BASS - Full Structural Segmentation: leaderboard
Metric: Label-matched intersection over union (%) on the 74 full structural segmentation questions: segment a whole song into its labelled sections with start and end times (free-form answers); BASS music understanding benchmark, audio language models get a text prompt (one of ten paraphrases) and a full commercial music recording from BASS's songs; raw score before chance normalization; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 32.34 | #145 |
| 2 | Voxtral-Small-24B-2507 | 32.16 | #565 |
| 3 | SALMONN-7B | 3.18 | #1274 |
| 4 | Phi-4 Multimodal Instruct | 2.71 | #896 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=bass-full-structural-segmentation · How It Works · Data refreshed daily, snapshot 2026-10-11.