OMD-Bench (Real, CoT) - Two-Modality Corruption: leaderboard
Metric: Accuracy (%) against the human evidence protocol's target (the answer the remaining modality supports, or abstention) when two modalities are replaced, mean of the three pairs (k=2), on the real split (stock video, recorded audio and human-written text) of OMD-Bench (255 knowledge questions about 27 anchors, objects or events shown at once in video, audio and text; five options including an explicit abstention; unparseable answers are excluded), chain-of-thought prompting; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen3 Omni 30B A3B Instruct | 67.5 | #362 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=omd-bench-real-cot-two-modality-corruption · How It Works · Data refreshed daily, snapshot 2026-10-11.