IPIBench - Reactive-to-Proactive: leaderboard

IPIBench (2026) streams videos to multimodal LLMs at one frame per second and checks whether they respond proactively at the right moment, manage monitoring tasks the user cancels or changes, and handle reactive questions interleaved with proactive tasks. This board is accuracy on reactive-to-proactive interaction (start tracking a new task from a reactive exchange); percent, higher is better.

Metric: Accuracy (%) on reactive-to-proactive interaction (start tracking a new task from a reactive exchange), on IPIBench (1,831 videos, 3,738 interaction instances from egocentric and exocentric datasets) under a unified streaming protocol at 1 FPS (offline models see a sliding window of the latest 16 frames); a proactive response counts only if the first trigger falls within one second of the ground-truth moment, multi-turn tasks only if every required trigger and answer is correct, reactive answers by bidirectional substring matching against up to 20 reference answers; higher is better. Source: arxiv.org. Saturation forecast: Around January 2028. 15 models tracked. Updated 14 Nov 2026.

Top models

#ModelScoreOverall rank
1Qwen 3.5 Plus32.25#242
2GPT-4o27.04#751
3GPT-5.427.04#135
4Qwen 3.5 9B (Harness: AURA2 IPIBench offline30frames; Thinking: Off)27.04
5Gemini 3 Pro24.76#145
6Qwen 3 VL 8B24.43#917
7GLM-4.6V22.15#710
8Gemini 2.5 Pro21.17#276
9JoyAI-VL-Interaction16.94
10Qwen 3.5 9B (Harness: AURA2 IPIBench offline30frames; Thinking: On)15.96
11InternVL3-8B7.17#1782
12MOSS-VL-Realtime7.17
13LLaVA-OneVision-7B3.91#2981
14Flash-VStream-7B0
15VideoLLM-online-8B0

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=ipibench-reactive-to-proactive · How It Works · Data refreshed daily, snapshot 2026-11-21.