SCENEBench (Audio Understanding) - Noise Localization (Direction Question): leaderboard
Metric: Accuracy (%) on SCENEBench noise localization: 6,000 ESC-50 environmental sound clips given an approaching, receding or oscillating amplitude envelope (three per source clip); a response is correct when it names the right motion pattern or a synonym, after a follow-up question about the movement of the sound source (FR2), with clips already solved in the free description counted correct; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen2-Audio-7B-Instruct | 57.23 | #859 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=scenebench-audio-understanding-noise-localization-direction-question · How It Works · Data refreshed daily, snapshot 2026-10-11.