EgoIntrospect - Memory Recall Prediction: leaderboard

Which of four details of a three-minute clip the wearer later recalled as vivid, asked on egocentric smart-glasses clips annotated by the wearers themselves (EgoIntrospect, 2026); native audio and video models hear the clip audio and the others read a transcript. Accuracy in percent on 200 four-choice questions; higher is better.

Metric: Accuracy (%) on 200 four-choice questions asking which detail of a 3-minute clip the wearer later recalled as vivid (distractors are other elements of the same clip written by Gemini and curated by the authors), fixation-overlaid egocentric clips from 15 held-out participants with the participant profile; native audio-video models hear the clip audio, the other models read its transcript; temperature 0 (0.6 for InternVL3.5 thinking mode); higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 19 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Qwen3 Omni 30B A3B Instruct55.5#930
2Qwen 3.5 Omni Plus52#254
3GPT-4o51.5#751
4Gemini 3.1 Pro (Preview)50#99
5MiniCPM-V-4.550#1441
6Qwen 3.5 397B A17B49#264
7LongVA-7B-DPO49#2188
8Kimi K2.5Thinking48.5#266
9Qwen 3 VL 8B Instruct47.5#1029
10Qwen 3 Omni 30B A3B (Thinking)47.5
11Gemini 3 Flash (Preview)47#152
12LLaVA-OneVision-7B47#2981
13Qwen 3 VL 8B (Thinking)43#910
14InternVL2.5-8B43#2370
15LLaVA-Video-7B-Qwen242.5#1978

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

Interactive version: theaggregate.ai/benchmark?slug=egointrospect-memory-recall-prediction · How It Works · Data refreshed daily, snapshot 2026-11-21.