Datasets:
|
Download release/README.md from UnpluggedAI/posa-bench: direct link, hf CLI and curl.
- Browser
- Download file 990 Bytes
-
https://huggingface.co/datasets/UnpluggedAI/posa-bench/resolve/main/release/README.md
- Command line
-
hf download hf://datasets/UnpluggedAI/posa-bench/release/README.md
-
curl -L -o README.md https://huggingface.co/datasets/UnpluggedAI/posa-bench/resolve/main/release/README.md
990 Bytes
| # POSA Bench preview-v0.1.0 | |
| This directory freezes the first fixed-context Pixel 7 baseline accepted for | |
| internal comparison. It is an **unranked directional preview**, not an official | |
| leaderboard release. | |
| The snapshot contains the exact questions, model manifest, methodology, | |
| scorer, raw responses, run metadata, automatic result artifacts, structured | |
| semantic-review decisions, and internal reviewed reports used for the preview. | |
| `manifest.json` records a SHA-256 and byte size for every frozen file. | |
| The release includes one completed repetition (seed 17) per model. Official | |
| ranking still requires seeds 17, 29, and 41, controlled thermal resets, and two | |
| independent blinded human reviewers plus adjudication. | |
| The POSA end-to-end Liquid LEAP result is deliberately excluded because it is | |
| a different track and includes retrieval and prompt-packing behavior. | |
| Verify the frozen snapshot with: | |
| ```sh | |
| python3 benchmark/scripts/freeze_preview.py --release preview-v0.1.0 --verify | |
| ``` | |