Add files using upload-large-folder tool

README.md

@@ -0,0 +1,286 @@
+---
+license: other
+license_name: upstream-undeclared
+license_link: https://huggingface.co/datasets/ThinkMorph/Jigsaw_Assembly
+pretty_name: "Spatial MMCoT v1 - thinkmorph_jigsaw"
+language:
+- en
+task_categories:
+- visual-question-answering
+- image-to-image
+tags:
+- spatial-reasoning
+- interleaved-reasoning
+- multimodal-chain-of-thought
+size_categories:
+- 1K<n<10K
+configs:
+- config_name: preview
+  default: true
+  data_files:
+  - split: train
+    path: preview/train/*.parquet
+  - split: validation
+    path: preview/validation/*.parquet
+- config_name: train
+  data_files:
+  - split: train
+    path: train/*.parquet
+  - split: validation
+    path: validation/*.parquet
+- config_name: meta
+  data_files:
+  - split: train
+    path: meta/train/*.parquet
+  - split: validation
+    path: meta/validation/*.parquet
+---
+
+# Spatial MMCoT v1 · `thinkmorph_jigsaw`
+
+ThinkMorph ([arXiv:2510.27492](https://arxiv.org/abs/2510.27492)) Jigsaw_Assembly: a picture cut into numbered parts, shown in an order that may be shuffled; decide the arrangement, draw the assembled picture, read it back, answer. According to the ThinkMorph paper (arXiv:2510.27492, appendix on data generation), GPT-4.1 wrote the plan and the read-back from the question and the ground-truth answer, and was told not to reveal the answer. The pictures come from 3 corpora: 3,203 rows are SAT renders of ProcTHOR houses (AI2-THOR assets); 1,821 rows are real photographs from ADE20K; 775 rows are real photographs from SUN RGB-D; the per-row corpus is `source_scene_corpus` in `meta` and `preview`. On photograph rows both the input and the target are the photograph. On 1,621 of the 5,799 rows (28.0%) the parts are already shown in the right order, so the target is the input with the separator strips and the number labels removed, not a rearrangement (592 of 1,123 on 1x2, 200 of 1,193 on 1x3, 580 of 1,114 on 2x1, 50 of 1,190 on 2x2, 199 of 1,179 on 3x1). Changes from upstream: images are re-encoded as JPEG; inputs are downscaled to a long edge of at most 512 px and targets to at most 768 px (smaller images keep their size). The part numbers are drawn upstream at a fixed size, and about half of the ADE20K photographs are 1,025-2,239 px on their long edge upstream, so on those rows the numbers shrink to a few pixels and are often unreadable. Every question states which position carries which number, so the task does not depend on reading them. Refused by the conversion's checks: 118 rows whose read-back names a different option from the label (`S0.readback_names_other_option`); 14 rows whose plan names a different option from the label (`S0.plan_names_other_option`). The plan usually decides the arrangement before the picture is drawn. Where the upstream plan named an option letter, every sentence that named one was moved to the start of the read-back (2,015 rows, flag `S4.plan_verdict_moved`). On most of these rows that sentence is the plan's closing verdict, so the read-back opens with the verdict before it re-examines the picture; on some it is the plan's comparison of the options, and there the plan no longer says which arrangement to draw (see Known issues).
+
+**Supervision kind** (`supervision_kind` in `meta`): `full_interleaved` on every row: the upstream trace itself interleaves text and target images (drawn or rendered states on most sources; the source note above says which), and the read-back comes after the target image it reads. The text is upstream's and was not checked against the images.
+
+Upstream: [`ThinkMorph/Jigsaw_Assembly`](https://huggingface.co/datasets/ThinkMorph/Jigsaw_Assembly). Licence: **undeclared**. The upstream repository declares no licence; this converted copy is shared for research use only, whatever terms the upstream authors set apply to it as well, and it will be taken down at their request. The ADE20K and SUN RGB-D photographs (the rows whose `source_scene_corpus` is `ade20k` or `sunrgbd`, in both `meta` and `preview`) are third-party images: neither ThinkMorph nor we hold their copyright, and they remain under the terms of ADE20K (https://groups.csail.mit.edu/vision/datasets/ADE20K/terms/) and SUN RGB-D, and of the collections those draw on. ADE20K's terms allow use only for non-commercial research and education, and allow the images to be passed on only to people who agree to those terms; download these rows only if you accept them. To leave the photographs out, filter on `source_scene_corpus`. The SAT renders come from SAT (MIT) and ProcTHOR / AI2-THOR assets. The reasoning text was generated with OpenAI's GPT-4.1 (ThinkMorph paper, appendix on data generation); OpenAI's terms for model outputs may bear on some uses, for example training models that compete with OpenAI's.
+
+## Known issues
+
+Rows with a measured per-row problem are listed in `reports/known_issues/`, one TSV per issue (a `# <description>` line, then `row_uid<TAB>split<TAB>detail` lines), so they can be filtered out. They are still in this release: no row was removed for these issues.
+
+| issue | rows | train | validation | what | how it was found | file |
+|---|---:|---:|---:|---|---|---|
+| `readback_opens_with_moved_comparison` | 172 | 164 | 8 | Flag S4.plan&#95;verdict&#95;moved and the read-back opens with sentences moved from the plan that read as a comparison of the options ('If I place Part 1 below Part 2 (option A), ...') rather than a verdict, so the read-back starts with pre-image reasoning; judged on the opening sentence only, so some listed rows' moved text does end in a verdict (mostly three- and four-part puzzles). | row flagged S4.plan&#95;verdict&#95;moved and the read-back's first sentence starts with If/Suppose, or starts with When/Testing/Checking/Placing/Comparing/Considering/Evaluating/Examining the options and does not itself give a verdict (matches, corresponds, the correct, the only, most natural, should be, natural order, ...); measured 2026-09-25; known&#95;issues.py sha1 655e5307; export 4aaee1f rows 83cf9966/ae9b703a | `reports/known_issues/readback_opens_with_moved_comparison.tsv` |
+| `text_states_other_arrangement` | 90 | 88 | 2 | The target image and &lt;answer&gt; show the labelled arrangement, but the text does not: the plan concludes, or the read-back says it assembled or chose, a different arrangement (or names another option's letter). | 3/4-part: a full position-&gt;part statement that is not hypothetical and not an enumerated rejected option, differs from the label (the plan's last such statement, any in the read-back); 2-part: a placing/'Part X should be left of Part Y' statement that puts the other part first; any part: 'matches/corresponds to (X)' with X not the answer, or '(X)' followed by option Y's text; measured 2026-09-25; known&#95;issues.py sha1 655e5307; export 4aaee1f rows 83cf9966/ae9b703a | `reports/known_issues/text_states_other_arrangement.tsv` |
+| `control_char_residue` | 49 | 49 | 0 | Removing upstream's control bytes left printable fragments in the text: a garbled &#92;boxed ('&#91;boxed{B}', 'cboxed{A}', 'oxed{A}', '&#91;0m') or a part number cut to '02 should be to the top of Part 1'; the model is trained to write them. | row flagged S0.control&#95;chars&#95;stripped and a thought contains 'boxed{X}' / 'oxed{X}' not preceded by a backslash, an ANSI remnant '&#91;0m' / '&#91;.}', or a '0&lt;digit&gt; should be' part name; measured 2026-09-25; known&#95;issues.py sha1 655e5307; export 4aaee1f rows 83cf9966/ae9b703a | `reports/known_issues/control_char_residue.tsv` |
+| `target_shared_with_other_row` | 12 | 12 | 0 | The same picture (decoded pixels identical) is the target of another shipped row: one photograph or render under two upstream file names, cut into different puzzles. | md5 of the decoded target pixels occurs on more than one shipped row; measured 2026-09-25; known&#95;issues.py sha1 655e5307; export 4aaee1f rows 83cf9966/ae9b703a | `reports/known_issues/target_shared_with_other_row.tsv` |
+| `validation_target_near_copy_of_train` | 1 | 0 | 1 | This validation row's target is a near-copy of a training row's target (the same scene shot from a slightly shifted viewpoint), so it is not fully held out. | validation target within 10 bits (of 256) of a training target by a 17x16 greyscale difference hash; the next-nearest validation-train pair on r2 is 43 bits apart; measured 2026-09-25; known&#95;issues.py sha1 655e5307; export 4aaee1f rows 83cf9966/ae9b703a | `reports/known_issues/validation_target_near_copy_of_train.tsv` |
+
+To leave the listed rows out (the snippet in the loader section downloads `reports/known_issues/` with the data):
+
+```python
+import glob, os
+root = "<root>/thinkmorph_jigsaw"
+drop = {line.split("\t")[0] for f in glob.glob(os.path.join(root, "reports/known_issues/*.tsv"))
+        for line in open(f) if line.strip() and not line.startswith(("#", "row_uid\t"))}
+# keep a row when its row_uid (a column of train/, meta/ and preview/) is not in drop
+```
+
+### Measured caveats
+
+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.
+
+- **Training-signal defect.** The target image and the answer always show the labelled arrangement (the labelled permutation is the closest re-assembly of the input's parts to the target on all 5,799 rows), but the upstream text does not always agree with them. On 90 rows (88 train, 2 validation; listed as `text_states_other_arrangement`), 1.6% of the rows, the plan argues for, or the read-back says it assembled, a different arrangement, often the identity order that is not among the options, and the row then answers the label. They include read-backs and plans that quote a different arrangement in the options' own wording (e.g. `775b42a13ffe2cf2`, `6ed2d5f38961ddb5`, validation `4869f3448140a813`), read-backs that pair '(A)' with option D's text and answer D (`070c5e56e50991e1`, `8df62bedf96e0c99`), and two-part plans that conclude the opposite arrangement in free wording (e.g. `eb30e46819a3aaae`). These rows train: text says X, image shows Y, answer Y.
+- The plan quotes the chosen option's text on 481 of the 5,799 rows. On rows flagged `S4.plan_verdict_moved` the sentences moved to the read-back are often the plan's comparison of the options ('If I place Part 1 below Part 2 (option A), ...') rather than its verdict, so the read-back opens with that pre-image comparison and the plan may no longer say which arrangement to draw (e.g. `f0c0ba3365bd5be7`). listed as `readback_opens_with_moved_comparison`: 172 rows, 164 train and 8 validation; the list is found from the read-back's opening sentence and also holds some rows whose moved text does end in a verdict, mostly among the three- and four-part puzzles.
+- **Training-signal defect.** Upstream wrote `\boxed{X}` in a form mangled into control bytes; the conversion stripped the control bytes (flag `S0.control_chars_stripped`) but left the printable remnant, so fragments such as `[boxed{B}`, `b[boxed{A}`, `cboxed{B}` and `boxed{A}` remain in the text, nearly all in the read-back (e.g. `96f5cd52dc2240f7`, `d40a6f95cec102f3`; listed as `control_char_residue`: 49 rows, 49 train and 0 validation; `99f7eab94a675c55` instead has a part number cut to '02 should be ...' in its plan), and the model is trained to write them. The letter in each boxed fragment equals the row's answer.
+- One picture can appear under two file names: 12 rows (12 train, 0 validation; listed as `target_shared_with_other_row`) share their target picture with another row (e.g. `ebd5964bccf870f9` and `01ce85a03268e9b2`; `b477f54ba9491c9d` and `48ccb324817e04b8` are the same puzzle with its answer worded differently). Validation row `96fb0688103dd21d` (ADE_train_00018130, listed as `validation_target_near_copy_of_train`) shows nearly the same street photograph as the target of training row `bdcb20276b29b58b` (ADE_train_00018569); it is cut into a different grid, so the check on input images does not catch it.
+
+## Size
+
+| split | rows | target image slots | distinct target images |
+|---|---:|---:|---:|
+| train | 5,654 | 5,654 | 5,648 |
+| validation | 145 | 145 | 145 |
+
+A slot is one target position in one row. Upstream has a few pictures under two file names, cut into different puzzles, so a few rows share a target picture (see Known issues).
+
+| task | train | validation |
+|---|---:|---:|
+| jigsaw_1x2 | 1,086 | 37 |
+| jigsaw_1x3 | 1,163 | 30 |
+| jigsaw_2x1 | 1,098 | 16 |
+| jigsaw_2x2 | 1,159 | 31 |
+| jigsaw_3x1 | 1,148 | 31 |
+
+Input images per row: 1. Target images per row (the images the model is trained to generate): 1.
+Image corpus (`source_scene_corpus`): procthor 3,203, ade20k 1,821, sunrgbd 775.
+
+## Row format
+
+One row is: input image(s) and a question, then K rounds of *thought → target image* (the target is the source's own ground-truth image, which the model is trained to generate), then a final thought (normally a read-back of the last target; where a source's final thought is something else, or often leaves out the answer, the source note or Known issues says so) and the answer; here K is 1. In the `train` config:
+
+```
+image_list        list<binary>  inputs first, then the K target images in order
+num_input_images  int64         how many of image_list are inputs
+instruction_list  list<string>  one element: system prompt + question + options
+output_text_list  list<string>  K+1 elements:
+  [0]   <think>plan 1</think><image_start>
+  [j]   <image_end><think>plan j+1</think><image_start>
+  [K]   <image_end><think>read-back</think><answer>answer</answer>
+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` 5,799) 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.
+
+`preview` shows the same rows with one column per slot: `input_image_i` for the inputs; for each of the K = `num_steps` rounds, the plan `thought_j` and its target `target_image_j`; and the read-back in `thought_1` on every row.
+
+`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`: undeclared 5,799.
+
+Flags on released rows (`filter_flags` in `meta` and `preview`, comma-separated):
+
+| flag | rows | meaning |
+|---|---:|---|
+| `S5.replay_unsupported` | 5,799 | no solver re-derives this task's answer from the trace, so S5 did not replay it |
+| `S4.plan_verdict_moved` | 2,015 | the plan named the option letter; those sentences were moved to the start of the read-back |
+| `S8.phash_near_but_distinct` | 1,373 | a target's perceptual hash is within 6 bits of an input image's, but its pixels differ, so it is not a copy; kept |
+| `S14.sampled_qa` | 200 | chosen for the S14 human spot-check (`reports/s14_sample.tsv`) |
+| `S0.control_chars_stripped` | 52 | control characters removed from the upstream text |
+| `S4c.freeform_unadjudicated` | 13 | S4c's generic check could not match the read-back's wording (e.g. 'statement (D', 'boxed{B') to the long option text; these rows are multiple choice, not free-form, and read by hand all 13 conclude the labelled option; kept |
+
+### Training with a BAGEL-family loader
+
+Every row here has one input image (`num_input_images` is 1), so the stock ThinkMorph `UnifiedEditIterableDataset` (https://github.com/ThinkMorph/ThinkMorph: `image_list[0]` as input, `image_list[j+1]` after `output_text_list[j]`) and the IPT release's version (which reads `num_input_images`) both read it as intended. Mixed with a source whose rows have more than one input image, only a loader that reads `num_input_images` is correct.
+
+The stock BAGEL edit loader (ByteDance-Seed/Bagel) cannot train these rows: it never reads `output_text_list` and expects each `instruction_list` element to be a list of paraphrases.
+
+`parquet_info.json` keys each training chunk as `<source>/<split>/<file>`, here `thinkmorph_jigsaw/train/chunk_00000.parquet`, with row-group counts read from the parquet footers. The loader matches a chunk only when its key equals the path it builds, `os.path.join(data_dir, file)`, and skips a chunk with no key without a warning: a source that is alone in its group then fails with `IndexError: list index out of range`, and in a mixed group it adds no rows. Download into a directory named after the source, not after the repository:
+
+```python
+from huggingface_hub import snapshot_download
+snapshot_download("yrlyrl/spatial-mmcot-thinkmorph_jigsaw", repo_type="dataset", local_dir="<root>/thinkmorph_jigsaw",
+                  allow_patterns=["train/*", "validation/*", "parquet_info.json", "reports/known_issues/*"])
+```
+
+Then either run from `<root>` with `data_dir: thinkmorph_jigsaw/train` and `parquet_info_path: thinkmorph_jigsaw/parquet_info.json`, or rebuild the index with absolute keys and use an absolute `data_dir`:
+
+```python
+import json, os
+root = "/abs/path/to/root"                      # the directory that holds thinkmorph_jigsaw/
+info = json.load(open(os.path.join(root, "thinkmorph_jigsaw", "parquet_info.json")))
+info = {os.path.join(root, k): v for k, v in info.items()}
+json.dump(info, open(os.path.join(root, "thinkmorph_jigsaw", "parquet_info_abs.json"), "w"))
+# data_dir = os.path.join(root, "thinkmorph_jigsaw", "train")   (spelled exactly so, no trailing slash)
+# parquet_info_path = os.path.join(root, "thinkmorph_jigsaw", "parquet_info_abs.json")
+```
+
+The Hugging Face cache (`.../snapshots/<hash>/train/`) or a folder named `spatial-mmcot-thinkmorph_jigsaw` 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 45 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 45 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
+
+| stage | rows |
+|---|---:|
+| upstream rows read | 6,000 |
+| refused before conversion (`S0raw`; each reason is in the table below) | 133 |
+| removed as benchmark evaluation items (S11) | 2 |
+| dropped at S5 by the phrase check ('does not make sense', 'discrepancy'); read by hand, 8 of these plans reject the arrangement they then choose or conclude a different option from the label, the others use the phrase to reject a wrong arrangement | 36 |
+| after conversion and per-row filters | 5,829 |
+| removed by S10 (none) | 0 |
+| removed by answer-prior balancing (S13) | 30 |
+| **released** | **5,799** |
+
+Every removed row has one line, with its reason, in `reports/`:
+
+| file | step | reason (the line's `flag`, or the field shown) | rows |
+|---|---|---|---:|
+| `build/dropped.jsonl` | S0raw | `S0.readback_names_other_option` | 118 |
+| `build/dropped.jsonl` | S0raw | `S0.plan_names_other_option` | 14 |
+| `build/dropped.jsonl` | S0raw | `S0.plan_enumerates_options` | 1 |
+| `build/dropped.jsonl` | S11 | `S11.bench_item` | 2 |
+| `build/dropped.jsonl` | S5 | `S5.self_contradiction` | 36 |
+| `s13_dropped.jsonl` | S13 | `step: letter` | 24 |
+| `s13_dropped.jsonl` | S13 | `step: answer` | 6 |
+
+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`).
+
+`S0raw` lines in `build/dropped.jsonl` were refused before a release row existed, so their `row_uid` field holds the converter's key for the upstream record, not a 16-hex `row_uid`; lines from later steps carry the `row_uid` the row had. No removed row appears in `meta` or `preview`.
+
+`S11.bench_item` marks upstream rows with an image that is an evaluation item of a benchmark (2 rows: puzzles whose target, the assembled photograph, is an image of the What's Up benchmark (Kamath et al., 2023), which is not among the benchmarks we report but is kept out of training; upstream ThinkMorph Jigsaw_Assembly still contains them); they were removed during conversion, so no such item ships.
+
+<details><summary>Per-step counters of the conversion</summary>
+
+6,000 upstream rows were read; `S0raw` refused 133 before a row existed and passed 5,867 to the first step. `S0` runs once more, last, on the final bytes. The reason for every refused, dropped or quarantined row is in the files above.
+
+| step | in | out | dropped | quarantined | rejected | repaired |
+|---|---:|---:|---:|---:|---:|---:|
+| S0raw (refused before conversion) | 6,000 | 5,867 | 0 | 0 | 133 | 0 |
+| S11 | 5,867 | 5,865 | 0 | 0 | 2 | 0 |
+| S4 | 5,865 | 5,865 | 0 | 0 | 0 | 0 |
+| S4c | 5,865 | 5,865 | 0 | 0 | 0 | 0 |
+| S5 | 5,865 | 5,829 | 36 | 0 | 0 | 0 |
+| S8 | 5,829 | 5,829 | 0 | 0 | 0 | 0 |
+| S9 | 5,829 | 5,829 | 0 | 0 | 0 | 0 |
+| S0 (final structural check, after S9) | 5,829 | 5,829 | 0 | 0 | 0 | 53 |
+
+</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). `scene_id` is the image file name. Nothing links SAT renders of one ProcTHOR house, or photographs of one place, so such images can sit on both sides, and one picture can appear under two file names (examples under Known issues). S12 saw 5,829 rows under 5,829 keys, one row per key, so the split is in effect per row. 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).
+
+### Answer-prior balancing (S13)
+
+Each (task, split) group is checked separately. An answer is the answer value compared as lower-cased text without a trailing full stop, with 'farther' read as 'further' and 'nearer' as 'closer' (for multiple choice, the option text, not the letter; where the candidates are drawn in the image, as in zebra_jigsaw and zebra_tetris, the answer is the letter itself). An answer is real when it holds at least 5 rows and 2% of the group; k is the number of real answers. Answer step: the target is max(30%, 1/k) when k >= 2, and max(30%, 1/d) over the d distinct answers when k = 1; a validation group uses the larger of its own target and its task's train target. A group is cut only when k >= 1 and its most common answer holds more than the target plus 5 percentage points; every answer is then capped at one common count, chosen so that none exceeds the target, and smaller answers keep all their rows. At the answer step, a group at or below that trigger, or with no real answer (k = 0), is left as it is, so its most common answer can hold up to the target plus 5 percentage points. A task whose train group has exactly two real answers is instead cut, in every split, so that its two largest answers have equal counts, with no trigger. Rank and label steps: then, in a group where every option value of every row is a number, the rank of the correct option among the sorted values, and after it, in a group where every trained answer is an option label, the label, are each capped by the same cut-and-trigger rule on their own counts (own target, validation included): capped, never evened out, so two labels are cut only when one exceeds 55%, and then only down to 50%. These steps can also cut groups the answer step left whole, including k = 0 groups, and can raise an answer's final share above its target; the run fails if a real answer ends above the target plus 5 percentage points. A train group of at least 20 rows in which one answer holds 90% or more fails the run. PET (exact_cells_pet) instead cuts each (question type x turn direction) cell to equal counts of its two answers; a PET cell that shows only one answer is removed.
+
+| task | split | pass | rule | rows in → out | real answers k | target | largest share, before → after | cut |
+|---|---|---|---|---:|---:|---:|---:|---|
+| jigsaw_1x2 | train | answer | `cap30[canon]` | 1,086 → 1,086 | 4 | 30.0% | 26.6% → 26.6% | no |
+| jigsaw_1x2 | train | letter | `cap30[letter]` | 1,086 → 1,086 | 2 | 50.0% | 51.6% → 51.6% | no |
+| jigsaw_1x2 | validation | answer | `cap30[canon]` | 41 → 37 | 4 | 30.0% | 36.6% → 29.7% | yes |
+| jigsaw_1x2 | validation | letter | `cap30[letter]` | 37 → 37 | 2 | 50.0% | 51.4% → 51.4% | no |
+| jigsaw_1x3 | train | answer | `cap30[canon]` | 1,163 → 1,163 | 6 | 30.0% | 16.9% → 16.9% | no |
+| jigsaw_1x3 | train | letter | `cap30[letter]` | 1,163 → 1,163 | 4 | 30.0% | 25.2% → 25.2% | no |
+| jigsaw_1x3 | validation | answer | `cap30[canon]` | 30 → 30 | 3 | 33.3% | 26.7% → 26.7% | no |
+| jigsaw_1x3 | validation | letter | `cap30[letter]` | 30 → 30 | 4 | 30.0% | 30.0% → 30.0% | no |
+| jigsaw_2x1 | train | answer | `cap30[canon]` | 1,098 → 1,098 | 4 | 30.0% | 26.3% → 26.3% | no |
+| jigsaw_2x1 | train | letter | `cap30[letter]` | 1,098 → 1,098 | 2 | 50.0% | 51.2% → 51.2% | no |
+| jigsaw_2x1 | validation | answer | `cap30[canon]` | 26 → 24 | 3 | 33.3% | 38.5% → 33.3% | yes |
+| jigsaw_2x1 | validation | letter | `cap30[letter]` | 24 → 16 | 2 | 50.0% | 66.7% → 50.0% | yes |
+| jigsaw_2x2 | train | answer | `cap30[canon]` | 1,159 → 1,159 | 24 | 30.0% | 4.3% → 4.3% | no |
+| jigsaw_2x2 | train | letter | `cap30[letter]` | 1,159 → 1,159 | 4 | 30.0% | 25.3% → 25.3% | no |
+| jigsaw_2x2 | validation | answer | `cap30[canon]` | 37 → 37 | 0 | 30.0% | 10.8% → 10.8% | no |
+| jigsaw_2x2 | validation | letter | `cap30[letter]` | 37 → 31 | 4 | 30.0% | 40.5% → 29.0% | yes |
+| jigsaw_3x1 | train | answer | `cap30[canon]` | 1,148 → 1,148 | 6 | 30.0% | 17.0% → 17.0% | no |
+| jigsaw_3x1 | train | letter | `cap30[letter]` | 1,148 → 1,148 | 4 | 30.0% | 25.4% → 25.4% | no |
+| jigsaw_3x1 | validation | answer | `cap30[canon]` | 41 → 41 | 6 | 30.0% | 22.0% → 22.0% | no |
+| jigsaw_3x1 | validation | letter | `cap30[letter]` | 41 → 31 | 4 | 30.0% | 36.6% → 29.0% | yes |
+
+S13 removed 30 rows from this source.
+
+### Text-only baselines
+
+Accuracy of guessers that never see an image. For each task the released training rows are split into two fixed halves by a hash of `row_uid`; each guesser is fitted on one half and scored once on the other (one held-out half, not cross-validation; `eval rows` below). The reference is chance (the mean of 1 / number of options) where every row is multiple choice, and otherwise the eval-half accuracy of always giving the answer most common in the fit half (when a task's top answers are nearly tied, this need not be the task's most common answer; the line after the table gives that answer's validation score). Accuracies are recounted from the stored rates and `eval rows`, so they are exact. A task is flagged when a text-only guesser beats its reference by more than 0.15 (for a free-form task, a guesser other than the most common answer). A flagged task can be partly answered from the text alone; an unflagged task passed only these probes, which do not prove the text carries no answer. Report scores on every task next to this baseline.
+
+Guessers: `keywords`: the most common answer per set of spatial words in the question; `last_mentioned`: the option named last in the question body; `letter_prior`: the most common answer letter; `majority`: the answer most common in the fit half; `option_prior`: the option text that won most often when shown; `template`: the most common answer per question wording (numbers masked, object names kept).
+
+| task | best text-only guesser | accuracy | reference | margin | eval rows | flagged |
+|---|---|---:|---:|---:|---:|---|
+| jigsaw_1x2 | `option_prior` | 0.495 | 0.500 (chance) | -0.005 | 547 | no |
+| jigsaw_1x3 | `letter_prior` | 0.243 | 0.250 (chance) | -0.007 | 575 | no |
+| jigsaw_2x1 | `option_prior` | 0.523 | 0.500 (chance) | +0.023 | 549 | no |
+| jigsaw_2x2 | `letter_prior` | 0.242 | 0.250 (chance) | -0.008 | 590 | no |
+| jigsaw_3x1 | `option_prior` | 0.241 | 0.250 (chance) | -0.009 | 553 | no |
+
+## Spot-check (S14)
+
+**Pending.** The S14 rows are chosen and flagged `S14.sampled_qa` in `meta` and `preview`; the human pass over them has not been signed off yet.
+
+## Citation
+
+Please cite ThinkMorph ([arXiv:2510.27492](https://arxiv.org/abs/2510.27492)), which built the puzzles and their reasoning traces:
+
+```bibtex
+@article{gu2025thinkmorph,
+  title={ThinkMorph: Emergent Properties in Multimodal Interleaved Chain-of-Thought Reasoning},
+  author={Gu, Jiawei and Hao, Yunzhuo and Wang, Huichen Will and Li, Linjie and Shieh, Michael Qizhe and Choi, Yejin and Krishna, Ranjay and Cheng, Yu},
+  journal={arXiv preprint arXiv:2510.27492},
+  year={2025}
+}
+```
+
+and the sources of the pictures:
+
+- SAT renders (`source_scene_corpus` = `procthor`): A. Ray et al., "SAT: Dynamic Spatial Aptitude Training for Multimodal Language Models", arXiv:2412.07755; and M. Deitke et al., "ProcTHOR: Large-Scale Embodied AI Using Procedural Generation", NeurIPS 2022.
+- ADE20K photographs (`ade20k`), as ADE20K asks: B. Zhou, H. Zhao, X. Puig, S. Fidler, A. Barriuso and A. Torralba, "Scene Parsing through ADE20K Dataset", CVPR 2017; and B. Zhou, H. Zhao, X. Puig, T. Xiao, S. Fidler, A. Barriuso and A. Torralba, "Semantic Understanding of Scenes through the ADE20K Dataset", IJCV (https://groups.csail.mit.edu/vision/datasets/ADE20K/).
+- SUN RGB-D photographs (`sunrgbd`): S. Song, S. Lichtenberg and J. Xiao, "SUN RGB-D: A RGB-D Scene Understanding Benchmark Suite", CVPR 2015; and, as the SUN RGB-D page requires of every user (https://rgbd.cs.princeton.edu/), the three datasets it contains: N. Silberman, D. Hoiem, P. Kohli and R. Fergus, "Indoor Segmentation and Support Inference from RGBD Images", ECCV 2012; A. Janoch, S. Karayev, Y. Jia, J. T. Barron, M. Fritz, K. Saenko and T. Darrell, "A Category-Level 3-D Object Dataset: Putting the Kinect to Work", ICCV Workshop on Consumer Depth Cameras for Computer Vision 2011; J. Xiao, A. Owens and A. Torralba, "SUN3D: A Database of Big Spaces Reconstructed using SfM and Object Labels", ICCV 2013.
+
+## Provenance
+
+The release files were written by our conversion code (the code repository is not public yet), `scripts/convert/export.py` at commit `4aaee1f4e946`, from build `thinkmorph_jigsaw_r2`. The build was made by `scripts/convert/run_source.py` from the same repository at commit `4aaee1f4e946`. S10, S12 and S13 ran before the export; `reports/export_manifest.json` pins every input the export read by SHA-1 (`build_manifest_sha1`, `s10_keep_sha1`, `s12_assignments_sha1`, `s13_balanced_keep_sha1`).
+
+Every row removed between upstream and this release has one line, with its reason, in `reports/`: `build/dropped.jsonl` (rows refused before conversion or dropped by a conversion step, S11 benchmark items included); `build/quarantine.jsonl` (rows set aside by S4c because an automatic check could not match the read-back's conclusion to the label); `s10_dropped.jsonl` (duplicates removed by S10); `s10_label_conflicts.jsonl` (rows S10 withheld because another row asks the identical question, options in the same order, of the same images with a different answer); `s13_dropped.jsonl` (rows removed by answer-prior balancing). `known_issues/` lists rows with a measured problem (see Known issues); `reports/` also holds the build manifest (absolute paths cut to basenames) and counters, the S14 sample list (`s14_sample.tsv`: row_uid, task, split) and `export_manifest.json`. In `reports/build/manifest.json`, `spec.source_scene_corpus` (`procthor`) is only the converter's fallback for rows that carry no corpus of their own; it does not describe every row. Each row's corpus is `source_scene_corpus` in `meta` (procthor 3,203, ade20k 1,821, sunrgbd 775). Part of [`yrlyrl/spatial-mmcot`](https://huggingface.co/datasets/yrlyrl/spatial-mmcot).

meta/train/chunk_00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1d0da3c3534059deb2ea1ace17a4a6526d2fdc3447db943902ada889caac510
+size 504996

meta/validation/chunk_00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:decf74c9aaaafb5e601ed78c6663baa8d554b9d1eb522b225c9c8fb7bc12b1aa
+size 19443

parquet_info.json

@@ -0,0 +1,10 @@
+{
+ "thinkmorph_jigsaw/train/chunk_00000.parquet": {
+  "num_row_groups": 45,
+  "num_rows": 5654
+ },
+ "thinkmorph_jigsaw/validation/chunk_00000.parquet": {
+  "num_row_groups": 2,
+  "num_rows": 145
+ }
+}

preview/train/chunk_00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20f62f6614b7504b2705f715ecad1ef985a6636b673204bd84f4bc93eb1fead8
+size 714879019

preview/validation/chunk_00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45b209d4fd9ac601414713425d9e1c3877fb84fea0dc4bf92704936a1733b2a8
+size 18249282

reports/build/counters.json

@@ -0,0 +1,93 @@
+{
+ "slug": "thinkmorph_jigsaw",
+ "targets_out": 5829,
+ "steps": {
+  "S11": {
+   "step": "S11",
+   "in_": 5867,
+   "out": 5865,
+   "dropped": 0,
+   "repaired": 0,
+   "quarantined": 0,
+   "rejected": 2,
+   "seconds": 0.004849672317504883
+  },
+  "S4": {
+   "step": "S4",
+   "in_": 5865,
+   "out": 5865,
+   "dropped": 0,
+   "repaired": 0,
+   "quarantined": 0,
+   "rejected": 0,
+   "seconds": 0.3604001998901367
+  },
+  "S4c": {
+   "step": "S4c",
+   "in_": 5865,
+   "out": 5865,
+   "dropped": 0,
+   "repaired": 0,
+   "quarantined": 0,
+   "rejected": 0,
+   "seconds": 0.21987438201904297
+  },
+  "S5": {
+   "step": "S5",
+   "in_": 5865,
+   "out": 5829,
+   "dropped": 36,
+   "repaired": 0,
+   "quarantined": 0,
+   "rejected": 0,
+   "seconds": 1.3420724868774414
+  },
+  "S8": {
+   "step": "S8",
+   "in_": 5829,
+   "out": 5829,
+   "dropped": 0,
+   "repaired": 0,
+   "quarantined": 0,
+   "rejected": 0,
+   "seconds": 41.04606056213379
+  },
+  "S9": {
+   "step": "S9",
+   "in_": 5829,
+   "out": 5829,
+   "dropped": 0,
+   "repaired": 0,
+   "quarantined": 0,
+   "rejected": 0,
+   "seconds": 0.6227550506591797
+  },
+  "S0": {
+   "step": "S0",
+   "in_": 5829,
+   "out": 5829,
+   "dropped": 0,
+   "repaired": 53,
+   "quarantined": 0,
+   "rejected": 0,
+   "seconds": 1.5328147411346436
+  },
+  "S0raw": {
+   "step": "S0raw",
+   "in_": 133,
+   "out": 0,
+   "dropped": 0,
+   "repaired": 0,
+   "quarantined": 0,
+   "rejected": 133,
+   "seconds": 0.0
+  }
+ },
+ "expectations": {
+  "S4.in_": {
+   "want": 5865,
+   "got": 5865,
+   "ok": true
+  }
+ }
+}

reports/build/dropped.jsonl

@@ -0,0 +1,171 @@
+{"row_uid": "00030", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "00069", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00077", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "00092", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text=None but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00098", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "00118", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "00133", "step": "S0raw", "flag": "S0.plan_names_other_option", "detail": "plan names ['A'] but the label is 'B'", "rejected": true}
+{"row_uid": "00138", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00157", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00158", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00216", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00219", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00225", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "180fbd3634337105", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "00270", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00291", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00304", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "00314", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00318", "step": "S0raw", "flag": "S0.plan_names_other_option", "detail": "plan names ['B'] but the label is 'A'", "rejected": true}
+{"row_uid": "00374", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "00386", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00440", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "00441", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00456", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "477b27dbd7a0f521", "step": "S11", "flag": "S11.bench_item", "detail": "thinkmorph_jigsaw_whatsup.txt"}
+{"row_uid": "00489", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "00505", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00513", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00515", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00516", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00517", "step": "S0raw", "flag": "S0.plan_names_other_option", "detail": "plan names ['A'] but the label is 'B'", "rejected": true}
+{"row_uid": "00567", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00582", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00593", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00598", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00640", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00676", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00686", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00692", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "00693", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00702", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "00724", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text=None but the label is 'A' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "00725", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00730", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "66d1d70b69b36cc4", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "00747", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00755", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00775", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00778", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "c077f4922100d8e8", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "00815", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00826", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00829", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00891", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00901", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter=None text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00939", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "00962", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00963", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "00972", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "00973", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "00977", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text=None but the label is 'A' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "00980", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "2f1bb696f7afaf42", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[1]: discrepancy"}
+{"row_uid": "01013", "step": "S0raw", "flag": "S0.plan_names_other_option", "detail": "plan names ['A'] but the label is 'B'", "rejected": true}
+{"row_uid": "01014", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "01018", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "01038", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text=None but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "01058", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 2 should be to the left of Part 1')", "rejected": true}
+{"row_uid": "01060", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "01113", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "126b0c3d851d7281", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "01148", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "01188", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the right of Part 2')", "rejected": true}
+{"row_uid": "fe8ce5bf9f2c57b0", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "28a0d6efd45b2b31", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "8767d4c8d6a83343", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[1]: discrepancy"}
+{"row_uid": "01739", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='D' but the label is 'B' ('The left part should be Part 1; the midd')", "rejected": true}
+{"row_uid": "01772", "step": "S0raw", "flag": "S0.plan_names_other_option", "detail": "plan names ['B', 'C', 'D'] but the label is 'A'", "rejected": true}
+{"row_uid": "216c230342c0e1f8", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "ba783309f24f26c1", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: doesn't make sense"}
+{"row_uid": "6f360a914b2415e8", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "02413", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02431", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='B' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02451", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02461", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the bottom of Part 2')", "rejected": true}
+{"row_uid": "e78b5326cf48cf36", "step": "S11", "flag": "S11.bench_item", "detail": "thinkmorph_jigsaw_whatsup.txt"}
+{"row_uid": "e3a37f8c2f1bc579", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "02530", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the bottom of Part 2')", "rejected": true}
+{"row_uid": "02542", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the bottom of Part 2')", "rejected": true}
+{"row_uid": "02544", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02600", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02604", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02612", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 1 should be to the bottom of Part 2')", "rejected": true}
+{"row_uid": "02624", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02648", "step": "S0raw", "flag": "S0.plan_names_other_option", "detail": "plan names ['A'] but the label is 'B'", "rejected": true}
+{"row_uid": "703acde9e3e4a0ea", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "02680", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02681", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02698", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02705", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02727", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02732", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='B' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02751", "step": "S0raw", "flag": "S0.plan_names_other_option", "detail": "plan names ['A'] but the label is 'B'", "rejected": true}
+{"row_uid": "02752", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02767", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "e2c27dddf4092fbd", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "02779", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "8d5e7e9d83da5a78", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "02815", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "87c149758bf4195e", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "02876", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02880", "step": "S0raw", "flag": "S0.plan_names_other_option", "detail": "plan names ['A'] but the label is 'B'", "rejected": true}
+{"row_uid": "02882", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the bottom of Part 2')", "rejected": true}
+{"row_uid": "02912", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the bottom of Part 2')", "rejected": true}
+{"row_uid": "02941", "step": "S0raw", "flag": "S0.plan_names_other_option", "detail": "plan names ['A'] but the label is 'B'", "rejected": true}
+{"row_uid": "02949", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "02950", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the bottom of Part 2')", "rejected": true}
+{"row_uid": "03009", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "03015", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "03030", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "03050", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "03062", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "dc0c3448fecde98b", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[1]: does not make sense"}
+{"row_uid": "03080", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "03082", "step": "S0raw", "flag": "S0.plan_names_other_option", "detail": "plan names ['A'] but the label is 'B'", "rejected": true}
+{"row_uid": "016a9ed0cf65871f", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "03086", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "8b9779356f18454c", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "111c7223beab7bfa", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "03109", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "03110", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 1 should be to the bottom of Part 2')", "rejected": true}
+{"row_uid": "03117", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the bottom of Part 2')", "rejected": true}
+{"row_uid": "03118", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "03147", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "a5021eeab22572e6", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "03154", "step": "S0raw", "flag": "S0.plan_names_other_option", "detail": "plan names ['A'] but the label is 'B'", "rejected": true}
+{"row_uid": "aff64bc021bca22a", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "03265", "step": "S0raw", "flag": "S0.plan_names_other_option", "detail": "plan names ['A'] but the label is 'B'", "rejected": true}
+{"row_uid": "03278", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the bottom of Part 2')", "rejected": true}
+{"row_uid": "03284", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "37c560c806b7f7b6", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "03319", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "03334", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "03366", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "03369", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='B' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "03408", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "b6607ab11c87a823", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "03431", "step": "S0raw", "flag": "S0.plan_names_other_option", "detail": "plan names ['A'] but the label is 'B'", "rejected": true}
+{"row_uid": "2a8ae1394e7556eb", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "03461", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the bottom of Part 2')", "rejected": true}
+{"row_uid": "03467", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the bottom of Part 2')", "rejected": true}
+{"row_uid": "03486", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='B' text='B' but the label is 'A' ('Part 1 should be to the bottom of Part 2')", "rejected": true}
+{"row_uid": "b9cf1aae493c7c8f", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "03536", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='A' but the label is 'B' ('Part 2 should be to the top of Part 1')", "rejected": true}
+{"row_uid": "c8c5d7451df16287", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "c14ec0da95e0410b", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "03704", "step": "S0raw", "flag": "S0.plan_enumerates_options", "detail": "plan is a line-start option list; its letters cannot be moved by sentence", "rejected": true}
+{"row_uid": "717b1cd32f623286", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "04357", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='D' text='A' but the label is 'D' ('The top-left part should be Part 1; the ')", "rejected": true}
+{"row_uid": "40f9cd1814af332c", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "ed146329d0dccd96", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: doesn't make sense"}
+{"row_uid": "c7dd60e89c87f342", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: doesn't make sense"}
+{"row_uid": "3b43f5f3b6c8cd23", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: does not make sense"}
+{"row_uid": "05322", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='C' text='B' but the label is 'C' ('The top part should be Part 1; the middl')", "rejected": true}
+{"row_uid": "8c8b6eccc751d438", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[1]: discrepancy"}
+{"row_uid": "05687", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='D' but the label is 'A' ('The top part should be Part 1; the middl')", "rejected": true}
+{"row_uid": "05761", "step": "S0raw", "flag": "S0.plan_names_other_option", "detail": "plan names ['C'] but the label is 'D'", "rejected": true}
+{"row_uid": "7b569e4f3bf02b43", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: doesn't make sense"}
+{"row_uid": "05804", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='C' text=None but the label is 'A' ('The top part should be Part 1; the middl')", "rejected": true}
+{"row_uid": "05916", "step": "S0raw", "flag": "S0.readback_names_other_option", "detail": "read-back concludes letter='A' text='D' but the label is 'A' ('The top part should be Part 3; the middl')", "rejected": true}
+{"row_uid": "4dac73c4c820db09", "step": "S5", "flag": "S5.self_contradiction", "detail": "thought[0]: doesn't make sense"}

reports/build/manifest.json

@@ -0,0 +1,39 @@
+{
+ "slug": "thinkmorph_jigsaw",
+ "raw": "thinkmorph_jigsaw",
+ "out": "thinkmorph_jigsaw_r2",
+ "raw_rows_read": 6000,
+ "raw_unusable": 133,
+ "rows_out": 5829,
+ "targets_out": 5829,
+ "chunks": 1,
+ "row_groups": 46,
+ "seconds": 177.3,
+ "spec": {
+  "slug": "thinkmorph_jigsaw",
+  "num_inputs": 1,
+  "max_k": 1,
+  "source_scene_corpus": "procthor",
+  "min_inputs": 1,
+  "license": "undeclared",
+  "license_commercial": false,
+  "replayable": false,
+  "min_image_edge": 160,
+  "copy_max_mean_diff": 1.0,
+  "copy_max_changed_frac": 0.02,
+  "allow_truncation": false,
+  "exclude_row_uids": "thinkmorph_jigsaw_whatsup.txt",
+  "exclude_row_uids_expected": 2,
+  "exclude_image_hashes": "",
+  "exclude_image_hashes_expected": 0,
+  "crop_black_bars": false,
+  "expected": {
+   "S4.in_": 5865
+  }
+ },
+ "gates_failed": false,
+ "limit": 0,
+ "converter_git_sha": "4aaee1f4e9469da41810aac049fa411c1834b809",
+ "converter_git_dirty": false,
+ "paths_redacted": "absolute build-machine paths cut to basenames for publication; build_manifest_sha1 in export_manifest.json is the original's sha1"
+}

reports/export_manifest.json

@@ -0,0 +1,704 @@
+{
+ "source": "thinkmorph_jigsaw",
+ "build": "thinkmorph_jigsaw_r2",
+ "export_git_sha": "4aaee1f4e9469da41810aac049fa411c1834b809",
+ "export_git_dirty": false,
+ "limit": 0,
+ "converter_git_sha": "4aaee1f4e9469da41810aac049fa411c1834b809",
+ "converter_git_dirty": false,
+ "build_counts": {
+  "raw_rows_read": 6000,
+  "raw_unusable": 133,
+  "rows_out": 5829
+ },
+ "build_manifest_sha1": "3a3d402b3a4eb4ec74c07f26beb0dde6225280d6",
+ "s10_keep_sha1": "9fd27b7ff9e2be8d9e0466795f251875210de152",
+ "s12_assignments_sha1": "d0195205cd5852e08769b99f9c72740f8ce9fcaa",
+ "s13_balanced_keep_sha1": "af73ed4ef438de882feeee30de4a25701f2a3f6a",
+ "license": "undeclared",
+ "license_commercial": false,
+ "upstream_repo": "ThinkMorph/Jigsaw_Assembly",
+ "donor": null,
+ "visual_options": null,
+ "num_inputs": 1,
+ "max_k": 1,
+ "preview_max_k": 1,
+ "rows": {
+  "train": 5654,
+  "validation": 145
+ },
+ "targets": {
+  "train": 5654,
+  "validation": 145
+ },
+ "distinct_targets_total": 5793,
+ "num_input_images": {
+  "train": [
+   1,
+   1
+  ],
+  "validation": [
+   1,
+   1
+  ]
+ },
+ "num_steps": {
+  "train": [
+   1,
+   1
+  ],
+  "validation": [
+   1,
+   1
+  ]
+ },
+ "distinct_targets": {
+  "train": 5648,
+  "validation": 145
+ },
+ "supervision_kind": {
+  "train": {
+   "full_interleaved": 5654
+  },
+  "validation": {
+   "full_interleaved": 145
+  }
+ },
+ "source_scene_corpus": {
+  "train": {
+   "procthor": 3121,
+   "ade20k": 1782,
+   "sunrgbd": 751
+  },
+  "validation": {
+   "procthor": 82,
+   "ade20k": 39,
+   "sunrgbd": 24
+  }
+ },
+ "filter_flags": {
+  "train": {
+   "S5.replay_unsupported": 5654,
+   "S4.plan_verdict_moved": 1966,
+   "S8.phash_near_but_distinct": 1344,
+   "S14.sampled_qa": 196,
+   "S0.control_chars_stripped": 52,
+   "S4c.freeform_unadjudicated": 13
+  },
+  "validation": {
+   "S5.replay_unsupported": 145,
+   "S4.plan_verdict_moved": 49,
+   "S8.phash_near_but_distinct": 29,
+   "S14.sampled_qa": 4
+  }
+ },
+ "answer_type": {
+  "train": {
+   "mcq_letter": 5654
+  },
+  "validation": {
+   "mcq_letter": 145
+  }
+ },
+ "answer_tag": {
+  "train": {
+   "mcq_letter": {
+    "rows": 5654,
+    "equals_answer_value": 0
+   }
+  },
+  "validation": {
+   "mcq_letter": {
+    "rows": 145,
+    "equals_answer_value": 0
+   }
+  }
+ },
+ "options_text_rows": {
+  "train": 5654,
+  "validation": 145
+ },
+ "text_checks": {
+  "train": {
+   "control_char_rows": 0,
+   "non_latin_rows": 0,
+   "empty_readback_rows": 0,
+   "placeholder_gap_rows": 0,
+   "boxed_differs_from_answer_rows": 0
+  },
+  "validation": {
+   "control_char_rows": 0,
+   "non_latin_rows": 0,
+   "empty_readback_rows": 0,
+   "placeholder_gap_rows": 0,
+   "boxed_differs_from_answer_rows": 0
+  }
+ },
+ "text_check_examples": {
+  "train": {},
+  "validation": {}
+ },
+ "row_license": {
+  "train": {
+   "undeclared": 5654
+  },
+  "validation": {
+   "undeclared": 145
+  }
+ },
+ "removed": {
+  "build/dropped.jsonl": {
+   "S0raw": {
+    "S0.readback_names_other_option": 118,
+    "S0.plan_names_other_option": 14,
+    "S0.plan_enumerates_options": 1
+   },
+   "S11": {
+    "S11.bench_item": 2
+   },
+   "S5": {
+    "S5.self_contradiction": 36
+   }
+  },
+  "build/quarantine.jsonl": {},
+  "s10_dropped.jsonl": {},
+  "s10_label_conflicts.jsonl": {},
+  "s13_dropped.jsonl": {
+   "S13": {
+    "S13.letter": 24,
+    "S13.answer": 6
+   }
+  }
+ },
+ "split": {
+  "policy": "scene_id",
+  "rows": 5829,
+  "keys": 5829,
+  "val_rows": 175,
+  "val_keys": 175,
+  "linked_by_input": null,
+  "components": null,
+  "margin": null,
+  "pixel_near_copy_checked": null,
+  "input_leak_hits": 0,
+  "shared_target_images": 0
+ },
+ "row_uid_sha1": {
+  "train": "83cf996628456291fc5edf5dbeb626d25dcd2f0d",
+  "validation": "ae9b703a37be06981c8b4fa8337f04ab6b01f02f"
+ },
+ "tasks": {
+  "jigsaw_1x2": {
+   "train": 1086,
+   "validation": 37
+  },
+  "jigsaw_1x3": {
+   "train": 1163,
+   "validation": 30
+  },
+  "jigsaw_2x1": {
+   "train": 1098,
+   "validation": 16
+  },
+  "jigsaw_2x2": {
+   "train": 1159,
+   "validation": 31
+  },
+  "jigsaw_3x1": {
+   "train": 1148,
+   "validation": 31
+  }
+ },
+ "answer_counts": {
+  "jigsaw_1x2/train": {
+   "Part 2 should be to the right of Part 1": 289,
+   "Part 1 should be to the left of Part 2": 287,
+   "Part 2 should be to the left of Part 1": 270,
+   "Part 1 should be to the right of Part 2": 240
+  },
+  "jigsaw_1x2/validation": {
+   "Part 1 should be to the right of Part 2": 11,
+   "Part 2 should be to the left of Part 1": 10,
+   "Part 2 should be to the right of Part 1": 9,
+   "Part 1 should be to the left of Part 2": 7
+  },
+  "jigsaw_1x3/train": {
+   "The left part should be Part 3; the middle part should be Part 2; and the right part should be Part 1.": 196,
+   "The left part should be Part 2; the middle part should be Part 1; and the right part should be Part 3.": 195,
+   "The left part should be Part 3; the middle part should be Part 1; and the right part should be Part 2.": 194,
+   "The left part should be Part 1; the middle part should be Part 2; and the right part should be Part 3.": 194,
+   "The left part should be Part 1; the middle part should be Part 3; and the right part should be Part 2.": 194,
+   "The left part should be Part 2; the middle part should be Part 3; and the right part should be Part 1.": 190
+  },
+  "jigsaw_1x3/validation": {
+   "The left part should be Part 2; the middle part should be Part 3; and the right part should be Part 1.": 8,
+   "The left part should be Part 1; the middle part should be Part 2; and the right part should be Part 3.": 6,
+   "The left part should be Part 2; the middle part should be Part 1; and the right part should be Part 3.": 5,
+   "The left part should be Part 3; the middle part should be Part 2; and the right part should be Part 1.": 4,
+   "The left part should be Part 3; the middle part should be Part 1; and the right part should be Part 2.": 4,
+   "The left part should be Part 1; the middle part should be Part 3; and the right part should be Part 2.": 3
+  },
+  "jigsaw_2x1/train": {
+   "Part 2 should be to the bottom of Part 1": 289,
+   "Part 1 should be to the top of Part 2": 284,
+   "Part 1 should be to the bottom of Part 2": 271,
+   "Part 2 should be to the top of Part 1": 254
+  },
+  "jigsaw_2x1/validation": {
+   "Part 1 should be to the top of Part 2": 5,
+   "Part 1 should be to the bottom of Part 2": 5,
+   "Part 2 should be to the top of Part 1": 4,
+   "Part 2 should be to the bottom of Part 1": 2
+  },
+  "jigsaw_2x2/train": {
+   "The top-left part should be Part 4; the top-right part should be Part 2; the bottom-left part should be Part 3; and the bottom-right part should be Part 1.": 50,
+   "The top-left part should be Part 3; the top-right part should be Part 4; the bottom-left part should be Part 1; and the bottom-right part should be Part 2.": 50,
+   "The top-left part should be Part 1; the top-right part should be Part 3; the bottom-left part should be Part 2; and the bottom-right part should be Part 4.": 50,
+   "The top-left part should be Part 2; the top-right part should be Part 3; the bottom-left part should be Part 1; and the bottom-right part should be Part 4.": 50,
+   "The top-left part should be Part 2; the top-right part should be Part 3; the bottom-left part should be Part 4; and the bottom-right part should be Part 1.": 50,
+   "The top-left part should be Part 3; the top-right part should be Part 2; the bottom-left part should be Part 1; and the bottom-right part should be Part 4.": 50,
+   "The top-left part should be Part 2; the top-right part should be Part 4; the bottom-left part should be Part 3; and the bottom-right part should be Part 1.": 50,
+   "The top-left part should be Part 4; the top-right part should be Part 3; the bottom-left part should be Part 1; and the bottom-right part should be Part 2.": 49,
+   "The top-left part should be Part 1; the top-right part should be Part 3; the bottom-left part should be Part 4; and the bottom-right part should be Part 2.": 49,
+   "The top-left part should be Part 3; the top-right part should be Part 1; the bottom-left part should be Part 2; and the bottom-right part should be Part 4.": 49,
+   "The top-left part should be Part 3; the top-right part should be Part 1; the bottom-left part should be Part 4; and the bottom-right part should be Part 2.": 49,
+   "The top-left part should be Part 2; the top-right part should be Part 1; the bottom-left part should be Part 3; and the bottom-right part should be Part 4.": 49,
+   "The top-left part should be Part 2; the top-right part should be Part 4; the bottom-left part should be Part 1; and the bottom-right part should be Part 3.": 48,
+   "The top-left part should be Part 1; the top-right part should be Part 4; the bottom-left part should be Part 2; and the bottom-right part should be Part 3.": 48,
+   "The top-left part should be Part 1; the top-right part should be Part 4; the bottom-left part should be Part 3; and the bottom-right part should be Part 2.": 48,
+   "The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 4; and the bottom-right part should be Part 3.": 48,
+   "The top-left part should be Part 4; the top-right part should be Part 2; the bottom-left part should be Part 1; and the bottom-right part should be Part 3.": 48,
+   "The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 3; and the bottom-right part should be Part 4.": 48,
+   "The top-left part should be Part 4; the top-right part should be Part 1; the bottom-left part should be Part 2; and the bottom-right part should be Part 3.": 47,
+   "The top-left part should be Part 4; the top-right part should be Part 1; the bottom-left part should be Part 3; and the bottom-right part should be Part 2.": 46,
+   "The top-left part should be Part 3; the top-right part should be Part 4; the bottom-left part should be Part 2; and the bottom-right part should be Part 1.": 46,
+   "The top-left part should be Part 4; the top-right part should be Part 3; the bottom-left part should be Part 2; and the bottom-right part should be Part 1.": 46,
+   "The top-left part should be Part 3; the top-right part should be Part 2; the bottom-left part should be Part 4; and the bottom-right part should be Part 1.": 46,
+   "The top-left part should be Part 2; the top-right part should be Part 1; the bottom-left part should be Part 4; and the bottom-right part should be Part 3.": 45
+  },
+  "jigsaw_2x2/validation": {
+   "The top-left part should be Part 4; the top-right part should be Part 3; the bottom-left part should be Part 2; and the bottom-right part should be Part 1.": 4,
+   "The top-left part should be Part 3; the top-right part should be Part 2; the bottom-left part should be Part 4; and the bottom-right part should be Part 1.": 4,
+   "The top-left part should be Part 4; the top-right part should be Part 1; the bottom-left part should be Part 2; and the bottom-right part should be Part 3.": 3,
+   "The top-left part should be Part 4; the top-right part should be Part 1; the bottom-left part should be Part 3; and the bottom-right part should be Part 2.": 3,
+   "The top-left part should be Part 2; the top-right part should be Part 1; the bottom-left part should be Part 4; and the bottom-right part should be Part 3.": 3,
+   "The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 4; and the bottom-right part should be Part 3.": 2,
+   "The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 3; and the bottom-right part should be Part 4.": 2,
+   "The top-left part should be Part 3; the top-right part should be Part 4; the bottom-left part should be Part 2; and the bottom-right part should be Part 1.": 2,
+   "The top-left part should be Part 2; the top-right part should be Part 4; the bottom-left part should be Part 1; and the bottom-right part should be Part 3.": 1,
+   "The top-left part should be Part 4; the top-right part should be Part 3; the bottom-left part should be Part 1; and the bottom-right part should be Part 2.": 1,
+   "The top-left part should be Part 4; the top-right part should be Part 2; the bottom-left part should be Part 1; and the bottom-right part should be Part 3.": 1,
+   "The top-left part should be Part 1; the top-right part should be Part 4; the bottom-left part should be Part 3; and the bottom-right part should be Part 2.": 1,
+   "The top-left part should be Part 3; the top-right part should be Part 1; the bottom-left part should be Part 4; and the bottom-right part should be Part 2.": 1,
+   "The top-left part should be Part 1; the top-right part should be Part 3; the bottom-left part should be Part 4; and the bottom-right part should be Part 2.": 1,
+   "The top-left part should be Part 1; the top-right part should be Part 4; the bottom-left part should be Part 2; and the bottom-right part should be Part 3.": 1,
+   "The top-left part should be Part 2; the top-right part should be Part 1; the bottom-left part should be Part 3; and the bottom-right part should be Part 4.": 1
+  },
+  "jigsaw_3x1/train": {
+   "The top part should be Part 3; the middle part should be Part 2; and the bottom part should be Part 1.": 195,
+   "The top part should be Part 1; the middle part should be Part 2; and the bottom part should be Part 3.": 194,
+   "The top part should be Part 2; the middle part should be Part 3; and the bottom part should be Part 1.": 191,
+   "The top part should be Part 3; the middle part should be Part 1; and the bottom part should be Part 2.": 191,
+   "The top part should be Part 2; the middle part should be Part 1; and the bottom part should be Part 3.": 189,
+   "The top part should be Part 1; the middle part should be Part 3; and the bottom part should be Part 2.": 188
+  },
+  "jigsaw_3x1/validation": {
+   "The top part should be Part 1; the middle part should be Part 3; and the bottom part should be Part 2.": 7,
+   "The top part should be Part 3; the middle part should be Part 1; and the bottom part should be Part 2.": 5,
+   "The top part should be Part 2; the middle part should be Part 3; and the bottom part should be Part 1.": 5,
+   "The top part should be Part 1; the middle part should be Part 2; and the bottom part should be Part 3.": 5,
+   "The top part should be Part 2; the middle part should be Part 1; and the bottom part should be Part 3.": 5,
+   "The top part should be Part 3; the middle part should be Part 2; and the bottom part should be Part 1.": 4
+  }
+ },
+ "majority_class_rate_train": {
+  "jigsaw_1x2": 0.26611418047882135,
+  "jigsaw_1x3": 0.16852966466036112,
+  "jigsaw_2x1": 0.26320582877959925,
+  "jigsaw_2x2": 0.04314063848144953,
+  "jigsaw_3x1": 0.16986062717770034
+ },
+ "s14_sample": {
+  "rows": 200,
+  "target": 200,
+  "replay_unsupported": true
+ },
+ "s13_groups": {
+  "thinkmorph_jigsaw_r2/jigsaw_1x2/train": {
+   "rule": "cap30[canon]",
+   "rows_in": 1086,
+   "rows_out": 1086,
+   "triggered": false,
+   "exceeds_target": false,
+   "level": 289,
+   "target": 0.3,
+   "k_real": 4,
+   "distinct": 4,
+   "majority_share_in": 0.2661,
+   "majority_share_out": 0.2661,
+   "letter": {
+    "rule": "cap30[letter]",
+    "rows_in": 1086,
+    "rows_out": 1086,
+    "triggered": false,
+    "exceeds_target": false,
+    "level": 560,
+    "target": 0.5,
+    "k_real": 2,
+    "distinct": 2,
+    "majority_share_in": 0.5157,
+    "majority_share_out": 0.5157
+   },
+   "majority_share_final": 0.2661
+  },
+  "thinkmorph_jigsaw_r2/jigsaw_1x2/val": {
+   "rule": "cap30[canon]",
+   "rows_in": 41,
+   "rows_out": 37,
+   "triggered": true,
+   "exceeds_target": false,
+   "level": 11,
+   "target": 0.3,
+   "k_real": 4,
+   "distinct": 4,
+   "majority_share_in": 0.3659,
+   "majority_share_out": 0.2973,
+   "letter": {
+    "rule": "cap30[letter]",
+    "rows_in": 37,
+    "rows_out": 37,
+    "triggered": false,
+    "exceeds_target": false,
+    "level": 19,
+    "target": 0.5,
+    "k_real": 2,
+    "distinct": 2,
+    "majority_share_in": 0.5135,
+    "majority_share_out": 0.5135
+   },
+   "majority_share_final": 0.2973
+  },
+  "thinkmorph_jigsaw_r2/jigsaw_1x3/train": {
+   "rule": "cap30[canon]",
+   "rows_in": 1163,
+   "rows_out": 1163,
+   "triggered": false,
+   "exceeds_target": false,
+   "level": 196,
+   "target": 0.3,
+   "k_real": 6,
+   "distinct": 6,
+   "majority_share_in": 0.1685,
+   "majority_share_out": 0.1685,
+   "letter": {
+    "rule": "cap30[letter]",
+    "rows_in": 1163,
+    "rows_out": 1163,
+    "triggered": false,
+    "exceeds_target": false,
+    "level": 293,
+    "target": 0.3,
+    "k_real": 4,
+    "distinct": 4,
+    "majority_share_in": 0.2519,
+    "majority_share_out": 0.2519
+   },
+   "majority_share_final": 0.1685
+  },
+  "thinkmorph_jigsaw_r2/jigsaw_1x3/val": {
+   "rule": "cap30[canon]",
+   "rows_in": 30,
+   "rows_out": 30,
+   "triggered": false,
+   "exceeds_target": false,
+   "level": 8,
+   "target": 0.3333333333333333,
+   "k_real": 3,
+   "distinct": 6,
+   "majority_share_in": 0.2667,
+   "majority_share_out": 0.2667,
+   "letter": {
+    "rule": "cap30[letter]",
+    "rows_in": 30,
+    "rows_out": 30,
+    "triggered": false,
+    "exceeds_target": false,
+    "level": 9,
+    "target": 0.3,
+    "k_real": 4,
+    "distinct": 4,
+    "majority_share_in": 0.3,
+    "majority_share_out": 0.3
+   },
+   "majority_share_final": 0.2667
+  },
+  "thinkmorph_jigsaw_r2/jigsaw_2x1/train": {
+   "rule": "cap30[canon]",
+   "rows_in": 1098,
+   "rows_out": 1098,
+   "triggered": false,
+   "exceeds_target": false,
+   "level": 289,
+   "target": 0.3,
+   "k_real": 4,
+   "distinct": 4,
+   "majority_share_in": 0.2632,
+   "majority_share_out": 0.2632,
+   "letter": {
+    "rule": "cap30[letter]",
+    "rows_in": 1098,
+    "rows_out": 1098,
+    "triggered": false,
+    "exceeds_target": false,
+    "level": 562,
+    "target": 0.5,
+    "k_real": 2,
+    "distinct": 2,
+    "majority_share_in": 0.5118,
+    "majority_share_out": 0.5118
+   },
+   "majority_share_final": 0.2632
+  },
+  "thinkmorph_jigsaw_r2/jigsaw_2x1/val": {
+   "rule": "cap30[canon]",
+   "rows_in": 26,
+   "rows_out": 24,
+   "triggered": true,
+   "exceeds_target": false,
+   "level": 8,
+   "target": 0.3333333333333333,
+   "k_real": 3,
+   "distinct": 4,
+   "majority_share_in": 0.3846,
+   "majority_share_out": 0.3333,
+   "letter": {
+    "rule": "cap30[letter]",
+    "rows_in": 24,
+    "rows_out": 16,
+    "triggered": true,
+    "exceeds_target": false,
+    "level": 8,
+    "target": 0.5,
+    "k_real": 2,
+    "distinct": 2,
+    "majority_share_in": 0.6667,
+    "majority_share_out": 0.5
+   },
+   "majority_share_final": 0.3125
+  },
+  "thinkmorph_jigsaw_r2/jigsaw_2x2/train": {
+   "rule": "cap30[canon]",
+   "rows_in": 1159,
+   "rows_out": 1159,
+   "triggered": false,
+   "exceeds_target": false,
+   "level": 50,
+   "target": 0.3,
+   "k_real": 24,
+   "distinct": 24,
+   "majority_share_in": 0.0431,
+   "majority_share_out": 0.0431,
+   "letter": {
+    "rule": "cap30[letter]",
+    "rows_in": 1159,
+    "rows_out": 1159,
+    "triggered": false,
+    "exceeds_target": false,
+    "level": 293,
+    "target": 0.3,
+    "k_real": 4,
+    "distinct": 4,
+    "majority_share_in": 0.2528,
+    "majority_share_out": 0.2528
+   },
+   "majority_share_final": 0.0431
+  },
+  "thinkmorph_jigsaw_r2/jigsaw_2x2/val": {
+   "rule": "cap30[canon]",
+   "rows_in": 37,
+   "rows_out": 37,
+   "triggered": false,
+   "exceeds_target": false,
+   "level": 4,
+   "target": 0.3,
+   "k_real": 0,
+   "distinct": 17,
+   "majority_share_in": 0.1081,
+   "majority_share_out": 0.1081,
+   "letter": {
+    "rule": "cap30[letter]",
+    "rows_in": 37,
+    "rows_out": 31,
+    "triggered": true,
+    "exceeds_target": false,
+    "level": 9,
+    "target": 0.3,
+    "k_real": 4,
+    "distinct": 4,
+    "majority_share_in": 0.4054,
+    "majority_share_out": 0.2903
+   },
+   "majority_share_final": 0.129
+  },
+  "thinkmorph_jigsaw_r2/jigsaw_3x1/train": {
+   "rule": "cap30[canon]",
+   "rows_in": 1148,
+   "rows_out": 1148,
+   "triggered": false,
+   "exceeds_target": false,
+   "level": 195,
+   "target": 0.3,
+   "k_real": 6,
+   "distinct": 6,
+   "majority_share_in": 0.1699,
+   "majority_share_out": 0.1699,
+   "letter": {
+    "rule": "cap30[letter]",
+    "rows_in": 1148,
+    "rows_out": 1148,
+    "triggered": false,
+    "exceeds_target": false,
+    "level": 292,
+    "target": 0.3,
+    "k_real": 4,
+    "distinct": 4,
+    "majority_share_in": 0.2544,
+    "majority_share_out": 0.2544
+   },
+   "majority_share_final": 0.1699
+  },
+  "thinkmorph_jigsaw_r2/jigsaw_3x1/val": {
+   "rule": "cap30[canon]",
+   "rows_in": 41,
+   "rows_out": 41,
+   "triggered": false,
+   "exceeds_target": false,
+   "level": 9,
+   "target": 0.3,
+   "k_real": 6,
+   "distinct": 6,
+   "majority_share_in": 0.2195,
+   "majority_share_out": 0.2195,
+   "letter": {
+    "rule": "cap30[letter]",
+    "rows_in": 41,
+    "rows_out": 31,
+    "triggered": true,
+    "exceeds_target": false,
+    "level": 9,
+    "target": 0.3,
+    "k_real": 4,
+    "distinct": 4,
+    "majority_share_in": 0.3659,
+    "majority_share_out": 0.2903
+   },
+   "majority_share_final": 0.2258
+  }
+ },
+ "blind_probe": {
+  "thinkmorph_jigsaw_r2/jigsaw_1x2": {
+   "probed": true,
+   "majority": 0.2395,
+   "keywords": 0.2395,
+   "template": 0.2395,
+   "option_prior": 0.4954,
+   "letter_prior": 0.4625,
+   "last_mentioned": 0.4625,
+   "chance": 0.5,
+   "reference": "chance",
+   "best": "option_prior",
+   "flag": false,
+   "eval_rows": 547,
+   "sigma3": 0.0641,
+   "details": {
+    "last_mentioned_rule": "plain; no option mentioned on 547 of 547 eval rows -> letter prior A",
+    "last_mentioned_no_mention_eval_rows": 547
+   }
+  },
+  "thinkmorph_jigsaw_r2/jigsaw_1x3": {
+   "probed": true,
+   "majority": 0.1461,
+   "keywords": 0.1461,
+   "template": 0.1461,
+   "option_prior": 0.24,
+   "letter_prior": 0.2435,
+   "last_mentioned": 0.2435,
+   "chance": 0.25,
+   "reference": "chance",
+   "best": "letter_prior",
+   "flag": false,
+   "eval_rows": 575,
+   "sigma3": 0.0626,
+   "details": {
+    "last_mentioned_rule": "plain; no option mentioned on 575 of 575 eval rows -> letter prior B",
+    "last_mentioned_no_mention_eval_rows": 575
+   }
+  },
+  "thinkmorph_jigsaw_r2/jigsaw_2x1": {
+   "probed": true,
+   "majority": 0.2404,
+   "keywords": 0.2404,
+   "template": 0.2404,
+   "option_prior": 0.5228,
+   "letter_prior": 0.5173,
+   "last_mentioned": 0.5173,
+   "chance": 0.5,
+   "reference": "chance",
+   "best": "option_prior",
+   "flag": false,
+   "eval_rows": 549,
+   "sigma3": 0.064,
+   "details": {
+    "last_mentioned_rule": "plain; no option mentioned on 549 of 549 eval rows -> letter prior A",
+    "last_mentioned_no_mention_eval_rows": 549
+   }
+  },
+  "thinkmorph_jigsaw_r2/jigsaw_2x2": {
+   "probed": true,
+   "majority": 0.0288,
+   "keywords": 0.0288,
+   "template": 0.0288,
+   "option_prior": 0.1932,
+   "letter_prior": 0.2424,
+   "last_mentioned": 0.2424,
+   "chance": 0.25,
+   "reference": "chance",
+   "best": "letter_prior",
+   "flag": false,
+   "eval_rows": 590,
+   "sigma3": 0.0618,
+   "details": {
+    "last_mentioned_rule": "plain; no option mentioned on 590 of 590 eval rows -> letter prior C",
+    "last_mentioned_no_mention_eval_rows": 590
+   }
+  },
+  "thinkmorph_jigsaw_r2/jigsaw_3x1": {
+   "probed": true,
+   "majority": 0.1573,
+   "keywords": 0.1573,
+   "template": 0.1573,
+   "option_prior": 0.2405,
+   "letter_prior": 0.2242,
+   "last_mentioned": 0.2242,
+   "chance": 0.25,
+   "reference": "chance",
+   "best": "option_prior",
+   "flag": false,
+   "eval_rows": 553,
+   "sigma3": 0.0638,
+   "details": {
+    "last_mentioned_rule": "plain; no option mentioned on 553 of 553 eval rows -> letter prior D",
+    "last_mentioned_no_mention_eval_rows": 553
+   }
+  }
+ },
+ "s13_rule": "Each (task, split) group is checked separately. An answer is the answer value compared as lower-cased text without a trailing full stop, with 'farther' read as 'further' and 'nearer' as 'closer' (for multiple choice, the option text, not the letter; where the candidates are drawn in the image, as in zebra_jigsaw and zebra_tetris, the answer is the letter itself). An answer is real when it holds at least 5 rows and 2% of the group; k is the number of real answers. Answer step: the target is max(30%, 1/k) when k >= 2, and max(30%, 1/d) over the d distinct answers when k = 1; a validation group uses the larger of its own target and its task's train target. A group is cut only when k >= 1 and its most common answer holds more than the target plus 5 percentage points; every answer is then capped at one common count, chosen so that none exceeds the target, and smaller answers keep all their rows. At the answer step, a group at or below that trigger, or with no real answer (k = 0), is left as it is, so its most common answer can hold up to the target plus 5 percentage points. A task whose train group has exactly two real answers is instead cut, in every split, so that its two largest answers have equal counts, with no trigger. Rank and label steps: then, in a group where every option value of every row is a number, the rank of the correct option among the sorted values, and after it, in a group where every trained answer is an option label, the label, are each capped by the same cut-and-trigger rule on their own counts (own target, validation included): capped, never evened out, so two labels are cut only when one exceeds 55%, and then only down to 50%. These steps can also cut groups the answer step left whole, including k = 0 groups, and can raise an answer's final share above its target; the run fails if a real answer ends above the target plus 5 percentage points. A train group of at least 20 rows in which one answer holds 90% or more fails the run. PET (exact_cells_pet) instead cuts each (question type x turn direction) cell to equal counts of its two answers; a PET cell that shows only one answer is removed.",
+ "s13_constants": {
+  "TARGET": 0.3,
+  "TRIGGER_GAP": 0.05,
+  "MIN_CLASS_ROWS": 5,
+  "MIN_CLASS_SHARE": 0.02,
+  "PURE_PRIOR": 0.9,
+  "BLIND_MARGIN": 0.15
+ }
+}

reports/known_issues/control_char_residue.tsv

@@ -0,0 +1,50 @@
+# Removing upstream's control bytes left printable fragments in the text: a garbled \boxed ('[boxed{B}', 'cboxed{A}', 'oxed{A}', '[0m') or a part number cut to '02 should be to the top of Part 1'; the model is trained to write them.
+96f5cd52dc2240f7	train	thought_1: arrangement is Part 1 on the left and Part 2 on the right: b[boxed{A}. Now that I have assembled the two part
+d40a6f95cec102f3	train	thought_1: My conclusion is boxed{A}. After assembling the image with Part 2
+f348a99abc2b3a90	train	thought_1: e, the correct arrangement is Part 1 to the left of Part 2: [boxed{A}]. Now that I have assembled the image w
+dc0a8d632f594880	train	thought_1: Part 1 should be to the right of Part 2, or in other words, [boxed{B}. Now that I have assembled the two part
+67f176f32261d05b	train	thought_1: Part 2 (toilet side) on the right, resulting in the answer [boxed{B}. After assembling the two vertical piec
+338864b14f378624	train	thought_1: The correct arrangement is boxed{B}. Now that I have assembled the image wi
+ce3ee44777624d43	train	thought_1: correct assembly is to place Part 1 to the right of Part 2: cboxed{B}. To ensure the coherence of the assembl
+2fd2c7adc0179ca6	train	thought_1: Therefore, the answer is [boxed{B}. To verify the correctness of my assemb
+fb2e51503d59a0c5	train	thought_1: easoning leads me to conclude that the correct assembly is: boxed{A}. To verify the correctness of the assem
+d397f3b5a12dd4fb	train	thought_1: correct assembly is to place Part 1 to the left of Part 2: boxed{B}. To verify the assembly, I examine the
+8fcc7bb4211790a8	train	thought_1: rrect arrangement is to place Part 1 to the left of Part 2: boxed{B}. Now that I have assembled the two part
+0f75e1adecdadd6a	train	thought_1: e Part 1 to the left of Part 2, which corresponds to answer boxed{B}. Now that I have assembled the image wi
+beec882b1a3b597a	train	thought_1: the correct arrangement to reconstruct the natural image is [boxed{B}: The left part should be Part 1; the mi
+66d2f241cf3edd74	train	thought_1: My conclusion is thus [boxed{C}. Now that I have assembled the image by
+defdcbbb6962b77c	train	thought_1: This matches [boxed{B}[0m: The left part should be Part 2; the
+cab8ed81df3bfbc5	train	thought_1: e sky/cloud patterns, it is clear that the correct order is boxed{D}. To ensure my selected arrangement (lef
+a32b4d09103133c3	train	thought_1: the correct arrangement to reconstruct the natural image is oxed{A}. To ensure the correctness of my previo
+48c8e40a79601d85	train	thought_1: In terms of the answer choices, this is boxed{A}. Now, having assembled the image accord
+01fa3ed3f4dbc062	train	thought_1: , the correct assembly is to place Part 1 on top of Part 2: boxed{B}. Now that I have assembled the two piec
+99f7eab94a675c55	train	thought_0: unters, and walls. In conclusion, the correct assembly is: 02 should be to the top of Part 1.
+2a8ca11f5a59c4da	train	thought_1: ive, and natural boundaries, the correct reconstruction is: boxed{B}. Now that I have assembled the puzzle w
+402a0f6d46dfae39	train	thought_1: Thus, the correct arrangement is cboxed{B}. Now that I have assembled the image by
+3485dc5d62f57819	train	thought_1: Thus, my answer is boxed{A}. Now that I have assembled the two puzz
+2c8d284cb8d74e0a	train	thought_1: Therefore, the correct assembly is [boxed{A}, where part 1 is placed on top of part
+00cb3d821f0a14f2	train	thought_1: bly is to place Part 1 at the bottom and Part 2 at the top: [boxed{B}. Now that I have assembled the image wi
+24cfc5a2d696e127	train	thought_1: Therefore, the correct assembly is: [boxed{A}, meaning Part 2 should be to the top of
+031d575186da3156	train	thought_1: n top and Part 2 on the bottom, which corresponds to answer [boxed{B}. Now that I have assembled the image by
+1238fce098520193	train	thought_1: ave part 2 at the bottom of part 1, corresponding to answer [boxed{B}. Now that I have assembled the image by
+a53f58ac72c24b8e	train	thought_1: Therefore, the correct arrangement is: cboxed{A}. Now that I have assembled the image by
+913b26edfd45c269	train	thought_1: to the bottom of Part 2. Therefore, the correct assembly is boxed{B}. Now that I have assembled the image wi
+d5d10b2a463cdc2b	train	thought_1: ce part 1 at the top and part 2 at the bottom, as in answer [boxed{A}]. Now that the two parts are assembled
+90ca0334a76f4aa7	train	thought_1: 1 at the bottom and part 2 at the top, resulting in answer [boxed{B}. Examining the assembled image after ar
+7df0e38947a34d60	train	thought_1: rect assembly is: Part 1 should be to the top of Part 2, or oxed{A}. Now that I have assembled the two part
+ed6c306d3540f9a1	train	thought_1: reconstruction is to place part 2 at the bottom of part 1: [boxed{B}[boxed{B}] Now that I have assembled the
+1b7c895e65192af0	train	thought_1: re items, I confidently conclude the correct arrangement is boxed{B}. Now that I have assembled the image ac
+6fd3bfd0aad319bb	train	thought_1: traits leads to the conclusion that the correct assembly is [boxed]{D}. Now that I have visualized the assembl
+012655f079392bb1	train	thought_1: features (furniture, paintings), the correct arrangement is boxed{D}. To confirm the correctness of my previ
+cb8955eecbeaf63a	train	thought_1: leaf edges, and flower placements, the correct assembly is [boxed{B}]. Upon reviewing the assembled image ba
+c986557791373a4e	train	thought_1: the sofa, lamp, flowers, and table, the correct assembly is [boxed{B}[.} Upon assembling the puzzle pieces ac
+9467780fbcea4956	train	thought_1: (countertop and trashcan) is bottom-right. So the Answer is boxed{C}. Now, I examine the assembled image to
+9b90988a280a4c71	train	thought_1: ighting, and wall features, the correct assembly is clearly boxed{A}. To verify the accuracy and coherence o
+a16e0c89d90fe85b	train	thought_1: Therefore, the correct statement is boxed{D}. Now that I have assembled the pieces a
+1cff3890b7651cd5	train	thought_1: t part should be Part 4. Therefore, the correct assembly is [boxed{C}. Now that I have reconstructed the imag
+59583f996cace214	train	thought_1: ty, spatial clues, and perspective, the correct assembly is [boxed{C}. Upon assembling the puzzle according t
+66cddf0cc2ce2f3e	train	thought_1: art should be Part 2. Therefore, the correct arrangement is boxed{C}. Upon assembling the four puzzle pieces
+bc3374d8410264d1	train	thought_1: Thus, the correct answer is [boxed{C}. Now that I have assembled the pieces i
+8a053a0bf79c95a9	train	thought_1: My final answer is: boxed{D}. To confirm the correctness of the reco
+35c17fecbbb85836	train	thought_1: eatures, and natural boundaries, the correct arrangement is oxed{B}. Now that I have assembled the image us
+0b85836a224bf615	train	thought_1: ndow, refrigerator, and countertop, the correct assembly is [boxed{C}. I have now assembled the image accordi

reports/known_issues/readback_opens_with_moved_comparison.tsv

@@ -0,0 +1,173 @@
+# Flag S4.plan_verdict_moved and the read-back opens with sentences moved from the plan that read as a comparison of the options ('If I place Part 1 below Part 2 (option A), ...') rather than a verdict, so the read-back starts with pre-image reasoning; judged on the opening sentence only, so some listed rows' moved text does end in a verdict (mostly three- and four-part puzzles).
+63f4a5fc8b841aaa	train	jigsaw_1x2, label A: If we try placing Part 2 to the left of Part 1 (arrangement A), the edge of the Fred building in Part 2 fits perfectly with the continuation of the brick building and its decorative features in Part 1.
+bce0594f50d4709f	train	jigsaw_1x2, label B: If I place Piece 2 on the left and Piece 1 on the right (arrangement B), the countertop's edge and pattern align perfectly, creating an uninterrupted surface.
+eec68d6ab5f77da8	train	jigsaw_1x2, label B: If I were to place Part 1 to the right of Part 2 (option A), the tiled area and garden plot would not align properly, and the background buildings would appear disjointed.
+a565085666d929c8	train	jigsaw_1x2, label B: If Part 1 were placed to the right of Part 2 (option A), the window would appear to abruptly meet the counter and the perspective would be visually inconsistent, breaking the natural flow of the room.
+dcc81c7d333a0bb4	train	jigsaw_1x2, label A: If I place piece 2 to the left of piece 1 (i.e., arrangement A), the sink and plunger occupy the left side of the wall, while the window, door, and their respective shadows align naturally on the right.
+6191ce9ec20c2888	train	jigsaw_1x2, label B: If I were to place piece 2 to the left of piece 1 (option A), the staircase structure would be interrupted and the visual lines would not connect naturally, breaking the flow of the stairs and railings.
+b80f858b3864b6ae	train	jigsaw_1x2, label B: When I imagine placing Piece 1 on the left and Piece 2 on the right (Option A), the red train is on the left and the blue train is on the right.
+4a929aae9da4e93f	train	jigsaw_1x2, label A: If I place Part 1 to the left of Part 2 (option B), the trash bin and wall corner would be in the middle of the scene, but the perspective lines and the baseboard would not align smoothly—there would be a jarring disc...
+fd43594a5930a4b8	train	jigsaw_1x2, label B: If I place piece 2 to the right of piece 1 (option A), the dresser and window would be on opposite sides of the room, but this would create an odd composition: the window would be far from the dresser, and the shadows...
+fefa247a18ae8e41	train	jigsaw_1x2, label B: If Piece 1 were placed to the left of Piece 2 (A), the left edge of Piece 1 would show empty wall and floor, and the bed would abruptly start, with no natural continuation from the left.
+9f3066a8f8b7962c	train	jigsaw_1x2, label B: If part 1 is placed on the left and part 2 on the right (option A), the table and chairs would appear to be floating in the middle of the room, and the doorway’s alignment would be off.
+4d456a8460b8b93f	train	jigsaw_1x2, label A: If I place Piece 1 on the left and Piece 2 on the right (Choice A), the table top forms an uninterrupted geometric shape, the chairs line up smoothly, and the background architecture makes sense: the open doorway is o...
+25d0e9c61bd971ab	train	jigsaw_1x2, label B: If I were to place Piece 2 to the left of Piece 1 (Choice A), the backgrounds would not align: the red wall would be on the right, and the bookshelf would be on the left, which is unnatural.
+fb48129709afe86f	train	jigsaw_1x2, label B: If I place Piece 2 on the left and Piece 1 on the right (option A), the window would be in the middle of the room, overlapping the dresser, which would break the natural flow of the room and disrupt the alignment of t...
+95f99f3178a97ef6	train	jigsaw_1x2, label B: If I imagine placing Piece 1 to the left of Piece 2 (option A), the left edge of Piece 2, which contains the white table, would be adjacent to the right edge of Piece 1, which contains the refrigerator and trash bin.
+9837c3a258eac6a3	train	jigsaw_1x2, label B: If Part 2 were to be placed to the left of Part 1 (Option A), the doorway would be adjacent to the table, which looks visually awkward and disrupts the flow of space.
+42322d50eddc17d9	train	jigsaw_1x2, label B: If I place Part 1 to the left of Part 2 (choice A), the door and hallway would be at the far left, with the room's furniture to the right.
+8939ff476c57cc4a	train	jigsaw_1x2, label B: If I place Piece 1 to the left of Piece 2 (Option A), the desk and window would be disconnected from the natural progression of the room—the cabinet would appear isolated on the left, and the window would not align wi...
+782262492ecaf0e8	train	jigsaw_1x2, label B: If I place piece 1 to the left of piece 2 (option A), the toilet and paper holder would awkwardly abut the curved edge of the sink, disrupting the continuity of the tiles and objects.
+98fafabbb8f4b04f	train	jigsaw_1x2, label B: If I place piece 1 to the left of piece 2 (option A), the doorway would appear on the left wall, and the television and nightstand would end up awkwardly separated by a corner, breaking the visual continuity of the room.
+e84e44999e0bac69	train	jigsaw_1x2, label B: If Part 1 is placed to the left of Part 2 (option A), the table legs and surface would not connect smoothly—the leftmost table leg in Part 1 would overlap awkwardly with the chairs and objects in Part 2, and the windo...
+c51b4a8d58f63301	train	jigsaw_1x2, label B: If I place Part 1 to the left of Part 2 (option A), the continuity of the wooden panel is disrupted: the left side of Part 2 would show the bed, and immediately to its right would be the wardrobe, which feels unnatura...
+e55c8fc59248efa9	train	jigsaw_1x2, label B: If I place part 2 to the left of part 1 (option A), the cabinet's halves would not align—the lamp would be off-center, and the objects on the cabinet would appear out of order.
+eaa76f12fc19d023	train	jigsaw_1x2, label A: If I place part 2 to the left of part 1 (arrangement A), the window and the sitting area naturally extend from the bed area, giving a coherent layout for a bedroom with a window seating space.
+274cbf366e176abe	train	jigsaw_1x2, label B: If part 1 were placed to the left of part 2 (option A), there would be an abrupt transition from the rug and table directly into the sideboard and trash bag, but the perspective and alignment would not match: the rug ...
+19ccb96d4c5f9031	train	jigsaw_1x2, label B: If I place piece 1 to the left of piece 2 (option A), the truncated right side of the sofa in piece 1 would not match up with the left armrest of the sofa in piece 2, as the armrests would be on opposite sides and the...
+0c8595eebb975d95	train	jigsaw_1x2, label A: If I were to place Part 2 to the left of Part 1 (option B), the visual elements would not join naturally: the lamppost and the signpost would be on the outside edges, and the building's roofline would be broken.
+9067609415f0e148	train	jigsaw_1x2, label B: If I were to place part 1 on the left and part 2 on the right (option A), the corner in part 1 would abruptly terminate into the window wall of part 2, which would disrupt the natural flow of the room.
+35ea77cfc97cfa45	train	jigsaw_1x2, label B: If I place part 1 on the left and part 2 on the right (option A), the image would show the tender first, followed by the locomotive, which is not how trains are structured—the locomotive leads, and the tender follows.
+d3a66a93be5de4c7	train	jigsaw_1x2, label B: If I place Part 2 to the right of Part 1 (option A), the objects do not line up naturally: the door and armchair would be separated, and the beds would not be side by side.
+537904f8a26aada1	train	jigsaw_1x2, label B: If Part 1 is placed on the left and Part 2 on the right (option A), the wall and outlet would be in the center of the assembled image, which is unnatural, and the bedspread would not align smoothly.
+2155e1e063b83b37	train	jigsaw_1x2, label B: If part 2 were placed to the left of part 1 (option A), the table would appear abruptly cut off by the left edge, and the perspective of the classroom would be disrupted.
+d476fcb51d78cf25	train	jigsaw_1x2, label A: If part 2 is placed to the left of part 1 (option A), the sloped ceiling's lines and wood joints flow naturally from right to left, without any abrupt mismatches.
+7c28f470d1ba1319	train	jigsaw_1x2, label B: If I were to place Piece 2 to the left of Piece 1 (Answer B), the boundary between the two pieces forms a continuous line for both the road and the trees, and the streetlight arm and lamp are reconnected.
+45d1b3b37c921635	train	jigsaw_1x2, label B: If piece 2 is placed to the right of piece 1 (answer A), the doorway and the vertical trim do not align smoothly—the right edge of piece 1 and the left edge of piece 2 both feature the wooden trim and counter, which w...
+43a80c9d0ecefc1a	train	jigsaw_1x2, label B: If we put part 2 on the left and part 1 on the right (option B), the shelf flows directly into the part of the room containing the stool, door, and chair, and the windows at the top are seamless.
+a9854287503746a5	train	jigsaw_1x2, label A: If I place Piece 2 to the left of Piece 1 (Choice A), the street shown in Piece 2 lines up naturally with the street and crosswalk in Piece 1.
+b8ab2cb647c620fd	train	jigsaw_1x2, label B: If I place piece 1 on the left and piece 2 on the right (choice A), the toilet would appear at the left side of the image, with the sinks and mirrors extending to the right, creating a continuous and realistic bathroo...
+95f53b3cb7132131	train	jigsaw_1x2, label A: If I place Piece 2 to the right of Piece 1 (Option A), the visual elements align seamlessly: the woven mat on the floor extends naturally, and the wall with its decorations continues smoothly.
+5498fb60ec68091f	train	jigsaw_1x2, label A: When I mentally place Part 2 on the left and Part 1 on the right (option A), the perspectives and the objects align naturally, forming a coherent scene.
+77ab1fadd27bdd96	train	jigsaw_1x2, label A: Placing Piece 1 on the left and Piece 2 on the right (Arrangement A) aligns the mirror halves and the wall sconces symmetrically, and the sink and rack appear in their logical positions.
+1e64957149961b05	train	jigsaw_1x2, label B: If I place piece 1 on the left and piece 2 on the right (option A), the refrigerator would be isolated and the kitchen counter would be separated from it, which feels unnatural.
+f0c0ba3365bd5be7	train	jigsaw_1x2, label B: If piece 2 is placed to the left of piece 1 (Arrangement B), the bed, hanging clothes, and poster line up seamlessly with the rest of the room, creating a natural flow from the sleeping area to the workspace and storage.
+8743b33ad0a996e5	train	jigsaw_1x2, label B: If I try placing Part 1 to the left of Part 2 (choice A), the orange chair would be on the leftmost side, and the blue chair would be abruptly cut off by the vertical white strip, not aligning smoothly into the next p...
+a7f29e6fcb1861af	train	jigsaw_1x2, label B: If I put Piece 1 to the left and Piece 2 to the right (option A), the couch would appear disconnected from the entertainment unit, and the foreground couch arm in Piece 2 would not align well.
+3a5fee438597522f	train	jigsaw_1x2, label A: If Piece 2 is placed on the left and Piece 1 on the right (Option A), the chest of drawers and the television will be on opposite sides of the room, matching the natural layout—furniture distributed across the space, ...
+445d1dbc4c957b71	train	jigsaw_1x2, label B: If I place Part 1 to the left of Part 2 (option A), the balcony and the rooflines do not align smoothly: the projecting bay and its railing seem disconnected from the porch and peaked roof.
+8bcd0b06caa549aa	train	jigsaw_1x2, label B: If we place piece 2 to the right of piece 1 (option B), all these details connect seamlessly, creating a coherent scene: the maroon column separates the two bins from the rest of the room, and the floor and wall detai...
+75111f96e25355b3	train	jigsaw_1x2, label A: If I place Piece 2 to the left of Piece 1 (option A), the staircase would begin on the left and curve awkwardly into a wall, breaking the natural flow of the architecture.
+570da9ce77331738	train	jigsaw_1x2, label A: If I place Piece 2 to the left of Piece 1 (Answer A), the edge of Piece 2, with its closer, larger frames, will transition naturally into the scene in Piece 1, where the people are standing further down the hallway an...
+4a50c95771f16659	train	jigsaw_1x2, label B: If I were to place Piece 1 on the right of Piece 2 (Choice A), the dresser would be on the far right, and the bed would be on the far left, breaking the continuity of furniture placement and room perspective.
+93411d701ba027c8	train	jigsaw_1x2, label B: If I place Part 2 to the left of Part 1 (option B), the gold chair in Part 2 aligns seamlessly with the ongoing table and people in Part 1, allowing the makeup table to form one continuous surface.
+161bd6dcb773a259	train	jigsaw_1x2, label A: If piece 1 is placed to the left of piece 2 (option A), the desk and wall connect seamlessly, and the refrigerator appears naturally at the right side of the room, adjacent to the desk.
+4567c9e9448bf62c	train	jigsaw_1x2, label B: If I place piece 2 to the right of piece 1 (option A), the table and bed would be separated by a wide gap, and the continuity of the furniture along the back wall would be broken.
+87704eea171833b1	train	jigsaw_1x2, label B: If piece 1 is placed on the left and piece 2 on the right (choice A), the table's edge and chairs do not align properly, and the perspective appears broken.
+798640130dc6bce3	train	jigsaw_1x2, label A: If I place Piece 2 to the left of Piece 1 (i.e., arrangement A), the leftmost edge of Piece 2 matches the rightmost edge of Piece 1: the building's curves and the horizontal bands align perfectly, and the structures a...
+364f158000b94a84	train	jigsaw_1x2, label B: If the pieces were assembled with Piece 1 on the left and Piece 2 on the right (Option A), the door frame and lamp would be visually disjointed, and the perspective of the room would not flow naturally.
+0eb0c4965ce8cba2	train	jigsaw_1x2, label B: If I place part 2 to the left of part 1 (choice A), the dresser and mirror would be awkwardly positioned next to the bed, and the room's layout would seem unnatural.
+f68b87556e40edaa	train	jigsaw_1x2, label B: If I position Part 2 to the left of Part 1 (i.e., B), the glass door and TV stand in Part 2 would naturally be on the left side of the room, followed by the table and laptop from Part 1 on the right.
+db994b3f4c0afb34	train	jigsaw_1x2, label A: If Part 2 is placed on the left and Part 1 on the right (option B), the kitchen counter would be immediately adjacent to the glass table.
+54dc35c3e64c940f	train	jigsaw_1x2, label A: If I place Part 1 on the left and Part 2 on the right (option B), the plant would be in the center of the room, and the window would be split awkwardly by the jigsaw seam, disrupting the natural flow.
+da4c9fb89876d582	train	jigsaw_1x2, label B: If I were to place Part 1 to the left of Part 2 (option A), the TV would be oddly positioned at the center, and the white wardrobe would be cut off from the left wall, disrupting the room's natural layout.
+87b337e38b0f2c0a	train	jigsaw_1x2, label B: If I place Part 1 to the left of Part 2 (choice A), the angles of the shelves would not match up: the shelving in Part 1 would abruptly end at the white strip, while the shelving in Part 2 would not continue smoothly ...
+131fba405c871fb0	train	jigsaw_1x2, label A: If I place Piece 1 on the left and Piece 2 on the right (A), the floral arrangement and fireplace on the left will lead naturally into the extended table and red chairs, with the rope and carpet flowing smoothly.
+e1bdd3131f6799e0	train	jigsaw_1x2, label B: If I were to put Part 2 on the left and Part 1 on the right (option A), the garage door would end up in the center of the house, which does not make architectural sense.
+264741c5ec025857	train	jigsaw_1x3, label C: Testing possible arrangements
+87e088c610bb4a36	train	jigsaw_1x3, label B: Testing the possible options
+b7f943467613f04b	train	jigsaw_1x3, label C: Testing possible arrangements
+70d2b1ffd1d1a764	train	jigsaw_1x3, label B: Testing possible combinations
+79ce19067c88b240	train	jigsaw_1x3, label D: Checking the options
+49b7f1759090316d	train	jigsaw_1x3, label D: - Considering the available options, if Piece 1 is left, 3 is middle, and 2 is right (Option D), let's check the alignment
+da7e3af303804383	train	jigsaw_1x3, label D: Testing each option
+6291bf5018532e93	train	jigsaw_1x3, label A: Checking the options
+066be484b0e2c12c	train	jigsaw_1x3, label A: When picturing the possible arrangements
+faf84cbd89a78280	train	jigsaw_1x3, label B: Testing the options
+7a3fa0cdc6f3114b	train	jigsaw_1x3, label C: Testing the options
+901ca079da5462eb	train	jigsaw_1x3, label C: Testing the options
+8790036e71f7d9ca	train	jigsaw_1x3, label C: If I place Part 1 on the left, Part 2 in the center, and Part 3 on the right (Option D), the perspectives and features do not align properly: the sidewalk and building lines on Part 1 and Part 2 do not continue cleanl...
+8e94077ebb4098e8	train	jigsaw_1x3, label B: Testing the options
+4f1730b658516197	train	jigsaw_1x3, label D: Testing possible arrangements
+7c385aa629c42176	train	jigsaw_2x1, label B: If I were to place Part 2 beneath Part 1 (Option A), the countertop and object would align perfectly, creating a seamless transition from the top piece down to the wider kitchen scene.
+9059c0e55ed7a170	train	jigsaw_2x1, label A: If I place Part 2 above Part 1 (option A), the open door and the more detailed room would be above a plain floor section, which makes sense in a natural spatial progression: the upper piece shows the room and doorway,...
+74869cb25af9f6ec	train	jigsaw_2x1, label B: If I were to place Part 1 above Part 2 (Option A), the bed would float above the wall, defying both gravity and interior design logic.
+54bc67813351d800	train	jigsaw_2x1, label B: If I were to place part 2 above part 1 (option A), the bed and trash can would be misaligned, breaking the continuity and natural flow of the scene.
+f55f572db91afe19	train	jigsaw_2x1, label B: - If Part 1 were placed above Part 2 (Option A), the floor would unnaturally transition into the tops of chairs and a green wall, which defies typical spatial arrangement—floors are at the bottom, walls at the top.
+b8bd1536d42272cd	train	jigsaw_2x1, label B: If I place Part 2 on top of Part 1 (option B), the upper portions of the furniture and room layout in Part 2 will naturally continue downward into the lower portions depicted in Part 1.
+b1c84252ac805aca	train	jigsaw_2x1, label A: If I place Part 2 above Part 1 (Option A), the features of the sofa, the floor, and the background wall merge seamlessly, creating a coherent living room scene.
+c5ebdb16ebcd192a	train	jigsaw_2x1, label B: If I were to place Part 2 above Part 1 (option A), the result would be visually incoherent: the water and street would be at the top, and the rooftops and sky at the bottom, which is unnatural.
+dc0e9ff97ba451c2	train	jigsaw_2x1, label B: If I were to place Part 1 above Part 2 (choice A), the legs of the chairs would float above the backs and tabletop, which would look unnatural and break the perspective.
+c22860448de8626f	train	jigsaw_2x1, label B: If I were to place Part 1 below Part 2 (option A), the sofa's top would be under its base, which is unnatural.
+9a0b8b5541a86e32	train	jigsaw_2x1, label B: If Part 1 were placed above Part 2 (option A), the orange floor would unnaturally continue above a scene that already has a complete room with furniture and walls, causing a visual disconnect.
+6e096f75bf69a80c	train	jigsaw_2x1, label B: If I place Part 2 on top of Part 1 (choice B), all these elements connect seamlessly: the tabletop edge, lamp bases, chair positions, carpet, and wooden leg all form a natural and continuous scene.
+5d3f1570479ec45b	train	jigsaw_2x1, label B: If I place piece 2 above piece 1 (option B), the bed in piece 2 aligns perfectly with the close-up in piece 1, creating a coherent perspective from near to far.
+7e45d7a5268bac22	train	jigsaw_2x1, label A: If I were to place Part 2 above Part 1 (answer A), the ceiling, wall signs, and upper parts of the doors would be on top, and the floor, lower walls, and door bases would be on the bottom, creating a coherent and real...
+68d04d33591b1d5f	train	jigsaw_2x1, label A: If I were to place Part 2 above Part 1 (option A), the desk, wall, and decorations would rest above the floor, which matches how rooms are structured in reality.
+0af39cda3ee44bec	train	jigsaw_2x1, label B: Placing piece 2 above piece 1 (option B) creates a seamless transition: the tiled wall, sink, and painting sit naturally above the floor, and the base of the sink aligns perfectly with the floor's surface in piece 1.
+0ae3ccd6be772a24	train	jigsaw_2x1, label B: If I place Part 1 beneath Part 2 (choice A), the sky and rooftops would appear below the street and cars, which is unnatural: lamp posts and pillars would be abruptly cut off in the wrong direction, and the scene woul...
+50996c1252dc4dfa	train	jigsaw_2x1, label B: If I place Part 2 above Part 1 (option A), the floor would be positioned above the wall and the bed's lower half would be above its upper half, which defies the laws of perspective and gravity.
+5dd8b44aa8a8794e	train	jigsaw_2x1, label B: If I were to place Part 2 above Part 1 (Option A), the floor would incorrectly be at the top and the refrigerator/counter would be upside down, breaking the natural layout.
+2bbb5db6d7632d95	train	jigsaw_2x1, label A: If I were to place Part 1 below Part 2 (choice A), all of these elements would align correctly: the countertop, objects on top, and the cabinet below.
+4a587a82906454ea	train	jigsaw_2x1, label B: If Piece 1 is placed above Piece 2 (option A), the chairs would float above the conference table and screen, which would not make sense spatially.
+95d4de88b5b8a0ca	train	jigsaw_2x1, label B: If Piece 1 is placed on top of Piece 2 (choice A), the lower sections of objects would appear above their upper sections, breaking the natural vertical flow of the room.
+d7d6447227a04e82	train	jigsaw_2x1, label B: If we try placing Piece 1 above Piece 2 (option A), the round cushion would appear to float above the rest of the room, which is visually and physically inconsistent, as cushions are normally placed on the floor.
+053b30826db1f7b1	train	jigsaw_2x1, label B: If I were to place Part 2 below Part 1 (option A), the blue vase and framed photos would unnaturally appear beneath the desk and floor, which would break the spatial logic of the scene.
+9947b653a790354c	train	jigsaw_2x1, label B: If the pieces were arranged with piece 1 above piece 2 (Option A), the visual flow would be broken—the bathroom would appear below the bedroom, disrupting the natural perspective and continuity of the doorway, floor, ...
+38eb905cf1fcb4be	train	jigsaw_2x1, label A: If I place piece 2 above piece 1 (option A), the door and shelving unit would be split awkwardly, and the perspective would not make sense, as the view of the room would be cut off at the top.
+8b4e75334d205c43	train	jigsaw_2x1, label B: If piece 2 were placed above piece 1 (option A), the floor and garbage bag would bizarrely appear above the room's walls and furnishings, which is illogical and breaks the expected perspective.
+a55ef50d0eded52a	train	jigsaw_2x1, label B: If I were to place Piece 1 above Piece 2 (Option A), the table legs and shadow would awkwardly appear above the tabletop and the rest of the room, breaking the natural continuity of furniture and space.
+66270e1207516f30	train	jigsaw_2x1, label B: If I placed Part 2 above Part 1 (option A), the chairs and lower part of the table would appear above the countertop and window, which would be unnatural and disrupt the continuity of the scene.
+ed93cb9799a9e828	train	jigsaw_2x1, label A: If I place Part 1 above Part 2 (option B), the trash bin would float in the air, which is unnatural.
+9e381367807927bf	train	jigsaw_2x1, label B: If I were to place Part 2 below Part 1 (Option A), the greenish-blue floor would be below the light floor, which would not make sense for the perspective of a continuous room—the foreground should lead into the backgr...
+b0ba323ad8babf90	train	jigsaw_2x1, label B: If Piece 1 were placed below Piece 2 (answer B), the path in the foreground would seamlessly extend into the background, with the shadows and edges lining up perfectly.
+b8d70634e523e9d5	train	jigsaw_2x1, label A: If I were to place Part 2 above Part 1 (option B), then the toilet would appear to float above the dresser; the parquet flooring would be interrupted in an unnatural way, and the perspective would be broken.
+eef41bab7f12df0f	train	jigsaw_2x1, label B: If I were to place Part 1 beneath Part 2 (choice A), the windows would be below the sidewalk, which defies normal perspective and structure.
+b7d39c46881f8d29	train	jigsaw_2x1, label A: If I were to place Part 2 below Part 1 (choice B), the window would awkwardly be below the sink and bathtub, which is unnatural and illogical for a bathroom layout.
+5ba9003b9616a21c	train	jigsaw_2x1, label B: Placing part 1 on top and part 2 on the bottom (answer B) aligns the floor and furniture perfectly, with the upper portion of the room above and the entrance below, reproducing a realistic viewpoint.
+22248cf510c43ff8	train	jigsaw_2x1, label B: If I were to place part 1 above part 2 (option A), the floor would appear above the walls and doors, which would break the natural logic of a room: the floor should always be below the wall and furniture.
+9bb7198d4133d45b	train	jigsaw_2x1, label B: If I were to place part 2 above part 1 (choice A), the scene would look unnatural: the floor would be floating above the doors and walls, and the lower half of the cart would be above its upper half.
+67b7de1acb7486f1	train	jigsaw_2x1, label A: If I place Part 2 above Part 1 (i.e., Option A), the table and chairs assemble into their correct forms: the tabletop sits atop its base, and the chairs are whole.
+ce56739c61281100	train	jigsaw_2x1, label B: Placing Part 2 above Part 1 (as in option A) would create a nonsensical arrangement, with the floor and shadows floating above furniture and bed frames.
+b68c6a32313d0755	train	jigsaw_2x1, label B: If I place Part 1 on top of Part 2 (option A), the red car appears above the rest of the scene, disconnected from its natural context, and the perspective feels off because the foreground (the car) should be at the bo...
+1a849a93c0a7ce2e	train	jigsaw_2x1, label B: If I place piece 2 above piece 1 (i.e., option B: part 2 should be to the top of part 1), the room's perspective and the continuity of the furniture lines up naturally: the ceiling is on top, the back wall and furnitu...
+fd95aa0e211b6dfb	train	jigsaw_2x1, label A: If I place part 1 below part 2 (arrangement B), the result would be physically correct: the floor (with chair/table legs and shadows) appears below the tabletop and wall, producing a coherent scene.
+06f6cee7853e5f2d	train	jigsaw_2x1, label B: If I were to place piece 1 below piece 2 (option A), the room would look unnatural because the ceiling and wall objects would be at the bottom, and the floor and lower furniture would be above them.
+381e1d7615d0d551	train	jigsaw_2x1, label B: If I consider alternative (A), placing Part 2 above Part 1, the visual continuity would be disrupted: the vanity base and toilet bowl would appear at the top of the image, with the sink and faucet below them, which is...
+1db719fc6a2c7a41	train	jigsaw_2x1, label B: If we place piece 2 above piece 1 (i.e., answer B), the dresser aligns perfectly, forming a single coherent piece of furniture from top to bottom.
+ed6c306d3540f9a1	train	jigsaw_2x1, label B: If we were to place piece 2 on top of piece 1 (choice A), the sinks and floor would float above the mirrors and lights, which is visually and physically unnatural.
+9891f184b3fbe53f	train	jigsaw_2x2, label B: If I put these together
+1b7c895e65192af0	train	jigsaw_2x2, label B: Checking the options
+744abc6f736fc69d	train	jigsaw_2x2, label D: Checking the options
+b815d7c1a9a591cf	train	jigsaw_2x2, label A: Testing the options
+f3bd8002b5ab7927	train	jigsaw_2x2, label D: Testing the options
+a794cb479912b51f	train	jigsaw_2x2, label A: Testing the options
+ca94258fbec91ebc	train	jigsaw_2x2, label B: Comparing these placements to the options
+d20456a184dd6481	train	jigsaw_2x2, label C: Checking the options
+963aa66608c4fd36	train	jigsaw_2x2, label B: Checking the options
+7aff0afc524d5935	train	jigsaw_2x2, label C: Examining the choices
+89a29bb336c6cd11	train	jigsaw_2x2, label D: If we align the countertop and backsplash, and match the drawer/countertop transitions, the only arrangement that fits all these cues is
+dff85540822bc003	train	jigsaw_2x2, label C: Testing the options
+ff6d743ab8003ea8	train	jigsaw_2x2, label B: When I assign the labels to their appropriate positions for the natural, reconstructed image, the arrangement is
+c986557791373a4e	train	jigsaw_2x2, label B: Placing these clues together
+48f8f9929764473b	train	jigsaw_2x2, label D: Comparing these deductions to the options given
+799d3c2f4fedd2dd	train	jigsaw_2x2, label B: If we assign positions
+ebc01f3faf1974d9	train	jigsaw_2x2, label B: Checking the four options
+72923e33e62f60a0	train	jigsaw_2x2, label B: If I arrange the pieces as in statement (A), the top row would be Piece 1 (stairs/floor) left and Piece 2 (right wall/corner) right.
+38a8f177708d1316	train	jigsaw_2x2, label A: Testing the options
+b9e8a06ada3636b5	train	jigsaw_2x2, label B: Testing possible arrangements
+030c77beec7b4b05	train	jigsaw_2x2, label C: Testing the possible configurations
+6d88d9880d64a7af	train	jigsaw_2x2, label D: Checking the options
+4ce289e694859b04	train	jigsaw_2x2, label C: Comparing to the provided choices
+ce98677d3881f0db	train	jigsaw_2x2, label C: Testing the options
+d3bcedd97367d09d	train	jigsaw_2x2, label B: Checking the options
+7f07e651c134e9f2	train	jigsaw_3x1, label C: Checking the provided answer choices
+fc1c059ead8d1594	train	jigsaw_3x1, label B: If I try placing Part 1 at the top, Part 2 in the middle, and Part 3 at the bottom (Option C), the alignment between the blanket and sofa does not flow naturally from Part 2 to Part 3—the blanket appears to 'jump' rat...
+315632d7cb9d74ce	train	jigsaw_3x1, label B: Testing the arrangements
+0a6c2be322ae04bd	train	jigsaw_3x1, label B: If I stack Piece 1 on top, Piece 3 in the middle, and Piece 2 at the bottom (Option B), the objects, window, and floor all align naturally, creating a coherent room scene.
+330d7ceaa2dc9433	train	jigsaw_3x1, label B: Testing the options
+8f1e4fd5e70118a0	train	jigsaw_3x1, label A: Testing the options
+93dcdbff9c93d8e3	train	jigsaw_3x1, label B: Testing possible arrangements
+e4c1bf7e5d79f72e	train	jigsaw_3x1, label B: Testing the provided options
+e1eccbf206ee791e	train	jigsaw_3x1, label B: If Piece 3 is on top, Piece 2 in the middle, and Piece 1 at the bottom, the correct sequence is
+76fdcae5437f05c8	train	jigsaw_3x1, label C: Testing possible combinations
+46373e48c63496ad	train	jigsaw_3x1, label D: Checking the options
+3c6dcab668234ff6	train	jigsaw_3x1, label B: Testing possible arrangements
+464e972bafbbe717	validation	jigsaw_1x2, label B: If Part 1 were placed to the left of Part 2 (option A), the visual continuity would be disrupted: the desk curves would not match up, the chair would not fit correctly under the desk, and the filing cabinet would not ...
+0b38c124f3975f87	validation	jigsaw_1x2, label B: If I place Part 1 to the left of Part 2 (option A), the double-door would be split incorrectly and the flooring transition would look unnatural, as the reddish-brown floor would abruptly meet the beige wall and light ...
+5b47fccadd17f6ef	validation	jigsaw_1x2, label B: If I place Piece 1 to the left of Piece 2 (option A), the dresser would be directly next to the open doors, which seems awkward and visually unnatural—the dresser appears to extend into the doorway space, disrupting t...
+3cc7ce55003a05d5	validation	jigsaw_1x2, label B: If I place piece 1 on the left and piece 2 on the right (option A), the table and chairs would be next to the couch, which does not align with typical room arrangements, as the door would be in the middle, splitting t...
+976c03a7c420d753	validation	jigsaw_1x2, label B: If I place Part 1 on the left and Part 2 on the right (option A), the patterns on the canopy and bedspread do not align smoothly—the canopy appears disjointed, and the window positions look inconsistent.
+1e1b41be758e6f07	validation	jigsaw_2x1, label B: If Part 1 were placed below Part 2 (choice A), the shelf would inexplicably be under the desk, which is not consistent with standard office setups.
+d7e81d80ae569e0b	validation	jigsaw_2x2, label D: Testing each option
+bd2635fdb313bf00	validation	jigsaw_3x1, label B: - Placing Piece 1 at the top, Piece 2 in the middle, and Piece 3 at the bottom (Option C) seems plausible, but the framed picture would be unnaturally high on the wall, and the lamp base would not align well.

reports/known_issues/summary.json

@@ -0,0 +1,47 @@
+{
+ "control_char_residue": {
+  "description": "Removing upstream's control bytes left printable fragments in the text: a garbled \\boxed ('[boxed{B}', 'cboxed{A}', 'oxed{A}', '[0m') or a part number cut to '02 should be to the top of Part 1'; the model is trained to write them.",
+  "rows": 49,
+  "train": 49,
+  "validation": 0,
+  "file": "known_issues/control_char_residue.tsv",
+  "detector": "row flagged S0.control_chars_stripped and a thought contains 'boxed{X}' / 'oxed{X}' not preceded by a backslash, an ANSI remnant '[0m' / '[.}', or a '0<digit> should be' part name",
+  "measured": "2026-09-25; known_issues.py sha1 655e5307; export 4aaee1f rows 83cf9966/ae9b703a"
+ },
+ "readback_opens_with_moved_comparison": {
+  "description": "Flag S4.plan_verdict_moved and the read-back opens with sentences moved from the plan that read as a comparison of the options ('If I place Part 1 below Part 2 (option A), ...') rather than a verdict, so the read-back starts with pre-image reasoning; judged on the opening sentence only, so some listed rows' moved text does end in a verdict (mostly three- and four-part puzzles).",
+  "rows": 172,
+  "train": 164,
+  "validation": 8,
+  "file": "known_issues/readback_opens_with_moved_comparison.tsv",
+  "detector": "row flagged S4.plan_verdict_moved and the read-back's first sentence starts with If/Suppose, or starts with When/Testing/Checking/Placing/Comparing/Considering/Evaluating/Examining the options and does not itself give a verdict (matches, corresponds, the correct, the only, most natural, should be, natural order, ...)",
+  "measured": "2026-09-25; known_issues.py sha1 655e5307; export 4aaee1f rows 83cf9966/ae9b703a"
+ },
+ "target_shared_with_other_row": {
+  "description": "The same picture (decoded pixels identical) is the target of another shipped row: one photograph or render under two upstream file names, cut into different puzzles.",
+  "rows": 12,
+  "train": 12,
+  "validation": 0,
+  "file": "known_issues/target_shared_with_other_row.tsv",
+  "detector": "md5 of the decoded target pixels occurs on more than one shipped row",
+  "measured": "2026-09-25; known_issues.py sha1 655e5307; export 4aaee1f rows 83cf9966/ae9b703a"
+ },
+ "text_states_other_arrangement": {
+  "description": "The target image and <answer> show the labelled arrangement, but the text does not: the plan concludes, or the read-back says it assembled or chose, a different arrangement (or names another option's letter).",
+  "rows": 90,
+  "train": 88,
+  "validation": 2,
+  "file": "known_issues/text_states_other_arrangement.tsv",
+  "detector": "3/4-part: a full position->part statement that is not hypothetical and not an enumerated rejected option, differs from the label (the plan's last such statement, any in the read-back); 2-part: a placing/'Part X should be left of Part Y' statement that puts the other part first; any part: 'matches/corresponds to (X)' with X not the answer, or '(X)' followed by option Y's text",
+  "measured": "2026-09-25; known_issues.py sha1 655e5307; export 4aaee1f rows 83cf9966/ae9b703a"
+ },
+ "validation_target_near_copy_of_train": {
+  "description": "This validation row's target is a near-copy of a training row's target (the same scene shot from a slightly shifted viewpoint), so it is not fully held out.",
+  "rows": 1,
+  "train": 0,
+  "validation": 1,
+  "file": "known_issues/validation_target_near_copy_of_train.tsv",
+  "detector": "validation target within 10 bits (of 256) of a training target by a 17x16 greyscale difference hash; the next-nearest validation-train pair on r2 is 43 bits apart",
+  "measured": "2026-09-25; known_issues.py sha1 655e5307; export 4aaee1f rows 83cf9966/ae9b703a"
+ }
+}

reports/known_issues/target_shared_with_other_row.tsv

@@ -0,0 +1,13 @@
+# The same picture (decoded pixels identical) is the target of another shipped row: one photograph or render under two upstream file names, cut into different puzzles.
+b477f54ba9491c9d	train	same target as 48ccb324817e04b8 (train, ade20k_00009-of-00010_001662_ADE_train_00020718)
+ebd5964bccf870f9	train	same target as 01ce85a03268e9b2 (train, sat_seg12_image_005021)
+48ccb324817e04b8	train	same target as b477f54ba9491c9d (train, ade20k_00009-of-00010_001660_ADE_train_00007586)
+ab928cf1c384946a	train	same target as 72e7c59e536a87cf (train, sat_seg03_image_001278)
+01ce85a03268e9b2	train	same target as ebd5964bccf870f9 (train, sat_seg12_image_005022)
+4e2c40d0e4c74e34	train	same target as a184a53a8fa974eb (train, ade20k_00009-of-00010_000898_ADE_train_00022577)
+72e7c59e536a87cf	train	same target as ab928cf1c384946a (train, sat_seg03_image_001279)
+75352c1832973336	train	same target as 4b9fd86db4872395 (train, sat_seg12_image_005281)
+a184a53a8fa974eb	train	same target as 4e2c40d0e4c74e34 (train, ade20k_00009-of-00010_000895_ADE_train_00019973)
+a4f3fe6977038d22	train	same target as 19354e0795f7096a (train, sat_seg13_image_005724)
+4b9fd86db4872395	train	same target as 75352c1832973336 (train, sat_seg12_image_005280)
+19354e0795f7096a	train	same target as a4f3fe6977038d22 (train, sat_seg13_image_005725)

reports/known_issues/text_states_other_arrangement.tsv

@@ -0,0 +1,91 @@
+# The target image and <answer> show the labelled arrangement, but the text does not: the plan concludes, or the read-back says it assembled or chose, a different arrangement (or names another option's letter).
+d40a6f95cec102f3	train	label A: plan puts Part 1 first: meet the pale flooring, which would be unnatural. However, putting Part 2 to the right of Part 1 allows the white wall to act as a struc
+235cfd1ee1b995e9	train	label A: plan puts Part 1 first: om's function and likely would not be obscured. Conversely, placing piece 2 to the right of piece 1 allows the refrigerator to sit naturall
+af28e3baf7456c7f	train	label A: plan puts Part 1 first: tive consistency and spatial logic of the room. Conversely, placing Part 2 to the right of Part 1 allows the tables and chairs to flow na
+79951484e6a90fbd	train	label A: plan puts Part 1 first: , which does not align with the natural viewpoint. However, placing piece 2 to the right of piece 1 allows the door to open into the room,
+eb30e46819a3aaae	train	label A: plan puts Part 1 first: erspective, objects, and lighting all support the idea that Piece 2 belongs to the right of Piece 1 for a natural, continuous street view.
+152ac4aff5b3c0fa	train	label B: plan puts Part 1 first: ld block access and sight lines to these features. However, putting part 2 to the right of part 1 allows a natural flow from the door and
+75111f96e25355b3	train	label A: read-back puts Part 1 first: ll, breaking the natural flow of the architecture. However, placing Piece 2 to the right of Piece 1 (option B), the staircase begins in the
+334f070f9b01f8b8	train	label B: plan puts Part 1 first: indicate that these items are more centrally located. When placing Part 2 to the right of Part 1, the floor hatch marks, red line, and r
+10d4031e348b0460	train	label A: plan puts Part 1 first: ms and a discontinuous carpet and valance line. Conversely, placing part 1 to the left of part 2 aligns the log beams perfectly, creates
+7678fe0b40853360	train	label D: plan states (1, 2, 3): gn well. Therefore, the most natural reconstruction places Part 1 on the left, Part 3 on the right, and Part 2 in the middle. This arrangement maintains t
+6c5ef7f688ac2c1e	train	label A: read-back states (1, 2, 3): ing these observations together, the correct assembly is: - Left: Part 1 (table edge, plain wall) - Middle: Part 2 (open door, visible bed) - Right: Part 3 (table, candlestick, picture,
+259a1ccd56cc22c2	train	label A: names (C): This corresponds to statement (C). However, reviewing the answer choices,
+f205bb08244c1337	train	label D: plan states (3, 2, 1): t logical arrangement to reconstruct the original image is: left—Piece 3, middle—Piece 2, right—Piece 1. This matches the visual flow | read-back states (3, 2, 1): e assembled the image according to my proposed arrangement (Piece 3 on the left, Piece 2 in the center, Piece 1 on the right), I will verify the coherence
+ad4d45e6914588c5	train	label D: plan states (1, 2, 3): nto piece 3. Looking at the base of the wall and the floor, piece 1 must be on the left, piece 2 in the middle, and piece 3 on the right to allow all lines, shadows,
+d9e2319ad9123cc2	train	label B: plan states (1, 2, 3): re is empty space and the painting continues. - The door in Piece 3 is at the right edge, meaning Piece 3 should be the rightmost segment, following the sink and toilet. - The flooring patterns and...
+b7485436d579b3bf	train	label B: plan states (1, 2, 3): The overall perspective and lighting remain consistent when Part 2 is placed in the middle, flanked by Part 1 on the left and Part 3 on the right. This arrangement allows the
+aee8491f4034f1b9	train	label A: plan states (1, 2, 3): dle), 3 (right). Therefore, the correct assembly should be: left part - Part 1; middle part - Part 2; right part - Part 3. | read-back states (1, 2, 3): ring to the answer choices, this matches statement (A): The left part should be Part 1; the middle part should be Part 2; and the right part should be Part 3. To verify the correctness of
+03dfb8bfb2bc202f	train	label B: names (A): This matches choice (A). But let’s check the options to ensure
+d9455cfb9aca8ef7	train	label B: plan states (1, 2, 3): rts in Part 1 and continues into Part 2. This suggests that Part 1 should be on the left and Part 2 in the middle, so the stick appears seamless. The black table or stand is partially visible in Pa...
+d328174cce4b5a1e	train	label A: read-back states (1, 2, 3): cording to the choices, this corresponds to answer (A): The left part should be Part 1; the middle part should be Part 2; and the right part should be Part 3. Upon examining the assembled | read-back states (1, 2, 3): onfirms that the correct order to reconstruct the image is: left part should be Part 1, middle part should be Part 2, and right part should be Part 3. The answer: (A).
+be3c09cdd93b9021	train	label A: plan states (1, 2, 3): of shadows and boundaries, the most natural arrangement is: Part 1 on the left, Part 2 in the middle, and Part 3 on the right.
+09e01d204a4c3ae8	train	label B: plan states (1, 2, 3): ally on the left, with the blue-green wall and furniture. - Piece 3 fits on the right, with the woven carpet and white stool, matching the rightmost part of the room. - Piece 2 fits in the center, ...
+9634f107a7b2d40b	train	label C: plan states (3, 2, 1): nd window at right. Therefore, the correct arrangement is: left: Part 3, center: Part 2, right: Part 1. | read-back states (3, 2, 1): ully examine the assembled image resulting from positioning Part 3 on the left, Part 2 in the center, and Part 1 on the right. The first aspect to check is
+4e6678bd35d58ebc	train	label A: plan states (1, 2, 3): nd Part 3 on the right. Therefore, the correct assembly is: left = Part 1, middle = Part 2, right = Part 3. Instead, the table needs to | names (B): This matches option (B). However, re-examining the options, opt
+82e5ce2108a5cfa6	train	label C: plan states (1, 2, 3): oom’s layout. Therefore, to reconstruct the original image: Part 2 should be in the center, Part 1 on the left, and Part 3 on the right.
+e4819bc42d22a4c6	train	label C: plan states (1, 2, 3): boundaries—it is clear that the correct reconstruction is: Left: Part 1, Middle: Part 2, Right: Part 3. | read-back states (1, 2, 3): ing with the given options, this matches statement (C): The left part should be Part 1; the middle part should be Part 2; and the right part should be Part 3. Now that I have reconstructe
+5c640f0443fd608d	train	label A: plan states (1, 2, 3): 1 (left side) to Part 3 (right side) strongly suggests that Part 1 should be on the left and Part 3 on the right. However, the monitor's positioning and the bar counter indicate that Part 2 must be...
+24726213280b131e	train	label A: read-back states (1, 2, 3, 4): natural arrangement to reconstruct the original image is: - Top-left: Piece 1 - Top-right: Piece 2 - Bottom-left: Piece 3 - Bottom-right: Piece 4 However, reviewing the answe
+995bb24b60684a55	train	label A: read-back states (3, 4, 1, 2): ccording to visual continuity, the best configuration is: - Top-left: Part 3 (headboard, left nightstand) - Top-right: Part 4 (window, right nightstand) - Bottom-left: Part 1 (footboard) - Bottom-r... | names (C): . My configuration puts Part 1 at bottom-left and Part 2 at bottom-right, which matches with (C). Examining the assembled image, I can n
+24fc551551376424	train	label D: read-back states (1, 2, 3, 4): a natural office scene. Therefore, the answer is: (D) The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 3; and the bottom-right part shoul...
+c282fb95913d47fd	train	label C: read-back states (1, 2, 3, 4): Therefore, the correct assembly is: Top-left: Part 1 Top-right: Part 2 Bottom-left: Part 3 Bottom-right: Part 4 This matches statement (C):
+047df34d80643a07	train	label A: read-back states (1, 4, 3, 2): Thus, the correct arrangement is: - Top-left: Piece 1 - Top-right: Piece 4 - Bottom-left: Piece 3 - Bottom-right: Piece 2 This matches statement (A): | read-back states (1, 4, 3, 2): 3 - Bottom-right: Piece 2 This matches statement (A): "The top-left part should be Part 1; the top-right part should be Part 4; the bottom-left part should be Part 3; and the bottom-right part shou...
+8faf2c6396e12a71	train	label A: read-back states (1, 2, 3, 4): nes, and people—it becomes clear the correct assembly is: - Top-left: Piece 1 - Top-right: Piece 2 - Bottom-left: Piece 3 - Bottom-right: Piece 4 Reviewing the options, this
+3fae5262bbf2db8c	train	label A: read-back states (1, 2, 3, 4): t) Looking at the options, this matches statement (A): The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 3; and the bottom-right part shou... | read-back states (1, 2, 3, 4): and undisturbed, confirming that the arrangement of pieces—top-left: Part 1, top-right: Part 2, bottom-left: Part 3, bottom-right: Part 4—is correct. Thus, the answer
+a49aff5c7c34417c	train	label B: read-back states (4, 2, 1, 3): alls, floors, and objects, matching the natural layout: - Top-left: Part 4 (window) - Top-right: Part 2 (door, plant) - Bottom-left: Part 1 (bed) - Bottom-right: Part 3 (open doorway, chair) Compar
+8ec60fbe31a6e1ce	train	label A: read-back states (1, 2, 3, 4): e furniture aligns naturally, the correct arrangement is: - Top-left: Part 1 - Top-right: Part 2 - Bottom-left: Part 3 - Bottom-right: Part 4 Reviewing the provided answe
+44163ae275cd61b5	train	label D: read-back states (1, 2, 3, 4): e continuity of objects and surfaces, the arrangement is: - Top-left: Part 1 - Top-right: Part 2 - Bottom-left: Part 3 - Bottom-right: Part 4 Reviewing the provided choic
+fe60e8b7ee9a3508	train	label D: read-back states (1, 2, 3, 4): angement, mapping the piece numbers to their positions, is: Top-left: Piece 1 Top-right: Piece 2 Bottom-left: Piece 3 Bottom-right: Piece 4 This corresponds to statemen | read-back states (1, 2, 3, 4): ttom-right: Piece 4 This corresponds to statement (D): The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 3; and the bottom-right part shou...
+8e6c4c2397e34a36	train	label A: plan states (1, 2, 3, 4): ft). Combining these observations, the correct assembly is: top-left: part 1, top-right: part 2, bottom-left: part 3, bottom-right: part 4. The continuity of the table, | read-back states (1, 2, 3, 4): horough visual check, I confirm that the answer is (A): The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 3; and the bottom-right part sho...
+13dd55b3450bf9d9	train	label B: read-back states (1, 4, 3, 2): Therefore, the correct arrangement is: - Top-left: Part 1 - Top-right: Part 4 - Bottom-left: Part 3 - Bottom-right: Part 2 This matches statement (B): | read-back states (1, 4, 3, 2): t 3 - Bottom-right: Part 2 This matches statement (B): The top-left part should be Part 1; the top-right part should be Part 4; the bottom-left part should be Part 3; and the bottom-right part shou...
+a15391a5cb027eb8	train	label A: read-back states (1, 2, 3, 4): these observations together, the correct arrangement is: - Top-left: Piece 1 (door/wall, picture, TV, blue wall) - Top-right: Piece 2 (window, plant, TV stand, doorframe) - Bottom-left: Piece 3 (wo...
+070c5e56e50991e1	train	label D: (A) quoted with the text of (D): fident that the correct arrangement is: (A) The top-left part should be Part 1; the top-right part should be Part 4; the bottom-left part should be Part 3; and the
+2d8330c2eaa6d334	train	label A: read-back states (1, 2, 3, 4): have visualized the assembled image based on the placement: Part 1 in the top-left, Part 2 in the top-right, Part 3 in the bottom-left, and Part 4 in the bottom-right. Upon close inspection, sever | read-back states (1, 2, 3, 4): rrangement is correct and coherent. The answer is: (A) The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 3; and the bottom-right part shou...
+ca94258fbec91ebc	train	label B: read-back states (1, 2, 3, 4): Comparing these placements to the options: - Top-left: Part 1 - Top-right: Part 2 - Bottom-left: Part 3 - Bottom-right: Part 4 But reviewing the options, o
+699fb178fcfe8f2a	train	label C: plan states (1, 2, 3, 4): are consistent as well. Therefore, the correct assembly is: top-left = Part 1, top-right = Part 2, bottom-left = Part 3, bottom-right = Part 4. | read-back states (1, 2, 3, 4): This matches statement (C): The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 3; and the bottom-right part should be Part 4. Now, with the...
+473c8c88f1a2dc52	train	label B: plan states (1, 2, 3, 4): tom left side of the shaft. Let's test this arrangement: - Top-left: Part 1 (pipes starting on the left wall, fixture present) - Top-right: Part 2 (brighter, floor edge visible, no pipes) - Bottom-...
+da57c46b599618ae	train	label A: read-back states (1, 2, 3, 4): s and the way the objects align, the correct assembly is: - Top-left: Part 1 - Top-right: Part 2 - Bottom-left: Part 3 - Bottom-right: Part 4 This corresponds to statemen
+025e682106db7fab	train	label D: plan states (1, 2, 3, 4): art 3. Putting this together, the correct arrangement is: - Top-left: Part 1 - Top-right: Part 2 - Bottom-left: Part 3 - Bottom-right: Part 4 However, let's check the opt | read-back states (1, 2, 3, 4): The only option that matches this arrangement is: (D) The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 3; and the bottom-right part shoul...
+f945c8cd70f5aef6	train	label D: plan states (1, 2, 3, 4): of the wall/floor from Part 1. Testing this arrangement: - Top-left: Part 1 (plant and table surface) - Top-right: Part 2 (door) - Bottom-left: Part 3 (empty floor) - Bottom-right: Part 4 (table le...
+54e447ac389effda	train	label B: read-back states (1, 2, 3, 4): Therefore, the correct placement is: - Top-left: Piece 1 (lamp and pillow) - Top-right: Piece 2 (window) - Bottom-left: Piece 3 (corner of bed) - Bottom-right: Piece 4 (top of bed with pillows) Th
+d200eed22c418c71	train	label C: read-back states (3, 2, 1, 4): Given this analysis: - Top-left: Part 3 (corner of the wall) - Top-right: Part 2 (bed edge and wall) - Bottom-left: Part 1 (door, TV, bed) - Bottom-right: Part 4 (white floor and TV base) Th
+2c650ef37bb9e1fb	train	label C: read-back states (3, 4, 1, 2): Therefore, the correct assembly is: Top-left: Piece 3 Top-right: Piece 4 Bottom-left: Piece 1 Bottom-right: Piece 2 Comparing this order to the
+6bb226b69783f6ba	train	label D: read-back states (1, 2, 3, 4): the puzzle pieces have been arranged as previously reasoned—Part 1 in the top-left, Part 2 in the top-right, Part 4 in the bottom-right, and Part 3 in the bottom-left—I can thoroughly examine the
+5b3870ef4d08885f	train	label C: read-back states (1, 2, 3, 4): Thus, the correct arrangement is: - Top-left: Piece 1 - Top-right: Piece 2 - Bottom-left: Piece 3 - Bottom-right: Piece 4 This matches statement (C):
+9467780fbcea4956	train	label C: read-back states (2, 1, 3, 4): Therefore, the correct arrangement is: Top-left = Piece 2 Top-right = Piece 1 Bottom-left = Piece 3 Bottom-right = Piece 4 Checking the answer options,
+4f71b63d7e498951	train	label D: read-back states (1, 2, 3, 4): correct arrangement to reconstruct the original image is: - Top-left: Part 1 - Top-right: Part 2 - Bottom-left: Part 3 - Bottom-right: Part 4 This corresponds to statement | read-back states (1, 2, 3, 4): Bottom-right: Part 4 This corresponds to statement (D): The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 3; and the bottom-right part sho...
+b166aba4d6ec0626	train	label A: read-back states (1, 2, 3, 4): correct assembly reconstructs the natural scene as follows: Top-left: Part 1, Top-right: Part 2, Bottom-left: Part 3, Bottom-right: Part 4, which matches statement (A).
+8da40af9e765f58d	train	label D: read-back states (2, 1, 4, 3): ged. Restating the answer in the specified format: (D) The top-left part should be Part 2; the top-right part should be Part 1; the bottom-left part should be Part 4; and the bottom-right part shou...
+aa0007683c2bfd50	train	label C: read-back states (1, 2, 3, 4): ction. Based on these observations, the answer is: (C) The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 3; and the bottom-right part shou...
+c906c78ba8e22eea	train	label B: read-back states (1, 4, 3, 2): Therefore, the correct arrangement is: - Top-left: Part 1 - Top-right: Part 4 - Bottom-left: Part 3 - Bottom-right: Part 2 This matches statement (B): | read-back states (1, 4, 3, 2): t 3 - Bottom-right: Part 2 This matches statement (B): The top-left part should be Part 1; the top-right part should be Part 4; the bottom-left part should be Part 3; and the bottom-right part shou...
+817a46129562b094	train	label D: read-back states (1, 2, 3, 4): the most coherent and visually natural arrangement is: The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 3; and the bottom-right part shou...
+1cff3890b7651cd5	train	label C: read-back states (1, 2, 3, 4): Arranging them as such: - Top-left: Part 1 (window, countertop, fridge) - Top-right: Part 2 (countertop continuation) - Bottom-left: Part 3 (fridge base, shelving unit, table edge, bag) - Bottom-ri...
+8df62bedf96e0c99	train	label D: (A) quoted with the text of (D): only possible correct arrangement is: (A) The top-left part should be Part 2; the top-right part should be Part 3; the bottom-left part should be Part 4; and the
+0327ed9c7d26d50a	train	label A: plan states (1, 2, 3, 4): ompleting the background. Thus, the correct arrangement is: Top-left: Part 1; Top-right: Part 2; Bottom-left: Part 3; Bottom-right: Part 4. Let's check this: - If Part | names (B): This corresponds to statement (B). However, the question requests the arr
+71622d84ed9c13a0	train	label D: read-back states (2, 1, 3, 4): sitions. Based on this verification, the answer is (D): The top-left part should be Part 2; the top-right part should be Part 1; the bottom-left part should be Part 3; and the bottom-right part sho...
+98a576cde61f024e	train	label A: read-back states (1, 2, 4, 3): Therefore, the correct assembly is: Top-left: Part 1 Top-right: Part 2 Bottom-left: Part 4 Bottom-right: Part 3 In terms of the given statem | read-back states (1, 2, 4, 3): terms of the given statements, this matches option (A): The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 4; and the bottom-right part sho...
+775b42a13ffe2cf2	train	label A: read-back states (1, 2, 3, 4): ts and walls all point towards the following arrangement: - Top-left: Piece 1 - Top-right: Piece 2 - Bottom-left: Piece 3 - Bottom-right: Piece 4 This matches statement (A): | read-back states (1, 2, 3, 4): assembled the four puzzle pieces according to my reasoning: Part 1 in the top-left, Part 2 in the top-right, Part 3 in the bottom-left, and Part 4 in the bottom-right. Now, I will carefully verify
+f398752d4d339f2b	train	label D: read-back states (1, 2, 3, 4): Therefore, the correct assembly is: Top-left: Part 1 Top-right: Part 2 Bottom-left: Part 3 Bottom-right: Part 4 This matches option (D): The
+7fc32d23a5d1c862	train	label A: plan states (1, 2, 3, 4): By matching these visual cues, the correct arrangement is: top-left = Part 1, top-right = Part 2, bottom-left = Part 3, bottom-right = Part 4. | read-back states (1, 2, 3, 4): t the arrangement is correct. Thus, the answer is (A): The top-left part should be Part 1; the top-right part should be Part 2; the bottom-left part should be Part 3; and the bottom-right part shou...
+c5080e4fe781a19c	train	label B: read-back states (2, 1, 4, 3): With this analysis, the arrangement should be: - Top-left: Piece 2 - Top-right: Piece 1 - Bottom-left: Piece 4 - Bottom-right: Piece 3 However, reviewing the answe
+a4e86525b8a6034c	train	label A: plan states (1, 2, 3, 4): the right of Piece 3 (bottom row, right). Assembling them: Top-left: Piece 1 (table and left wall) Top-right: Piece 2 (floor and chair legs) Bottom-left: Piece 3 (bunk bed and left wall) Bottom-rig...
+4c85e6584c4872d8	train	label D: read-back states (1, 2, 3): Therefore, the logical arrangement is: Top: Piece 1 (wall, TV, lamp, table) Middle: Piece 2 (floor) Bottom: Piece 3 (couch, lamp base, furniture
+ce815564e1405dea	train	label A: plan states (1, 2, 3): re, the correct order to reconstruct the original image is: Top – Part 1, Middle – Part 2, Bottom – Part 3. | names (B): This corresponds to option (B). However, upon further inspection, I re
+62e23a8deda77bf1	train	label D: read-back states (3, 2, 1): Therefore, the correct assembly is: - Top: Piece 3 - Middle: Piece 2 - Bottom: Piece 1 This matches statement (D): | read-back states (3, 2, 1): led image created according to my previous reasoning, where Part 3 is at the top, Part 2 is in the middle, and Part 1 is at the bottom. My goal is to check for visu
+b5a24bd76f831ba8	train	label A: plan states (1, 2, 3): is absent in Part 3, suggesting the order should begin with Part 1 at the top. - The table's edge appears in Part 2 and continues into Part 3, indicating Part 2 should be in the middle and Part 3 a...
+8b8ab86926f585f2	train	label A: plan states (1, 2, 3): all the object continuity, perspective, and boundaries is: Top = Part 1, Middle = Part 2, Bottom = Part 3.
+de19e74508eb5444	train	label A: read-back states (1, 2, 3): orrectness of the arrangement. Thus, the answer is: (A) The top part should be Part 1; the middle part should be Part 2; and the bottom part should be Part 3.
+610faf4db230fc46	train	label A: plan states (1, 2, 3): refore, the most compelling and seamless reconstruction is: Part 1 at the top, Part 2 in the middle, and Part 3 at the bottom. This preserves object contin
+c5b3350c0ffd0af5	train	label C: plan states (1, 2, 3): een Parts 1 and 3. Thus, the natural arrangement should be: Top: Part 1, Middle: Part 2, Bottom: Part 3.
+cb155ee41193fac6	train	label A: read-back states (1, 2, 3): Therefore, the correct assembly is: Top: Part 1 Middle: Part 2 Bottom: Part 3 Among the choices, this matc | read-back states (1, 2, 3): of my previous analysis. Therefore, the answer is (A): The top part should be Part 1; the middle part should be Part 2; and the bottom part should be Part 3.
+9d2d8cda26c78e69	train	label A: read-back states (1, 2, 3): This matches statement (A): The top part should be Part 1; the middle part should be Part 2; and the bottom part should be Part 3. Having assembled the image u | read-back states (1, 2, 3): ement is coherent and correct. Thus, the answer is (A): The top part should be Part 1; the middle part should be Part 2; and the bottom part should be Part 3.
+a3a7c6852c988f89	train	label D: read-back states (3, 2, 1): Therefore, the correct arrangement is: Top: Part 3 Middle: Part 2 Bottom: Part 1 This corresponds to statemen | read-back states (3, 2, 1): assembled the image using the sequence I determined—placing Part 3 at the top, Part 2 in the middle, and Part 1 at the bottom—I can evaluate the visual coh
+89b812cd00e60134	train	label D: read-back states (3, 2, 1): and naturalness of the arrangement. The answer is: (D) The top part should be Part 3; the middle part should be Part 2; and the bottom part should be Part 1.
+4c069e8a1183d5cf	train	label A: plan states (1, 2, 3): ral boundaries. Therefore, the correct arrangement is: the top part should be Part 1; the middle part should be Part 2; and the bottom part should be Part 3. | read-back states (1, 2, 3): iscontinuity or visual error. Thus, the answer is: (A) The top part should be Part 1; the middle part should be Part 2; and the bottom part should be Part 3.
+7ad50a9ccefbf777	train	label A: plan states (3, 2, 1): cts at the top. Therefore, the correct arrangement is: the top part should be Part 3, the middle part should be Part 2, and the bottom part should be Part 1. | read-back states (3, 2, 1): at I have assembled the image in the proposed order—placing Part 3 at the top, Part 2 in the middle, and Part 1 at the bottom—I can verify the coherence of
+e074a2e41b4d900c	train	label A: plan states (1, 2, 3): t that reconstructs the original scene naturally is placing Part 1 at the top, Part 2 in the middle, and Part 3 at the bottom. | read-back states (1, 2, 3): After assembling the image by placing Part 1 at the top, Part 2 in the middle, and Part 3 at the bottom, I carefully examined the vis
+91b39bca18883802	train	label C: plan states (1, 2, 3): beled parts. Looking at the labels, the correct order is: - Top: Part 1 - Middle: Part 2 - Bottom: Part 3 But, the answer choices sugg
+6ed2d5f38961ddb5	train	label D: read-back states (1, 2, 3): Therefore, the correct assembly is: - Top: Part 1 - Middle: Part 2 - Bottom: Part 3 Among the given options, thi | read-back states (1, 2, 3): g the given options, this corresponds to statement (D): The top part should be Part 1; the middle part should be Part 2; and the bottom part should be Part 3. To verify the correctness of
+4869f3448140a813	validation	label C: read-back states (1, 2, 3): iven these observations, the most natural arrangement is: - Top: Part 1 (upper wall, TV) - Middle: Part 2 (carpet/floor) - Bottom: Part 3 (bed, chair, dresser, bear) | read-back states (1, 2, 3): he jigsaw pieces in the order determined previously—placing Part 1 at the top, Part 2 in the middle, and Part 3 at the bottom—I closely examine the resulti
+8b13feeb1d334247	validation	label A: read-back states (1, 2, 3): een wall, furniture, and floor, the correct arrangement is: Top: Part 1 Middle: Part 2 Bottom: Part 3 This corresponds to statemen | names (D): or, the correct arrangement is: Top: Part 1 Middle: Part 2 Bottom: Part 3 This corresponds to statement (D). However, looking at the options, state

reports/known_issues/validation_target_near_copy_of_train.tsv

@@ -0,0 +1,2 @@
+# This validation row's target is a near-copy of a training row's target (the same scene shot from a slightly shifted viewpoint), so it is not fully held out.
+96fb0688103dd21d	validation	1 bits from train bdcb20276b29b58b (ade20k_00009-of-00010_000212_ADE_train_00018569)

reports/s13_dropped.jsonl

@@ -0,0 +1,30 @@
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "81277c04c5ba9850", "split": "val", "reason": "prior_downsample", "step": "answer", "answer": "part 1 should be to the right of part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "84a4deb80fc1d17a", "split": "val", "reason": "prior_downsample", "step": "answer", "answer": "part 1 should be to the right of part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "edc0e055f19974d6", "split": "val", "reason": "prior_downsample", "step": "answer", "answer": "part 1 should be to the right of part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "ef0a273cae79305d", "split": "val", "reason": "prior_downsample", "step": "answer", "answer": "part 1 should be to the right of part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "f74801245842146b", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "part 1 should be to the bottom of part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "5687e8513c6c8770", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "part 1 should be to the bottom of part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "ce0e0c25e15e4d7f", "split": "val", "reason": "prior_downsample", "step": "answer", "answer": "part 1 should be to the top of part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "38382176b5d44614", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "part 1 should be to the top of part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "e9fe11fb738bfdad", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "part 1 should be to the bottom of part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "fa1915f2e234f996", "split": "val", "reason": "prior_downsample", "step": "answer", "answer": "part 1 should be to the top of part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "33a2d2411f016635", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "part 2 should be to the bottom of part 1"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "2ea4b696a9ab0d8f", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "part 1 should be to the top of part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "c32bfaf7bc983e94", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "part 1 should be to the top of part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "810a7ddedde8452f", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "part 2 should be to the top of part 1"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "d72554b42a702e51", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top-left part should be part 3; the top-right part should be part 4; the bottom-left part should be part 2; and the bottom-right part should be part 1"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "2e1c0fd58524e0bf", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top-left part should be part 4; the top-right part should be part 1; the bottom-left part should be part 3; and the bottom-right part should be part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "3aec12ee948f920f", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top-left part should be part 2; the top-right part should be part 1; the bottom-left part should be part 4; and the bottom-right part should be part 3"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "679a3087512476cb", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top-left part should be part 1; the top-right part should be part 4; the bottom-left part should be part 3; and the bottom-right part should be part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "1135de1dc3cdff30", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top-left part should be part 4; the top-right part should be part 2; the bottom-left part should be part 1; and the bottom-right part should be part 3"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "daa1d63ab48afece", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top-left part should be part 3; the top-right part should be part 1; the bottom-left part should be part 2; and the bottom-right part should be part 4"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "671a0a90f8225a08", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top part should be part 2; the middle part should be part 1; and the bottom part should be part 3"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "f66db7cf3836dc88", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top part should be part 3; the middle part should be part 2; and the bottom part should be part 1"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "e6ca437b40d6a1e6", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top part should be part 2; the middle part should be part 3; and the bottom part should be part 1"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "e07ce5754c61b5d3", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top part should be part 3; the middle part should be part 1; and the bottom part should be part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "51edaa501b6f705b", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top part should be part 1; the middle part should be part 3; and the bottom part should be part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "bf84d5bf29aae1db", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top part should be part 1; the middle part should be part 2; and the bottom part should be part 3"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "549c16058b2482df", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top part should be part 2; the middle part should be part 3; and the bottom part should be part 1"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "925eeff0cfd4cdca", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top part should be part 2; the middle part should be part 1; and the bottom part should be part 3"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "858a1261e2e6b4f8", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top part should be part 1; the middle part should be part 3; and the bottom part should be part 2"}
+{"build": "thinkmorph_jigsaw_r2", "row_uid": "f51e1a8d2b5bdd1b", "split": "val", "reason": "prior_downsample", "step": "letter", "answer": "the top part should be part 2; the middle part should be part 1; and the bottom part should be part 3"}

reports/s14_sample.tsv

@@ -0,0 +1,201 @@
+# row_uid	task	split -- the S14 spot-check rows; also flagged S14.sampled_qa
+9e153466d27880b9	jigsaw_2x1	train
+72b13f21b8b2fc18	jigsaw_3x1	train
+4494cb910ae7a804	jigsaw_1x2	train
+d249734aacd5b098	jigsaw_1x3	train
+8df62bedf96e0c99	jigsaw_2x2	train
+1e1dede80a95ae80	jigsaw_3x1	train
+4f1730b658516197	jigsaw_1x3	train
+cf46e240bc5bef85	jigsaw_2x1	train
+1a6757e762441e00	jigsaw_3x1	train
+c7b7d66f26f53216	jigsaw_2x1	train
+d27411d3caa85a12	jigsaw_2x2	train
+e389b44eb6068684	jigsaw_2x1	train
+92c20f4c4c4c234f	jigsaw_1x3	train
+7d8daa66de715057	jigsaw_3x1	train
+0ccba55cb3fa13c8	jigsaw_2x1	train
+a7510d6a78f0223d	jigsaw_1x3	train
+2a531010b4ac8dd5	jigsaw_2x2	train
+7fcf33ec36cb8d24	jigsaw_1x2	train
+c4dd31af031744be	jigsaw_2x1	train
+0f57f77ef03a36ec	jigsaw_1x3	train
+e30718984fa4a37d	jigsaw_1x3	train
+4df936cabf4ef41e	jigsaw_1x3	train
+580f601e6b846bed	jigsaw_1x2	validation
+1565110529c99488	jigsaw_1x3	train
+9db13d99ebf5c2e3	jigsaw_3x1	train
+21c9afdd85e56db7	jigsaw_3x1	train
+4b9f440a7806360d	jigsaw_3x1	train
+53f7b21459727c62	jigsaw_2x1	train
+8867314a659debcd	jigsaw_1x3	train
+52703e0a16a5b3e1	jigsaw_3x1	train
+212a96b33840c40b	jigsaw_2x1	train
+4a8ee937e925ca78	jigsaw_1x3	train
+9df5732860e455fc	jigsaw_3x1	train
+974e6686c82491fd	jigsaw_1x2	train
+1e0d98134175d2b8	jigsaw_2x1	train
+826abd171d3b152a	jigsaw_2x2	train
+7b3894151856846e	jigsaw_1x3	train
+28829cffed4edd31	jigsaw_1x2	train
+97a11215b731f09b	jigsaw_2x1	train
+3218d91a3fbf6090	jigsaw_2x2	train
+ab0a7b3444c83758	jigsaw_2x1	train
+4b43d52778f4670a	jigsaw_2x2	train
+a86d9bd155aba81c	jigsaw_2x1	train
+2c1fd5f1d2ebe514	jigsaw_3x1	train
+cea2379c8ab1080c	jigsaw_2x2	train
+fc07493556bec635	jigsaw_1x2	train
+43bfc7f7bd5393ad	jigsaw_1x3	train
+1327a2ca50386b94	jigsaw_3x1	train
+2d679ecd355bd512	jigsaw_1x2	train
+a9399498b9984e5d	jigsaw_1x2	train
+45e417182c1d5f39	jigsaw_2x2	train
+22a3929aa2745ad2	jigsaw_2x2	train
+891dd83ae3eb1f7d	jigsaw_3x1	train
+e576a2700ddf0366	jigsaw_3x1	train
+f4c1a0586c82a89f	jigsaw_2x1	train
+ff323a73beb668bf	jigsaw_1x3	train
+174e0898d3fa50de	jigsaw_2x2	train
+dcee652a341750bb	jigsaw_3x1	train
+c625f29961e9d09d	jigsaw_2x1	train
+3482c78d81a1f33b	jigsaw_1x3	train
+f424b0e0cb2e7781	jigsaw_1x3	train
+8c7ad341aeb14efd	jigsaw_1x3	train
+6e85507455b08fed	jigsaw_3x1	train
+d48afde923c3d0dd	jigsaw_2x1	train
+bde0faab1f3780e5	jigsaw_2x1	train
+36977369bfbdef91	jigsaw_1x2	validation
+e10c449c852522f1	jigsaw_1x2	train
+dc9f442b62539da1	jigsaw_2x1	train
+ef67e864c2ab2004	jigsaw_1x3	train
+4edb16bf9e34fb90	jigsaw_2x1	train
+79159d17acefe237	jigsaw_1x2	train
+4c89136989510af2	jigsaw_1x3	train
+2cba1b63944b7a15	jigsaw_2x2	train
+35ac17d6a5c10b6f	jigsaw_2x2	train
+7aa3ab3537a5010c	jigsaw_2x2	train
+82f717650ad49eb2	jigsaw_2x1	train
+28e524e88ac057f2	jigsaw_2x1	train
+1d9d56230d8b9f0d	jigsaw_1x2	train
+7a91914d9b6df36d	jigsaw_2x1	train
+4e63f5f14069a284	jigsaw_2x2	validation
+bfd80e590159fbbc	jigsaw_1x3	train
+2aa97aa22607d9bf	jigsaw_3x1	train
+55837c889e46dc4b	jigsaw_1x3	train
+7d7ce5dffdc12976	jigsaw_2x2	train
+a6615638b2aff613	jigsaw_2x2	train
+ef9405d1c19621a9	jigsaw_2x1	train
+8a80eaa37e4d5da1	jigsaw_1x3	train
+ebc01f3faf1974d9	jigsaw_2x2	train
+f3f8d63313965b18	jigsaw_2x2	train
+c74de36dc572feaf	jigsaw_2x2	train
+131fba405c871fb0	jigsaw_1x2	train
+436aee1572ead478	jigsaw_2x2	train
+c92ebea48c267ddb	jigsaw_3x1	train
+f972150703aa1227	jigsaw_2x2	train
+0b449b650d7c71b0	jigsaw_2x1	train
+4b714f69e5185e75	jigsaw_1x2	train
+aebc480512e2792f	jigsaw_2x1	train
+91830425c635239f	jigsaw_2x1	train
+1f13ad6ed9c5b0ae	jigsaw_2x1	train
+80aa32ebc466f367	jigsaw_2x1	train
+b4a3fa6ce8982436	jigsaw_2x2	train
+69e289ed245c3de8	jigsaw_1x3	train
+48897ff509ab58b5	jigsaw_1x3	train
+9059c0e55ed7a170	jigsaw_2x1	train
+adea6a52540a0c77	jigsaw_3x1	train
+83ac43efc782279a	jigsaw_2x1	train
+f39e4994ab5d0cb6	jigsaw_2x1	train
+325fdde41bcebb79	jigsaw_1x2	train
+6cb1dbc5e49eeaa7	jigsaw_2x1	train
+e784e05eae620dbe	jigsaw_3x1	train
+e26c5a80bbd62584	jigsaw_1x3	train
+8d9c87196062b017	jigsaw_2x2	train
+f945c8cd70f5aef6	jigsaw_2x2	train
+43eb44a74b838857	jigsaw_1x3	train
+ab25fce66b5300c5	jigsaw_2x1	train
+12ac5d210785fda6	jigsaw_2x1	train
+40383ab396f20a8e	jigsaw_3x1	train
+6ceeeedfebb7f531	jigsaw_2x1	train
+0bbc2fb19e2aa7ae	jigsaw_1x2	train
+db943f79a47cfbe4	jigsaw_1x3	train
+efa805b4e7a9a1d2	jigsaw_3x1	train
+259a1ccd56cc22c2	jigsaw_1x3	train
+120e6f426236a61d	jigsaw_1x3	train
+194a9ccfd886828b	jigsaw_1x2	train
+d32918c2826322db	jigsaw_2x1	train
+40225bef3ddd94e8	jigsaw_2x1	train
+b8903f98eb0aada0	jigsaw_2x2	train
+d93eccfd2dfb81a3	jigsaw_3x1	train
+4a1e9b3bd5b0adec	jigsaw_2x2	train
+6afc41a7bc66f5e4	jigsaw_1x3	train
+cefdd93357145599	jigsaw_1x3	train
+55d2668e91c47d91	jigsaw_2x1	train
+8c075397cc300cc1	jigsaw_3x1	train
+1b8d8b4a5d33b79e	jigsaw_2x2	train
+a9f06dc3a55331fa	jigsaw_2x2	train
+0bf696621e8d94e3	jigsaw_1x2	train
+aeb5d0a0d9fc3bd2	jigsaw_2x2	train
+d6fb256ea6256761	jigsaw_1x2	train
+b9b31cee62b83c1a	jigsaw_3x1	train
+fdd141cd474f9ee5	jigsaw_3x1	train
+2f2d85f6f7c95584	jigsaw_1x2	train
+7580b578db61d577	jigsaw_2x2	train
+8de6199bc65d1350	jigsaw_1x3	train
+f767629d45ba86cf	jigsaw_3x1	train
+debcbc78f931823b	jigsaw_1x3	train
+395e1aa2bc4817a4	jigsaw_1x2	train
+c0a4a3f08d367f2a	jigsaw_1x2	train
+27da67e67eb02e2f	jigsaw_2x2	train
+11e9f948bd96bc73	jigsaw_1x3	train
+3265e482ff9c92d1	jigsaw_1x3	train
+485b0c5fbc8360f9	jigsaw_2x2	train
+d1d69a2482f20d3f	jigsaw_2x2	train
+c0cd0d3151be0007	jigsaw_1x3	train
+92c52e76d3d4e3e9	jigsaw_2x1	train
+919ae328eebdb90d	jigsaw_1x3	train
+54bf0bda51fb9579	jigsaw_1x3	train
+b5fd9920da0f5b57	jigsaw_3x1	train
+98fafabbb8f4b04f	jigsaw_1x2	train
+c5080e4fe781a19c	jigsaw_2x2	train
+3aca848e8c7a074d	jigsaw_1x2	train
+940c12e1e9e1eff9	jigsaw_1x2	train
+f31e1256dd655a16	jigsaw_1x2	train
+787944f935a659a4	jigsaw_1x3	train
+6e159531837da8f2	jigsaw_3x1	train
+a7e523a05139938b	jigsaw_2x2	train
+d1404308e9ae3027	jigsaw_3x1	train
+c9184404a34f828f	jigsaw_3x1	train
+73c55c03f7ed022b	jigsaw_1x2	train
+29923d12a7f5ffd8	jigsaw_1x2	train
+f6d1258fb024a722	jigsaw_3x1	train
+b8dc2079397a7f6a	jigsaw_2x2	train
+6d22da7aa1fccba2	jigsaw_3x1	train
+963aa66608c4fd36	jigsaw_2x2	train
+99119336d24db7aa	jigsaw_1x2	train
+912643959ed71177	jigsaw_2x2	train
+13d40ab2ea4488e8	jigsaw_2x2	train
+17a71fa7add13967	jigsaw_3x1	train
+f32d61b3e79f0e67	jigsaw_2x2	train
+3ba6173a30991f33	jigsaw_1x3	train
+05f2e05bcdfd7320	jigsaw_1x3	train
+3e92c32f7efe1f74	jigsaw_1x2	train
+91648f92bdd795bd	jigsaw_2x2	train
+84c52ef41ec50bf8	jigsaw_3x1	train
+95ec2eac74277579	jigsaw_1x3	train
+2b96cdbf8c5445ca	jigsaw_1x2	train
+c812b48bca7d1880	jigsaw_3x1	train
+5329c75ca6ff473f	jigsaw_3x1	train
+296f33b6823f7470	jigsaw_3x1	train
+c0d496823c8d8384	jigsaw_3x1	train
+d35f9e1deb1002cb	jigsaw_3x1	train
+a3a7c6852c988f89	jigsaw_3x1	train
+c45a55192cd17dbd	jigsaw_1x2	train
+c77ac1a6d8279288	jigsaw_3x1	train
+bb72c118a24fb935	jigsaw_1x2	train
+dcc81c7d333a0bb4	jigsaw_1x2	train
+976a46f1d939aa3a	jigsaw_1x2	train
+7e43569f29399c15	jigsaw_1x2	train
+d5a761cbaa5c0d17	jigsaw_1x2	train
+d00fe744ab4226c0	jigsaw_1x2	train
+3ba3ba012cffe41b	jigsaw_1x2	validation

train/chunk_00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa7c3a2602365d8316f6965795f6d57a8d291b39bb94aeb2173bcb1b971dd9e8
+size 711937773

validation/chunk_00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:275274b4505a2189114a1b17f52d9d3ca737f849d4e22d70f723a05c66f161a8
+size 18170599