Video-MemComp - VideoMME Medium: leaderboard

Video-MemComp v1: streaming evaluation of official development sets. Table 1 mixes reproduced starred rows and results from prior papers; frame sampling and average KV budget differ by system and remain in row identity. Backbone checkpoint revision and decoding are not fully specified; source-qualified systems are retained. This is video QA under cache compression, not an unmodified backbone score. VideoMME Medium. Metric: Accuracy (%).

Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 15 models tracked.

Top models

#ModelScoreOverall rank
1Video-MemComp Qwen2-VL-7B [frames 768; KV 100K]63.9
2Video-MemComp Qwen2-VL-7B + StreamMem [frames 0.5/0.2fps; KV 6K]62.4
3Video-MemComp Qwen2-VL-7B + InfiniPot-V [frames 768; KV 6K]60.8
4Video-MemComp Qwen2-VL-7B + Video-MemComp [frames 768; KV 6K]60.2
5Video-MemComp Qwen2.5-VL-3B + StreamMem [frames 4.0/0.5 fps; KV 6K]60.1
6Video-MemComp Qwen2.5-VL-3B + Video-MemComp [frames 768; KV 6K]60
7Video-MemComp Qwen2.5-VL-3B [frames 768; KV 100K]58.1
8Video-MemComp LLaVA-OneVision-7B [frames 32; KV 6K]57.4
9Video-MemComp InternVL3.5-2B + Video-MemComp [frames 64/100; KV 6K]57.4
10Video-MemComp LLaVA-OneVision-7B + Video-MemComp [frames 0.5/0.2 fps; KV 6K]57.2

Interactive version: theaggregate.ai/benchmark?slug=video-memcomp-videomme-medium · How It Works · Data refreshed daily, snapshot 2026-10-27.