TAR-Bench - Temporal Localization: leaderboard

Metric: Mean IoU (%; the predicted anomaly interval against the annotated interval; 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 2030. 11 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview)32.2
2Qwen 3.5 27B (Non-reasoning)26.7
3Qwen 3 VL 8B Instruct25.1
4Gemma 4 31B (IT)22.3
5Qwen 3 VL 32B Instruct21.1
6Qwen 3.5 9B (Non-reasoning)20.4

Interactive version: theaggregate.ai/benchmark?slug=tar-bench-temporal-localization · How It Works · Data refreshed daily, snapshot 2026-09-29.