewrite"
+ ],
+ "budget": 24576,
+ "drop_files_sha1": {
+  "/media/ruy032/lvr/annotate/r4_label_doubts.json": "a52ff5510740b5a8aea6eab47d5d31c7f68207ed"
+ },
+ "lists": {
+  "drop_doubts": "/media/ruy032/lvr/annotate/r4_label_doubts.json",
+  "drop_list": []
+ },
+ "git_sha": "996b875a1a62fb87e9307abfa1ee59d8a5c980a8",
+ "this_source": {
+  "source": "vdrop",
+  "rows": {
+   "train": 6543,
+   "validation": 264
+  },
+  "dropped": {
+   "label_doubt": 208
+  },
+  "repaired": 0,
+  "est_tokens_max": 10402
+ }
+}

reports/subset_dropped.jsonl

@@ -0,0 +1,208 @@
+{"row_uid": "535863599d26d589", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The supplied answer C (right) conflicts with the panorama, which places the window to the left of the second view's forward direction."}
+{"row_uid": "01d65e5cb09e842d", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama places the red window front-left of the second viewpoint, conflicting with the behind-right option."}
+{"row_uid": "95ef5f33859e8919", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The first photo also shows a yellow shelving unit in the corner by the dark-curtained window, so a yellow shelf could be said to appear in both views."}
+{"row_uid": "89a59264a5c61f06", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The slatted window also appears in both input frames, so option A is not clearly ruled out by the images."}
+{"row_uid": "40ec0131c768fd8c", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The shared window's curtains look pale beige in the two photos, so calling them green rests mainly on the panorama's tint."}
+{"row_uid": "57c698b9dad2236a", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The shared window's curtains read violet rather than a clear blue, so option A's colour label is only a loose match."}
+{"row_uid": "eb5c24f1c335cf95", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The green item in the shallow white dish could be read as a potted cactus, which would bring the count to three."}
+{"row_uid": "e693ab9a1e87bcc7", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The door appears at the right edge of the second view, which reads as front-right rather than front-left."}
+{"row_uid": "41d0b8a0eca600ca", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The drapes on that window read as beige as much as yellow, so the window with beige curtains also appears in both photos."}
+{"row_uid": "ceef92a7a86252aa", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama places the shelf beside the green door in the front-right of the second view, not in behind-left."}
+{"row_uid": "5d411a242597468e", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "Option A calls the sofa brown, but in both inputs and the panorama the sofa is pink; the lamp beside it is still the only option's object visible in both views."}
+{"row_uid": "d17366c8deea6951", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The sofa appears pink rather than brown, and the yellow side table with the beige lamp of option A also appears in both photos, so the descriptors in B are not fully supported."}
+{"row_uid": "3c69a8bdd1638a97", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama places that door on the left side of the second view, so the supplied option (right) is not supported by the pictures."}
+{"row_uid": "4b30609ef8f3e3ea", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The tan-slatted window also appears in both views, so option A's window is likewise visible in both; separating A from B rests only on the curtain-versus-slats wording."}
+{"row_uid": "46b74658d4f6ecee", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The sink reads as pale white-green in the photos, so the option's word 'green' is only loosely matched."}
+{"row_uid": "b6e07f571907f56c", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama places the gray pot behind-left of the second view, contradicting option D (front-left)."}
+{"row_uid": "77d03e6e2ce0204d", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The white door next to the red cabinet also lies toward the far wall and reads as nearly as distant as the plant."}
+{"row_uid": "36d4ce83e44818e1", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The white desk with the monitor by the blinds window also lies in the overlap of the two photos, so the panorama does not single out the red lamp uniquely."}
+{"row_uid": "3a01648c10a85566", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama places the white door on the left of the second view, which conflicts with the expected right."}
+{"row_uid": "3868ce86e9c9b863", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The supplied option front-right is not supported by the panorama, which places the monitor left of the second view's center."}
+{"row_uid": "b98850c11bc62e91", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "No blue curtains are visible on the window in any picture; it has a dark frame with pale blue glass."}
+{"row_uid": "77ff1d5f8d047b18", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "Both the second photo and the panorama place the black desk on the cabinet-and-shelf side, i.e. the right of the second view, which conflicts with a left label."}
+{"row_uid": "4a6642fd9d40139c", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "In the ring the wooden door is fairly far left of the small window, near the boundary between front-left and behind-left as seen from the second view."}
+{"row_uid": "e9c16d2656ca34ed", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The sofa I match to option C reads as white or cream in all three views, not purple."}
+{"row_uid": "e2ae07983d55b91a", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The second photo appears to show the window and the desk on the right side of the frame, which would suggest right rather than left."}
+{"row_uid": "fbe400d78e2b74c3", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "In the panorama the beige-panel window also appears right beside the brown shutters window, so it could instead be read as the closest option."}
+{"row_uid": "53e55fa98291ce28", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The second photo's frame ends right at the stove, so the plant sits at its extreme edge and is hard to confirm there."}
+{"row_uid": "fe04f68d0dba9750", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The panorama shows four cabinets while the provided correct answer is three."}
+{"row_uid": "9d8b45829803d861", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The left-right assignment of the second view within the panorama is not fully unambiguous to me, so I am not certain the beige bed falls on its left rather than its right."}
+{"row_uid": "50900ca450330621", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The white-curtain window is visible within the front of the second photo, so the photos place it in front rather than behind-right."}
+{"row_uid": "8a230648c2c383ab", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "In the second photo the pink door actually appears at the left edge of the frame, which seems to contradict reading it as right."}
+{"row_uid": "2ab507831152e689", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The bed beside that lamp appears pink rather than blue in all the pictures, so the option's color does not match what is visible."}
+{"row_uid": "8794d683a74e153f", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The light brown shelf also appears in both photos, so the panorama does not uniquely single out option C."}
+{"row_uid": "f04cf07c3d7dd05f", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "Option A calls the frame beige, but the bed's frame reads as dark green in every image."}
+{"row_uid": "ad76162ab7abc65e", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The second photo places the beige-curtained window at the left of the view, so the pictures support (D) left rather than the stated label (B) right."}
+{"row_uid": "4289fc80cd7d1a0c", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The brown-shutter window of option C is visible only in the first photo, not the second, so it cannot be the object seen from both perspectives."}
+{"row_uid": "4e2e25acb9767903", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The panorama places the black plant container at the far end of the low cabinet, farther from the light brown cabinet than the monitor on the black desk, so the picture supports A rather than the stated B."}
+{"row_uid": "66bac019ed6a57be", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "Option A calls the bed frame brown, though the panorama shows a pink frame with only the mattress brown."}
+{"row_uid": "78ea7463e13e9c0d", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "In the panorama the gray pot sits on the far side of the bed, so read strictly from the second image's facing it looks more behind-right than cleanly left."}
+{"row_uid": "1b115437e3f47d0e", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The curtain beside the pale-framed bed reads as amber slats rather than beige, so option B's curtain colour is not clearly supported, though that bed is the only one present in both views."}
+{"row_uid": "08bdf489f0f71078", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The second photo crops the green-curtain desk at its extreme right edge, so the lamp and monitor there are only barely inside that frame."}
+{"row_uid": "5d678d63dece1be4", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "In the second photo the brown-curtained window appears left of the frame's center, which would suggest front-left rather than front-right."}
+{"row_uid": "a97d2d199aa1a526", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The expected label A (light brown cabinet) is not supported, because the tan bookcase lies only at the first view's edge and outside the second view's arc."}
+{"row_uid": "59f74370528c734a", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The second photo itself frames the TV near its center, which from that camera alone would read as front rather than left."}
+{"row_uid": "5565bed9373c723f", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The blue plant cabinet also sits under that same framed window in both views, so the window near a blue plant container is not clearly absent from one perspective."}
+{"row_uid": "bdf43606b5dd1abd", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The second photo plainly places the green‑curtained desk on the right of the frame, yet the expected label is left."}
+{"row_uid": "293654736c77f2a5", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The second image also shows a gray-pot plant beside its own beige floor lamp on the right, so the phrase could select that one, which would point right instead."}
+{"row_uid": "a1583ddcd5448fa0", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The desk under the right-hand monitor looks light-toned in the first photo but black in the second."}
+{"row_uid": "5d83d7ca86802a93", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama places the yellow shelf to the left of the white door, so from the second view it reads as behind-left rather than front-right."}
+{"row_uid": "f5bc058a1b8d45a9", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The second photo's own window also has bluish vertical blinds with a small white desk nearby, so the phrase 'blue curtains' could instead pick out that front-left desk."}
+{"row_uid": "4e134c27d964a029", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama places the door beside the white desk on the left, so the expected right is not supported by the image."}
+{"row_uid": "78b103f0769e6cf9", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "In the panorama the door sits between the yellow shelf and the bed, which from the second camera reads as front-right rather than front-left."}
+{"row_uid": "3082e73714c1a400", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The strip places that floor lamp at the front, beside the centered beige window with the bed left and the desks right, not to the left as the expected label says."}
+{"row_uid": "6bdf0f76ad374786", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The pot on that right dark cabinet reads as pink rather than brown in both photos, so the \"brown plant container\" label only loosely matches what the pictures show."}
+{"row_uid": "1bbf578976e2370c", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The second input image itself shows the framed art within its frame at the left edge, which would suggest left rather than behind."}
+{"row_uid": "4c393e88a040a3f3", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The stated correct answer is left, but both the second photo and the panorama place the brown door on the right side of that view's facing direction."}
+{"row_uid": "7e52c97d07b00732", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The panorama reveals a pale-curtained window on the wall hidden from both photos, giving three windows rather than the two the photos alone suggest, so the visible total conflicts with the expected count of two."}
+{"row_uid": "64762da15756b8ae", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "Both views place the dark brown cabinet on the right of the second photo, giving front-right, which conflicts with the indicated front-left."}
+{"row_uid": "941b3f51e7438711", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "Both the second photo and the panorama place the red cabinet to the right of the second viewpoint's forward direction, so the images support front-right rather than the front-left given as correct."}
+{"row_uid": "a436fdde91a1d191", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "In the second photo the door appears right of the frame's center, which reads as front-right rather than front-left, so the pictures conflict with the stated direction."}
+{"row_uid": "0df49025480d5828", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The lamp stands left of the second view's centre, which I read as front-left rather than front-right."}
+{"row_uid": "8921775c2a759826", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The stated correct label is front-right, but both the second photo and the panorama place the purple door to the left of that viewpoint's center, which reads as front-left."}
+{"row_uid": "10a2f8906bd32aa2", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The blue-curtain window sits on the right side of the second view in both the photo and the panorama, which I read as front-right rather than front-left."}
+{"row_uid": "f409841819508f9b", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The drawn panorama places the yellow door at the right end of the wall the second photo faces (front-right), not front-left."}
+{"row_uid": "b4491e09b1625d88", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The toilet appears white in the images, while option A describes it as grey."}
+{"row_uid": "26a6fb5cda9e0a65", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "In the panorama the wrapped-edge dark brown cabinet on the right sits at a similar separation from the cream-curtained window, so the back-wall corner on the door side is read as the farthest rather than being unambiguously measured."}
+{"row_uid": "3b6b48b5d03fc07c", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The window beside the dark brown cabinet sits at the right edge of the panorama, also a long span from the green window, so the picture keeps A and C close."}
+{"row_uid": "bf780fb123c96e25", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The short red shelf stands directly beside the blue one and is also visible in both views, so the pictures support D as well."}
+{"row_uid": "05a24c5040ee6a89", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "In this panorama the floor gap between the grey pot and the toilet looks at least as small as the gap to the shelf, so the picture does not clearly favor the green shelf."}
+{"row_uid": "35cb6954c69bf0aa", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The yellow shelf beside the door is also visible in both photos, so option C could arguably hold as well."}
+{"row_uid": "b31bea7f5a61184b", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The panorama places the grey-curtain window directly in front of the first view, so the supplied label behind-right is not supported by the images."}
+{"row_uid": "b25843ec63e8a241", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The grey-curtain window's near edge also meets the black-curtain window at the corner, so a nearest-edge reading would put it as close as the brown-shutter window."}
+{"row_uid": "56a9820058338e5b", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "In this view the black pot at the blinds window also lies close to the green canvas, so the gap to the brown-shutter window is not visually decisive."}
+{"row_uid": "177bf9922baa0beb", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The dark brown cabinet under the pink curtain is visible in both input views, so option A also seems to qualify."}
+{"row_uid": "8f8a52eb12beda7d", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The panorama places the brown-shutter window between the door and the purple-curtain window, so the images support B rather than the expected A."}
+{"row_uid": "b18f2ad50652a892", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "In the first photo the tall cabinet at the left edge appears nearly as close as the brown-curtained window, so the closeness call rests on the panorama's layout rather than a single clear size cue."}
+{"row_uid": "15c4623acbd28719", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The yellow cabinet by the green door is also visible in both photos, so the evidence does not uniquely single out option A."}
+{"row_uid": "9b1e236fc39e7ba2", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The pale shelf on the left wall is also a short shelf and is far from that window, so the verdict depends on which shelf the option names."}
+{"row_uid": "f29e2f03f8770135", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The low cabinet directly beside the door could itself be read as the dark brown cabinet, which would make it closer than the orange window."}
+{"row_uid": "9eb25c80251b263c", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The red shelf sits just past the corner of the first view's left wall, so it could equally be read as behind-left (D)."}
+{"row_uid": "db98e6ad6d8dda31", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The small green framed panel could also be read as a second piece of wall art, which would make the total 2."}
+{"row_uid": "70c3c21f42ea29aa", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The panorama puts the sofa at the left edge about as far from the beige door as the blue shelf at the right edge, so it does not clearly favor one over the other."}
+{"row_uid": "6f141e2e883a3e2d", "split": "train", "task": "counting", "reason": "label_doubt", "note": "All three views show three doorways, so the supplied option D (1) is not supported by the pictures."}
+{"row_uid": "2659bfb5aeb82384", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The expected option is left, but the panorama places the tall shelf on the wall opposite the first photo's facing wall, i.e., behind that viewpoint."}
+{"row_uid": "cce42b63cb22a6f5", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The panorama places the door next to the white shelf directly behind the sofa, nearer than the yellow shelf, which conflicts with the yellow-shelf answer."}
+{"row_uid": "3836ef1fd751e873", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The panorama actually places the gray sofa between the plant and the beige-shutter window, so the sofa reads as closer than the window."}
+{"row_uid": "8bca45439fa16ae1", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The same white shelf also appears just left of the dark doorway in the panorama, so a left turn from the first view reaches it as well, which the picture could read as (D) left."}
+{"row_uid": "edf8ebe8729206e8", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The curtains appear pale and pink under the warm light rather than blue, but this does not change the left/right placement relative to the second view."}
+{"row_uid": "8c6539ead424abe3", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The second photo shows that same red shelf at its right edge, which reads as right rather than left."}
+{"row_uid": "67c8a0282e8402e0", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The panorama shows four doors, so the picture supports (C) 4 rather than the provided answer (A) 1."}
+{"row_uid": "144d430eea635f81", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The stated correct option (B) 2 is not supported by the panorama, which shows four distinct doors."}
+{"row_uid": "98fed1f64a25857a", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The panorama clearly shows three potted plants, so the indicated option D (1) is not supported by the picture."}
+{"row_uid": "4b72d0a3dd076a3b", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The small yellow unit also shows open cubbies, so it could be read as a second small shelf."}
+{"row_uid": "682d38c8e4c39e1b", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "A strict corner-view projection from the first image would place the light-brown-shelf wall behind-right rather than front-right."}
+{"row_uid": "08fb43ad31ce8d29", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The stated correct answer A is not supported; the panorama places the shelf next to a wall art farther from the fridge than the window next to a wall art."}
+{"row_uid": "0800d9a867a014fc", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama places the window beside the left wall art at front-left and the window beside the right wall art at behind-right, so the supplied front-right is not what the picture shows."}
+{"row_uid": "09686eea6c349fed", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The panorama shows the oven in the immediate foreground of the first view while the yellow cabinet is on the far wall, so the visible evidence favors the oven, option A."}
+{"row_uid": "3a2534a55fa74269", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The provided correct answer was right, but the panorama and the first photo's left-edge sliver both place the blue shelf to the left of the first view."}
+{"row_uid": "823b8de78d47c77f", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The stated answer is front, but the panorama places the potted plant to the right of the first viewpoint, so the picture does not support front."}
+{"row_uid": "c6ce02e6f08f19d2", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The sofa renders as white/cream in both photos rather than the blue named in the options."}
+{"row_uid": "6e1ba7415b9d5c9a", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The panorama renders the sofa large in the right foreground, which could also be read as the sofa being close to the camera rather than far."}
+{"row_uid": "8042356a705f28bb", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The panorama also shows a low cabinet between the plant and the blue shelf, so if that cabinet is the 'short shelf' it would be nearer than the plant."}
+{"row_uid": "cf7e569633fee7c2", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The panorama shows an empty room with no furniture, so it does not depict the sofa or table and cannot by itself support the answer."}
+{"row_uid": "5788b2daa3f4524a", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The panorama I drew shows an unfurnished room with no sofa, so it does not match the furnished room in the two input photos."}
+{"row_uid": "948bd146b7aa251d", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The curtain visible in the panorama reads dark-red rather than white, so the white-curtain window's identity rests on the first photo."}
+{"row_uid": "dea8304c538c51aa", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama renders the room without furniture, so the cabinet's side is fixed by matching the door and doorway across the two photos rather than by seeing it in the panorama."}
+{"row_uid": "bf1376421b3bc664", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The panorama places the curtained window on the wall opposite the picture wall, i.e. behind the second image's viewpoint, so the picture does not support left."}
+{"row_uid": "23a7b8b200d56eca", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The expected label is left, but the second photo and this panorama both place the brown-bedframe bed to the right of the second camera's facing direction."}
+{"row_uid": "c3b83da295ae2213", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The only window with drapes shows green, not white, side curtains, though it is the sole curtained window and the one to the right."}
+{"row_uid": "df8b4295e5ce48ec", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The tall white cabinet is visible in both views as well, so option B is also present in both."}
+{"row_uid": "481544a66f2ce237", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The glossy black box filling image 2's right foreground could be read as the dishwasher seen close up, which would put option A nearest, so the panorama only weakly settles the camera at the door end."}
+{"row_uid": "1a6bd0c7d27403c3", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The panorama places the orange-curtain window at the back-left, not the front-right given as the correct label."}
+{"row_uid": "f89b40094d9f747b", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The window beside the white cabinet appears as a plain pale pane with no visible grey curtains, so that part of option A is not clearly shown."}
+{"row_uid": "7b68370ccb6bc6f1", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The large multi-pane window shows wooden lattice frames rather than brown curtains, so I match option A to it as the only other candidate window visible in the panorama."}
+{"row_uid": "bf55d199831f20cd", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The panorama places the refrigerator at the far right counter rather than beside the blue shelf, so the 'next to' half of option A is not directly visible."}
+{"row_uid": "377ed854824b8c9b", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The desk near the gray plant container stands at the far wall, farther than the white desk, so the picture does not support option C as the nearest object."}
+{"row_uid": "139995afaaacc127", "split": "train", "task": "anchor", "reason": "label_doubt", "note": "The bed frame reads as dark green in the photos rather than strictly black."}
+{"row_uid": "3980048a5fad0610", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "Image 1 shows the orange-framed bed in the immediate foreground, so the photos do not visibly confirm that the unseen beige-curtain window behind the camera is actually nearer than that bed."}
+{"row_uid": "5a1b4fba328a5370", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "In the second photo the curtained window actually sits to the right of the frame's centre, which points to front-right rather than the front-left read from the panorama."}
+{"row_uid": "d7c4b09f0da79b11", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "I do not actually see purple curtains on the window beside the purple bed; it reads as plain glass, so the option's 'purple curtains' label is not visually confirmed."}
+{"row_uid": "b24e29be9b1cb17c", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The two low white cube units also look like open shelving, so the panorama could support a count higher than one."}
+{"row_uid": "b3b32ee5c5364195", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The drawn panorama shows a windowed corridor and a paneled door rather than the floral-walled room, so it does not itself display the tan shelf to confirm its direction."}
+{"row_uid": "e120acdd0604eddc", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The panorama I drew shows a windowed hall and a paneled door rather than the floral-walled room and the pink sofa, so it does not itself display the sofa."}
+{"row_uid": "dacfdafc95c90248", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The 360 panorama shows an empty windowed corridor rather than the furnished room, so it contains no tan shelf and the direction is read from the first image instead."}
+{"row_uid": "f409b7b13ec0af36", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The expected label is left, but the panorama and the second photo both put the orange door on the right of that view."}
+{"row_uid": "4fa58cf2977db851", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The given correct answer is 1, but the panorama clearly shows four separate doorways, so the images do not support that label."}
+{"row_uid": "ce059db735e65e1a", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The door shown in the views and the panorama is dark green, not brown, so the colour in option A does not match the picture."}
+{"row_uid": "daa22303cf337a1c", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "In the second photo the green shelf is drawn at the right edge of the frame, so that photo taken by itself would point to right rather than front."}
+{"row_uid": "035fe0ebf06c8cc3", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The panorama shows five separate window openings, contradicting the stated correct answer D (1)."}
+{"row_uid": "8c00a034d10397d3", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "Reading the panorama literally, the grey-curtained desk lies on the far side of the bed from the door, which from the second viewpoint looks more like behind or left than clearly right."}
+{"row_uid": "8bfe2c5ac538180b", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The panorama's walls are drawn bare of the bed that the two photos clearly show, so the single bed is not itself visible in the full view."}
+{"row_uid": "5291419e148a000e", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The panorama I sketched renders the room with different walls and omits the black door, so the door's bearing relies on the first view rather than the wrap."}
+{"row_uid": "f09ea25e50eefa92", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama places the tall bed just past the left edge of the first view (behind-left), which conflicts with the front-left label."}
+{"row_uid": "281f41de3ee3f1c0", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The stated correct label B (front-right) is not supported: in both the first photo and the panorama the curtained window lies to the left of the first view's direction."}
+{"row_uid": "b67028d80dfe1a5e", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The wrap places the brown shelf on the right of the first view, so the picture does not support the expected left."}
+{"row_uid": "80805e7473cb426b", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The stated key is front-right, but both photos and the panorama place the brown-curtain window to the left of the first view's center, i.e. front-left."}
+{"row_uid": "f37be0573b9d5b05", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The panorama places the black shelf and its window on the back wall behind the first viewpoint, so front-right is not what the picture shows."}
+{"row_uid": "72f2589363f33b09", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The window I match to C shows a brown frame but no clearly visible brown curtains."}
+{"row_uid": "99fa4df47baafe75", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "Option B calls the curtains brown, but in all three pictures they read as dark teal-green, though it is plainly the same single curtained window."}
+{"row_uid": "31979ef73f0e2e8f", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The white cabinet also borders the same curtained window on the left wall, so the two options point at nearly the same window and the distance gap is very tight."}
+{"row_uid": "5220e76cc5a75653", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The curtains read as dark teal-green in the images rather than brown, though it is the only window with curtains so option C refers to it."}
+{"row_uid": "e8efc82e33de7aae", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "Option B's 'brown curtains' do not appear as curtains in the panorama; the back window has a brown frame, so I matched B to it."}
+{"row_uid": "87545eaf6633cf72", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The window nearest the bed shows brown frames rather than visible curtains, and the only curtained window has teal slats, so the option's 'brown curtains' is a loose label."}
+{"row_uid": "7a054c0d67e03fdc", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The curtains on that window read as teal-green rather than the brown stated in option C, and the door reads as brown-textured rather than the grey in the question."}
+{"row_uid": "a6a7600988e483cf", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The panel immediately beside the bed reads as a brown door, and the only curtains I can clearly see are blue on the far window, so calling the bed-side panel a window with brown curtains is uncertain."}
+{"row_uid": "6b3cdf135d22a474", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "Reading the panorama, the white-curtain window falls to the front-right of the second photo's viewpoint rather than the front-left."}
+{"row_uid": "7a9b166362dfc335", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The panorama shows the black shelf directly beside the green-curtained window, clearly closer than the brown cabinet, which conflicts with answer B."}
+{"row_uid": "bf41f1c331e547de", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "My panorama reading makes the brown-curtain window (C) the nearest object, which conflicts with the provided option D (brown cabinet)."}
+{"row_uid": "050880a8c46dca96", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The slats on the leftmost window look dark brown in both photos rather than clearly green, so I am taking the question's wording to identify that window as the green-curtained one."}
+{"row_uid": "26d46b8cb6b088fc", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The stated correct answer C is not supported here: the panorama places the wood cabinet by the brown-blind window on the opposite wall, while the black shelf is directly under the green-draped window."}
+{"row_uid": "d4020a79e7b8590f", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama places the brown-side-table window on the wall opposite the bed, i.e. behind the second view, not to its right as option B would require."}
+{"row_uid": "cf106e8211000890", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The green unit's open bays holding books could be read as a second shelf, which would support 2 rather than 1."}
+{"row_uid": "088f1227eacabe44", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama places the pot on the wall behind the second view, so the picture-supported direction is behind-right rather than front-right."}
+{"row_uid": "d530d0a59e6cf312", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The panorama shows five windows, so it does not match a count of four."}
+{"row_uid": "e5926c52dc0b1009", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The doorway beside the shelves looks dark rather than distinctly grey, so I identify it as the grey door only because the other doorway reads green."}
+{"row_uid": "cfa083f41b394a9a", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The thin bright panel at the panorama's left edge could be read as a third window; I treat it as the seam wrap of the curtained window."}
+{"row_uid": "673c809c6d4bf16b", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The corner shelving looks like two separate boxes and the unit under the window is open-fronted, so a strict count of open shelving could give more than one."}
+{"row_uid": "51657bc144668f71", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The prompt's 'brown curtains' best matches the brown wooden frame of the long window whose slats look grey, so the grey option and the reference window appear to be the same window."}
+{"row_uid": "19c8fcb6bde79f24", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "Both the second photo and the panorama place the brown cabinet on the right side of the second viewpoint, so the visible evidence gives front-right rather than the stated front-left."}
+{"row_uid": "34a404a15384e8a4", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama places the curtained-window cabinet on the right (front-right), not on the front-left that the option key indicates."}
+{"row_uid": "721d972ad4d81729", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The curtain colours named in the options are not clearly visible in the renders, so I matched windows by their positions and the options' labels rather than by observed colour."}
+{"row_uid": "422bfb941c6062f6", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The blue bed sits just past the left edge of the second view, so its bearing could equally be read as behind-left rather than front-left."}
+{"row_uid": "e923bcbd2226998e", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The panorama shows two open shelving units (the wooden bookshelf and the black cubby unit), which conflicts with the expected count of 1."}
+{"row_uid": "eee0e97afe946d6d", "split": "train", "task": "counting", "reason": "label_doubt", "note": "The second photo seems to show a second potted plant at the base of the dark stand, yet the panorama shows only one plant container in that corner."}
+{"row_uid": "15bbcc352e1d6f4b", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The option describes the curtains as black, but in the photos they look pale grey or pinkish."}
+{"row_uid": "085f2363febbf6b8", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "Option B describes the brown cabinet as being near a window with black curtains, yet in the panorama that cabinet sits beside the grey-curtained window."}
+{"row_uid": "50b536a1e16ad17d", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "Option A calls the curtains black, but in both photos and the panorama they look pale grey, so I matched the cabinet by its position beside the curtained window."}
+{"row_uid": "a013e61d0f3b31af", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "Option D's phrase 'near a window with black curtains' does not match the panorama, where the brown cabinet stands in the corner by the door, far from that window."}
+{"row_uid": "86244fab0bf9a15b", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The door next to the oven reads as dark brown in both views rather than grey, though it is the only door there."}
+{"row_uid": "4bee71646d0078b1", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "Because the dishwasher sits at the panorama's right edge, it could alternatively be read as behind the first viewpoint rather than left."}
+{"row_uid": "1ad5ee6ff8ac0508", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The expected label is behind, but the panorama puts the brown shelf on the doorway wall, which the first photo shows at its right, so it reads as right."}
+{"row_uid": "fb1f011871f4bc61", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The cabinet sits right at the left edge of the first view, so behind-left is also a plausible reading."}
+{"row_uid": "2aac52018035900b", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama places the door beyond the left edge of the second view, so it reads as behind-left rather than the expected front-left."}
+{"row_uid": "73d3ad9ea7a42c47", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The windowed wall sits close to the right edge of the second view, so it could plausibly read as behind-right rather than front-right."}
+{"row_uid": "f6862bccd2a5a4a2", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The question says 'brown curtains', but no curtains are visible; I treated the small brown-framed window as the referent."}
+{"row_uid": "3cbf9053dd9dde93", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The panorama places the short shelf directly beside the curtained window, with the purple door and white cabinet on the opposite wall, so the image supports A and B rather than D."}
+{"row_uid": "cea91e233d29e4c0", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The panorama places the short shelf directly beside the curtained window, so the expected answer D (short shelf) is not supported by what the picture shows."}
+{"row_uid": "c82a42d1f15d13e9", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama places the curtained window on the right of the first view, so the expected front-left is not supported by the images."}
+{"row_uid": "997292c2f6befd8f", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The curtained window is also the window next to the white cabinet, so the picture supports B and C equally rather than distinguishing them."}
+{"row_uid": "661ce33d5a47a273", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The brown-frame bed is not visible inside the first photo's frame, which would more naturally place it behind-left rather than front-left."}
+{"row_uid": "0b4181b373dde6b0", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The panorama puts the white-curtained window wall opposite the bed wall, i.e. behind the second viewpoint, not to its right as the expected option states."}
+{"row_uid": "afe05e224694d248", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama puts the black shelf on the rear wrap‑around edge, i.e. behind the second‑photo view, so a rightward direction is not supported by the pictures."}
+{"row_uid": "18a8764584820461", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The panorama places the gray-framed bed beside the white-curtain window and the beige-curtain window at the far end, which contradicts option B."}
+{"row_uid": "4a99a5651bc39e6b", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The bed in option B stands right beside the green-curtain window, so the panorama keeps the bed and the window at almost the same distance from the white-curtain window rather than clearly separating them."}
+{"row_uid": "cb70c725c4ba94ce", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The drape on that far-end window looks pinkish rather than green in the photos, but it is the only curtained window, so option C must refer to it."}
+{"row_uid": "3483cf5c498e8c88", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The white cabinet stands right at the head of that bed, so it could equally be read as the nearest object."}
+{"row_uid": "744cde8877d01a12", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The green bed also stands close to the short white shelf, so which bed the question calls 'next to a white shelf' is not unambiguously settled by the pictures."}
+{"row_uid": "c075a89cdc7e0518", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The panorama places the curtained window just beyond the left edge of the second view, so the picture supports left rather than the expected right."}
+{"row_uid": "b258ed070a11637c", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The door sits close to the rear of the first view, so the panorama leaves left versus behind somewhat ambiguous."}
+{"row_uid": "e3e1b868ad85eac1", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The panorama places the brown-curtain window to the right of the first view's center, not the left, so the pictured evidence supports right rather than left."}
+{"row_uid": "85f7ee334d6f001e", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The panorama puts the shelf beside the pale door on the right-hand wall of the second view, so the picture supports right rather than the given left."}
+{"row_uid": "0c3f563952521971", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "In the first photo the pink-framed doorway is clipped at the left edge rather than centered, so its appearing at the side there is in tension with the panorama placing it in the forward sector."}
+{"row_uid": "8f9ef7a5cff3723d", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The orange slat panel hugs the camera's left edge in the first view as well, so the panorama leaves the desk and the orange window both near the camera without fully separating their closeness."}
+{"row_uid": "c71f500b3f5cfe53", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The panorama places the curtained window on the far side of the bed from the second viewpoint (behind), which conflicts with the front label."}
+{"row_uid": "cc8dfe489db82319", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The expected option is front, but the panorama places the curtained window on the wall opposite the door the second view faces, which reads as behind."}
+{"row_uid": "d93818c1fc534328", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The left window's bright panel reads partly like open sky and the shelf beside it is partly white, so calling it the brown shelf by a white-curtained window is not fully certain."}
+{"row_uid": "48e7b0874b93c954", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "In the second photo the picture sits close to the centre of the view, so front is also a plausible reading."}
+{"row_uid": "add85a8a65367152", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The panorama shows the door as dark purple, not grey, so the 'grey door' label does not match its visible colour."}
+{"row_uid": "808d55dcad0fe4a9", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama places the oven near its right edge, so from the second image's heading it could also be read as just past the right edge (behind-right) rather than front-right."}
+{"row_uid": "abdbae13588252fe", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The panorama places the curtained window front-left of the second viewpoint, not front-right as the provided answer states."}
+{"row_uid": "f8ba6011d954676e", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "Both the second view and the panorama place the curtained window to the left of the washer, i.e. front-left, so the stated front-right is not supported."}
+{"row_uid": "30b3dbf30d090e13", "split": "train", "task": "perspective_taking", "reason": "label_doubt", "note": "The panorama I sketched shows the adjoining dark room, not the pale room, so the brown shelf is not in it and the direction rests on the two photos rather than on that view."}
+{"row_uid": "b1fa1e02bfacf0dc", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The panorama I drew does not contain the plant or the shelf, so the direction is read from the second photo rather than from the panorama."}
+{"row_uid": "240e859760b00378", "split": "train", "task": "relative_distance", "reason": "label_doubt", "note": "The oven also fills the near left edge of image 2, so the pictures do not clearly show the gray-curtained window to be nearer than the oven."}
+{"row_uid": "4d2416530243dbde", "split": "train", "task": "relative_direction", "reason": "label_doubt", "note": "The first photo is too dark to confirm which wall it faces, so the right-side reading rests on matching its faint edge to the black-shelving wall."}
+{"row_uid": "e851f057ee4b85f1", "split": "validation", "task": "anchor", "reason": "label_doubt", "note": "The second photo shows the desk and monitor only at its extreme left edge, so their presence there is read off the panorama rather than clearly seen."}
+{"row_uid": "7cde7468719deece", "split": "validation", "task": "perspective_taking", "reason": "label_doubt", "note": "The panorama places the curtained window at the left of the second image's view, not on the right that option D indicates."}
+{"row_uid": "d31dee39246c9482", "split": "validation", "task": "relative_distance", "reason": "label_doubt", "note": "In the first photo the grey floor lamp looks small on the far wall while the bed is near at the left, which suggests the lamp is farther and conflicts with choosing the bed."}

train/chunk_00000.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1b8b07be753a00aafee8feedb7e8ca3cfb60bee0c6fe6194452effb29b88a358
-size 190603319
+oid sha256:be3ff2b97180dfdd9f88e23cd8f492b33cc1d19aa2c2df25a21c509ca3a6609c
+size 191574442

validation/chunk_00000.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b85bf481a7de97e36c79d7e14cd89361ca313e2fdda471c906c41eb847d3c8f8
-size 7149020
+oid sha256:a8f02f8ac040c196c665a313a63582d62035924b4c2be9e24043be3a4e647699
+size 7193107