| { |
| "source": "messytable", |
| "build": "messytable_r2", |
| "export_git_sha": "4aaee1f4e9469da41810aac049fa411c1834b809", |
| "export_git_dirty": false, |
| "limit": 0, |
| "converter_git_sha": "949af62f8ab701c58da5cdeed32946242c571c83", |
| "converter_git_dirty": false, |
| "build_counts": { |
| "raw_rows_read": 1880, |
| "raw_unusable": 0, |
| "rows_out": 1819 |
| }, |
| "build_manifest_sha1": "86277d88c96db3d53023c731ab75dc979bca4385", |
| "s10_keep_sha1": "9fd27b7ff9e2be8d9e0466795f251875210de152", |
| "s12_assignments_sha1": "d0195205cd5852e08769b99f9c72740f8ce9fcaa", |
| "s13_balanced_keep_sha1": "af73ed4ef438de882feeee30de4a25701f2a3f6a", |
| "license": "undeclared", |
| "license_commercial": false, |
| "upstream_repo": "leo66666/messytable", |
| "donor": null, |
| "visual_options": null, |
| "num_inputs": [ |
| 2, |
| 7 |
| ], |
| "max_k": 1, |
| "preview_max_k": 1, |
| "rows": { |
| "train": 1490, |
| "validation": 27 |
| }, |
| "targets": { |
| "train": 1490, |
| "validation": 27 |
| }, |
| "distinct_targets_total": 1517, |
| "num_input_images": { |
| "train": [ |
| 2, |
| 7 |
| ], |
| "validation": [ |
| 2, |
| 3 |
| ] |
| }, |
| "num_steps": { |
| "train": [ |
| 1, |
| 1 |
| ], |
| "validation": [ |
| 1, |
| 1 |
| ] |
| }, |
| "distinct_targets": { |
| "train": 1490, |
| "validation": 27 |
| }, |
| "supervision_kind": { |
| "train": { |
| "full_interleaved": 1490 |
| }, |
| "validation": { |
| "full_interleaved": 27 |
| } |
| }, |
| "source_scene_corpus": { |
| "train": { |
| "messytable": 1490 |
| }, |
| "validation": { |
| "messytable": 27 |
| } |
| }, |
| "filter_flags": { |
| "train": { |
| "S5.replay_unsupported": 1490, |
| "S14.sampled_qa": 197 |
| }, |
| "validation": { |
| "S5.replay_unsupported": 27, |
| "S14.sampled_qa": 3 |
| } |
| }, |
| "answer_type": { |
| "train": { |
| "integer": 1490 |
| }, |
| "validation": { |
| "integer": 27 |
| } |
| }, |
| "answer_tag": { |
| "train": { |
| "integer": { |
| "rows": 1490, |
| "equals_answer_value": 1490 |
| } |
| }, |
| "validation": { |
| "integer": { |
| "rows": 27, |
| "equals_answer_value": 27 |
| } |
| } |
| }, |
| "options_text_rows": { |
| "train": 0, |
| "validation": 0 |
| }, |
| "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": 1490 |
| }, |
| "validation": { |
| "undeclared": 27 |
| } |
| }, |
| "removed": { |
| "build/dropped.jsonl": {}, |
| "build/quarantine.jsonl": { |
| "S4c": { |
| "S4c.cot_label_conflict": 61 |
| } |
| }, |
| "s10_dropped.jsonl": {}, |
| "s10_label_conflicts.jsonl": {}, |
| "s13_dropped.jsonl": { |
| "S13": { |
| "S13.answer": 302 |
| } |
| } |
| }, |
| "split": { |
| "policy": "scene_id", |
| "rows": 1819, |
| "keys": 191, |
| "val_rows": 27, |
| "val_keys": 3, |
| "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": "be1d3dc74b8a4971dcf142e460fba3e32d6375c5", |
| "validation": "6070fad95fff608a1877f80e70583c51b386bce9" |
| }, |
| "tasks": { |
| "multiview_counting": { |
| "train": 1490, |
| "validation": 27 |
| } |
| }, |
| "answer_counts": { |
| "multiview_counting/train": { |
| "2": 447, |
| "4": 362, |
| "3": 347, |
| "1": 184, |
| "5": 119, |
| "6": 25, |
| "7": 4, |
| "8": 2 |
| }, |
| "multiview_counting/validation": { |
| "4": 9, |
| "2": 8, |
| "3": 5, |
| "1": 4, |
| "5": 1 |
| } |
| }, |
| "majority_class_rate_train": { |
| "multiview_counting": 0.3 |
| }, |
| "s14_sample": { |
| "rows": 200, |
| "target": 200, |
| "replay_unsupported": true |
| }, |
| "s13_groups": { |
| "messytable_r2/multiview_counting/train": { |
| "rule": "cap30[canon]", |
| "rows_in": 1792, |
| "rows_out": 1490, |
| "triggered": true, |
| "exceeds_target": false, |
| "level": 447, |
| "target": 0.3, |
| "k_real": 5, |
| "distinct": 8, |
| "majority_share_in": 0.418, |
| "majority_share_out": 0.3, |
| "majority_share_final": 0.3 |
| }, |
| "messytable_r2/multiview_counting/val": { |
| "rule": "cap30[canon]", |
| "rows_in": 27, |
| "rows_out": 27, |
| "triggered": false, |
| "exceeds_target": false, |
| "level": 9, |
| "target": 0.3333333333333333, |
| "k_real": 3, |
| "distinct": 5, |
| "majority_share_in": 0.3333, |
| "majority_share_out": 0.3333, |
| "majority_share_final": 0.3333 |
| } |
| }, |
| "blind_probe": { |
| "messytable_r2/multiview_counting": { |
| "probed": true, |
| "majority": 0.2982, |
| "keywords": 0.2982, |
| "template": 0.4213, |
| "chance": null, |
| "reference": "majority", |
| "best": "template", |
| "flag": false, |
| "eval_rows": 731, |
| "sigma3": 0.0555, |
| "details": {} |
| } |
| }, |
| "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 |
| } |
| } |
