TAR-Bench - Binary QA: leaderboard
TAR-Bench (2026) evaluates vision-language models on 960 human-curated questions about 80 held-out traffic-anomaly clips, zero-shot and without reasoning mode. The Binary QA task covers yes/no questions about the anomaly, one positive and one negative per clip. Score is accuracy, in percent; higher is better.
Metric: Accuracy (%; yes/no questions about the anomaly, one positive and one negative per clip; zero-shot in non-reasoning (direct-answer) mode through VLMEvalKit, mean of 3 runs, on TAR-Bench: 960 human-curated annotations for 80 held-out traffic-anomaly clips from 17 public YouTube videos). Source: arxiv.org. Saturation forecast: Around October 2028. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 27B (Non-reasoning) | 72.5 | #359 (Qwen 3.5 27B) |
| 2 | Qwen 3.5 9B (Non-reasoning) | 67.3 | #803 (Qwen 3.5 9B) |
| 3 | Qwen 3 VL 32B Instruct | 65 | #585 |
| 4 | Gemini 3.1 Pro (Preview) | 64.4 | #94 |
| 5 | Qwen 3 VL 8B Instruct | 59.6 | #981 |
| 6 | Gemma 4 31B (IT) | 56.5 | #342 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=tar-bench-binary-qa · How It Works · Data refreshed daily, snapshot 2026-11-13.