RefMem-Bench - Multi-Choice: leaderboard
Reflective memory in long multi-session dialogues on RefMem-Bench (Lin et al., 2026, Connecting the Dots): questions ask for latent interpretations such as beliefs, habits, boundaries and plans that must be pieced together from cues scattered across the history, with images given as indexed captions. Multi-choice format with one to four correct options scored by exact set match; accuracy in percent averaged over eight reflective-memory dimensions on the test split, higher is better.
Metric: Answer accuracy (%; multi-choice questions with one to four correct options, exact set match; mean of the eight reflective-memory dimensions over the 20,279-question test split; dialogue history with image captions in context, temperature 0). Source: arxiv.org. Saturation forecast: Around February 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek R1 Distill Qwen 14B | 24.6 | #2594 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=refmem-bench-multi-choice · How It Works · Data refreshed daily, snapshot 2026-11-12.