arXiv:2611.24995v1 Announce Type: new Abstract: Understanding an egocentric environment requires reasoning about how surrounding entities affect the observer's actions. Existing evaluations of multimodal foundation models (MFMs) examine observer localization, motion trajectories, and action feasibility, but leave the discovery and interpretation of evolving observer-entity interactions insufficiently assessed. We introduce EgoPass-Bench, a benchmark for observer-relative interaction understanding in real-world egocentric video. EgoPass-Bench comprises 284 manually reviewed evaluation clips recorded using Ray-Ban Meta (Gen 2) smart glasses and 2,201 distinct questions across five tasks assessing motion perception, interaction judgments about specified entities, and relevant-entity discovery without target cues. Our evaluation of eight frame-based and six native-video configurations reveals a pronounced discrepancy between motion perception and interaction understanding. The strongest evaluated model, GPT-6 Astra, correctly recognizes object motion in 91.37% of 394 comparisons involving the same targets, but correctly identifies changes in passage constraints in only 66.50%. Across all 574 entity-discovery questions, it trails the two-person human baseline by $23.65 \pm 7.45$ percentage points (paired 95% CI), while exceeding the always-empty baseline (36.76%) by at least 18.9 percentage points. Further analysis reveals frequent inclusion of entities that neither pose a plausible physical conflict nor constrain passage, showing that accurate motion recognition does not ensure reliable relevant-entity discovery.
EgoPass-Bench: Identifying Objects That May Affect the Wearer's Passage in Egocentric Video
A new benchmark, EgoPass-Bench, finds that even the strongest multimodal model tested, GPT-6 Astra, correctly identifies changes in passage constraints in only 66.50% of cases despite recognizing object motion in 91.37% of 394 comparisons involving the same targets. The benchmark, built from 284 manually reviewed clips recorded on Ray-Ban Meta (Gen 2) smart glasses and 2,201 questions across five tasks, shows GPT-6 Astra trailing a two-person human baseline by 23.65 ± 7.45 percentage points on 574 entity-discovery questions while beating the always-empty baseline of 36.76% by at least 18.9 points. The authors conclude that accurate motion recognition does not ensure reliable discovery of entities relevant to the wearer's passage.
Evaluate egocentric models on relevant-entity discovery without target cues, not just motion recognition; GPT-6 Astra scored 91.37% on motion but only 66.50% on passage-constraint changes.
more in playbook: Evals and testing for LLM appsRun your AI side-project on zahid.host
EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.