IPIBench - Interleaved Reactive-Proactive Requests: 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 the interleaved reactive-proactive request tasks (mean of reactive-to-proactive, reactive-under-proactive and reactive-after-proactive); percent, higher is better.

Metric: Accuracy (%) on the interleaved reactive-proactive request tasks (mean of reactive-to-proactive, reactive-under-proactive and reactive-after-proactive), 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 March 2028. 15 models tracked. Updated 14 Nov 2026.

Top models

#ModelScoreOverall rank
1Qwen 3.5 Plus28.28#250
2GPT-5.423.79#138
3GPT-4o23.63#754
4Gemini 3 Pro23.04#146
5GLM-4.6V22.76#721
6Qwen 3 VL 8B22.41#927
7Gemini 2.5 Pro22.32#280
8Qwen 3.5 9B (Harness: AURA2 IPIBench offline30frames; Thinking: On)19.83
9Qwen 3.5 9B (Harness: AURA2 IPIBench offline30frames; Thinking: Off)15.02
10JoyAI-VL-Interaction6.53
11MOSS-VL-Realtime6.02
12InternVL3-8B5.5#1795
13LLaVA-OneVision-7B2.6#3002
14VideoLLM-online-8B0.88
15Flash-VStream-7B0.26

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=ipibench-interleaved-reactive-proactive-requests · How It Works · Data refreshed daily, snapshot 2026-11-22.