VideoMemBench - 64 Frames - Episodic Dependency Accuracy: leaderboard
Video-grounded multi-turn dialogue memory under sequential self-rollout, with model answers retained in history. Turn accuracy, episodic dependency accuracy and answer-chain accuracy are separate, overlapping views; the overall score is their equal mean. 326 dialogues; sequential self-rollout with full previous model answers; uniform timestamped video frames; no audio/subtitles; complete conversation history; exact deployment checkpoints and thinking modes unreported; native MC extraction and GPT-5.4 mini open-answer grading; percentages; 64 frames. Metric: Accuracy (%).
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 235B A22B (Benchmark Sweep Run: source=VideoMemBench; frames=64; self-rollout; exact-checkpoint-and-thinking-mode=unreported) | 39.3 | |
| 2 | Qwen 3.5 27B (Benchmark Sweep Run: source=VideoMemBench; frames=64; self-rollout; exact-checkpoint-and-thinking-mode=unreported) | 36.7 | |
| 3 | Qwen 3.5 35B A3B (Benchmark Sweep Run: source=VideoMemBench; frames=64; self-rollout; exact-checkpoint-and-thinking-mode=unreported) | 36.7 | |
| 4 | Qwen 3.5 122B A10B (Benchmark Sweep Run: source=VideoMemBench; frames=64; self-rollout; exact-checkpoint-and-thinking-mode=unreported) | 36.5 | |
| 5 | Qwen 3 VL 32B (Benchmark Sweep Run: source=VideoMemBench; frames=64; self-rollout; exact-checkpoint-and-thinking-mode=unreported) | 36.1 | |
| 6 | Qwen 3.5 9B (Benchmark Sweep Run: source=VideoMemBench; frames=64; self-rollout; exact-checkpoint-and-thinking-mode=unreported) | 33.2 | |
| 7 | Qwen 3 VL 30B A3B (Benchmark Sweep Run: source=VideoMemBench; frames=64; self-rollout; exact-checkpoint-and-thinking-mode=unreported) | 32.9 | |
| 8 | Qwen 3 VL 8B (Benchmark Sweep Run: source=VideoMemBench; frames=64; self-rollout; exact-checkpoint-and-thinking-mode=unreported) | 32 | |
| 9 | Qwen 3 VL 4B (Benchmark Sweep Run: source=VideoMemBench; frames=64; self-rollout; exact-checkpoint-and-thinking-mode=unreported) | 30.3 | |
| 10 | Keye-VL-1.5-8B (Benchmark Sweep Run: source=VideoMemBench; frames=64; self-rollout; exact-checkpoint-and-thinking-mode=unreported) | 24.9 |
Interactive version: theaggregate.ai/benchmark?slug=videomembench-64-frames-episodic-dependency-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-06.