CASU - Counterfactual Reasoning: leaderboard
Context-aware auditory scene understanding: four-option questions on short semi-synthetic clips that mix speech, sound events and background, asking how an interpretation of the scene would change if part of the acoustic evidence were different; CASU benchmark (Zhang et al., UC Irvine, 2026). Accuracy in percent, higher is better.
Metric: Accuracy (%; four-option multiple-choice questions on semi-synthetic auditory scenes that combine speech, sound events and background; counterfactual reasoning: how an interpretation changes if the acoustic evidence were altered). Source: arxiv.org. Saturation forecast: Estimated already saturated. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen3 Omni 30B A3B Instruct | 74.5 | #902 |
| 2 | Gemini 2.0 Flash | 73.83 | #702 |
| 3 | Qwen2.5-Omni-7B | 65.38 | #1797 |
| 4 | Voxtral-Small-24B-2507 | 55.99 | #1291 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=casu-counterfactual-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-12.