FMSU-Bench - Background Sound (Chinese): leaderboard

FMSU-Bench (2026) tests fine-grained speech understanding in Chinese and English with multiple-choice questions whose distractors include subtle acoustic confusions and text-only semantic traps, plus a tagged transcription task. This board is accuracy on identifying the background sound in the Chinese subset; percent, higher is better.

Metric: Accuracy (%) of identifying the background sound (BS), on the FMSU-Bench Chinese subset (500 to 1,000 manually verified items per task, from audiovisual sources), multiple-choice questions whose distractors include fine-grained acoustic confusions and text-only semantic traps; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3.1 Pro (Preview)66.7#99
2Gemini 3 Flash60.2#175
3Qwen 3 Omni 30B A3B57.6#847
4Kimi-Audio50.1#1735
5MiMo-Audio46.5#931
6Gemini 2.5 Flash46.4#463
7Audio Flamingo 342.1
8FMSU-Bench Step-Audio 2 (checkpoint unspecified)31.7
9Qwen2.5-Omni-7B29.1#1884
10Qwen2-Audio (FMSU-Bench checkpoint unspecified)12.2

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=fmsu-bench-background-sound-chinese · How It Works · Data refreshed daily, snapshot 2026-11-21.