Datasets:
Download exclusions.json from hotchpotch/s1mb-dataset: direct link, hf CLI and curl.
- Browser
- Download file 1.04 kB
-
https://huggingface.co/datasets/hotchpotch/s1mb-dataset/resolve/5b26cb8e8cf62ace3daf24eb8299face735b6e13/exclusions.json
- Command line
-
hf download hf://datasets/hotchpotch/s1mb-dataset@5b26cb8e8cf62ace3daf24eb8299face735b6e13/exclusions.json
-
curl -L -o exclusions.json https://huggingface.co/datasets/hotchpotch/s1mb-dataset/resolve/5b26cb8e8cf62ace3daf24eb8299face735b6e13/exclusions.json
1.04 kB
| { | |
| "version": "s1mb.open_jev.curated.v2.20260928", | |
| "scope": "Explicitly excluded from packaged cases and active evaluation; not quarantine or missing data.", | |
| "excluded_cases": 138, | |
| "tasks": { | |
| "open_jev__tile_platformer-v1:choice": "The one-hot BFS teacher rejects equally shortest successful actions: 74/100 audited cases have multiple optimal first actions. Neither the shortest-path objective nor the teacher tie order is fully specified in the question.", | |
| "open_jev__vizdoom-basic-v1:choice": "The requested momentum-aware script policy uses a velocity multiplier of 8 and a lateral-error deadband of 10 that are not disclosed in the inference input. Labels measure a hidden weak-script convention, not a uniquely specified action.", | |
| "open_jev__trex_runner-v1:choice": "Only four original test cases exist. One answer changes accuracy by 25 percentage points, while equal benchmark weighting gives this unstable estimate the same weight as larger tasks." | |
| }, | |
| "retained_vizdoom_tasks": [ | |
| "noul", | |
| "score" | |
| ] | |
| } | |