EgoCoT-Bench - Hand-Object Association: leaderboard

EgoCoT-Bench (2026) Hand-Object Association subtask: from an egocentric manipulation video, determine which hand interacts with which object. Four-way multiple choice accuracy in percent; higher is better.

Metric: Answer accuracy (%, four-way multiple choice, strict exact match) of zero-shot MLLMs on egocentric manipulation videos of EgoCoT-Bench (351 clips, 3,172 human-verified questions built from spatio-temporal scene graphs), Hand-Object Association subtask (determine which hand interacts with which object); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 19 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5.163.16#248
2InternVL3.5-2B63.16#3313
3GPT-5.262.28#190
4Qwen 3.5 Plus62.28#250
5Qwen 3 VL 32B62.28#624
6Qwen 3.5 27B61.4#380
7Qwen 3.5 122B A10B61.4#350
8InternVL3.5-8B61.4#1618
9Qwen 3 VL Plus61.4#514
10Qwen 3 VL 30B A3B61.06#791
11InternVL3.5-4B59.65#1946
12Qwen 3 VL 8B59.29#927
13Qwen 3.5 397B A17B58.77#270
14LLaVA-OneVision-1.5-8B58.77#1876
15Qwen 3 VL 235B A22B57.02#486

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

Interactive version: theaggregate.ai/benchmark?slug=egocot-bench-hand-object-association · How It Works · Data refreshed daily, snapshot 2026-11-22.