EgoThinker-7B: benchmark results

Provider: Other. Released 2025-10-29. Access: Open.

Aggregate Rating 1551 ± 51, rank #1612 of 10556 rated models, from 20 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
EgoActionBench - Temporal m-tIoU13.7Mean matched temporal IoU (0-100; all 1,407 episodes (Ego4D,77.8
EgoActionBench - Temporal18.9Temporal Overall (0-100; mean of m-tIoU and micro-averaged F61.1
EgoActionBench20.8EgoActionScore (0-100; mean of Temporal Overall and Semantic44.4
EgoActionBench - Ego4D subset22.6EgoActionScore (0-100; mean of Temporal Overall and Semantic44.4
EgoActionBench - EPIC-KITCHENS-100 subset20.4EgoActionScore (0-100; mean of Temporal Overall and Semantic33.3
EgoActionBench - EgoLife subset18.9EgoActionScore (0-100; mean of Temporal Overall and Semantic33.3
EgoActionBench - Semantic Hand33.7Hand score (0-100; hand-label agreement of the prediction se33.3
EgoActionBench - Temporal F124.1Micro-averaged action F1 from one-to-one matching (0-100; al33.3
EgoActionBench - Semantic Motion19.5Motion agreement judged by the language-model judge (0-100; 30.6
EgoActionBench - Semantic Verb19Verb agreement judged by the language-model judge (0-100; al27.8
EgoActionBench - Semantic22.8Semantic Overall (0-100; applicable Hand, Verb, Objects, Too22.2
EgoActionBench - Semantic Objects19.1Objects agreement judged by the language-model judge (0-100;16.7

Interactive version: theaggregate.ai/model?slug=egothinker-7b · How It Works · Data refreshed daily, snapshot 2026-11-17.