Datasets:
Download reports/subset_dropped.jsonl from yrlyrl/spatial-mmcot-thinkmorph_jigsaw: direct link, hf CLI and curl.
- Browser
- Download file 1.41 kB
-
https://huggingface.co/datasets/yrlyrl/spatial-mmcot-thinkmorph_jigsaw/resolve/main/reports/subset_dropped.jsonl
- Command line
-
hf download hf://datasets/yrlyrl/spatial-mmcot-thinkmorph_jigsaw/reports/subset_dropped.jsonl
-
curl -L -o subset_dropped.jsonl https://huggingface.co/datasets/yrlyrl/spatial-mmcot-thinkmorph_jigsaw/resolve/main/reports/subset_dropped.jsonl
1.41 kB
| {"row_uid": "fa6dcfb6faaa8f90", "split": "train", "task": "jigsaw_2x2", "reason": "label_doubt", "note": "The prompt ties label 1 to the top-left display cell, but the parts are shuffled and their printed numbers are not legible in the input, so I matched by content; if the labels followed display reading order the correct arrangement would be the unlisted identity."} | |
| {"row_uid": "a97e3d46ed7b37ac", "split": "train", "task": "jigsaw_3x1", "reason": "label_doubt", "note": "The reassembled view supports option C (1 top, 2 middle, 3 bottom), not the indicated option D."} | |
| {"row_uid": "f7f1e98a44a999c5", "split": "train", "task": "jigsaw_3x1", "reason": "label_doubt", "note": "In the input the gap between the facade and the street is not a clear white strip, so that part of the three-way split is inferred from the question's framing rather than a visible boundary."} | |
| {"row_uid": "8bc42661da439025", "split": "validation", "task": "jigsaw_1x3", "reason": "label_doubt", "note": "The reassembled view locks the parts as 1-2-3 (option C), with the white bed corner at the right edge, not the 3-1-2 order (option B) named as correct."} | |
| {"row_uid": "d7e81d80ae569e0b", "split": "validation", "task": "jigsaw_2x2", "reason": "label_doubt", "note": "The reassembled image visually places Part 3 on the bottom-left and Part 4 on the bottom-right, but Option (A) lists them as Part 4 and Part 3 respectively."} | |