| # Video-AMME CI | |
| Video-AMME is a 50-case CI dataset derived from `zhaochenyang20/Video_MME_ci`. | |
| Each example keeps the Video-MME video and moves the question, answer | |
| choices, and answer-format instruction into a spoken WAV file. | |
| ## Files | |
| - `data/test.jsonl`: metadata and source Video-MME references. | |
| - `audios/*.wav`: spoken question/options/instruction. | |
| - `videos/*.mp4`: present only when built with `--copy-videos`. | |
| ## Generation | |
| - TTS model: `fishaudio/s2-pro` | |
| - Max samples requested: `50` | |
| - Videos copied into this dataset: `False` | |
| The benchmark request text must contain routing/format instructions only; the | |
| semantic question content lives in `audio_text` and the WAV file. | |