S-EMBER - Interval mIoU: leaderboard

Streaming egocentric memory retrieval (S-EMBER, Wang et al., Meta, 2026): 9,448 questions about 388 hours of first-person video from Ray-Ban Meta glasses, each asked at a trigger time and answerable only from what the wearer saw earlier, with open models given 128 uniformly sampled frames and proprietary models their platform limits. The model also names the time interval that supports its answer; mean temporal IoU with the annotated interval, from 0 to 1, higher is better.

Metric: Mean temporal IoU (0-1) between the predicted evidence interval and the annotated one (S-EMBER: 9,448 questions on 3,141 egocentric Ray-Ban Meta videos, answered from the stream up to the question's trigger time; open models at 128 uniformly sampled frames, proprietary models at their platform limits). Source: arxiv.org. Saturation forecast: Around February 2028. 5 models tracked.

Top models

#ModelScoreOverall rank
1Gemini 3.1 Pro (Preview)0.48#91
2Qwen 3 VL 32B Instruct0.41#582
3GPT-5.40.33#121

No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.

Interactive version: theaggregate.ai/benchmark?slug=s-ember-interval-miou · How It Works · Data refreshed daily, snapshot 2026-11-12.