ObsDriveBench - Risk-aware Decision-making: leaderboard

Metric: Accuracy (%; dimension average of risk identification, fallback action and trajectory evaluation; multiple-choice questions on synchronized camera, LiDAR BEV and radar BEV inputs from K-Radar under fog, rain, snow and low light, answers matched by an LLM judge; official inference settings). Source: arxiv.org. Saturation forecast: Around December 2026. 15 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview)80.22
2Qwen 3 VL 32B79.71
3Qwen 2.5 VL 7B67.51
4GPT-566.8
5InternVL3-8B64.75
6InternVL3-38B57.9

Interactive version: theaggregate.ai/benchmark?slug=obsdrivebench-risk-aware-decision-making · How It Works · Data refreshed daily, snapshot 2026-09-29.