Personalized Memory Rubrics Baselines - PersonaBench Noise 0.3: leaderboard
Personalized Memory Rubrics Baselines - PersonaBench Noise 0.3: Native setting score (%). Personalized Memory Rubrics arXiv 2611.14017v1: long-context, RAG, Mem0, A-Mem and LightMem use unmodified Qwen2.5-7B-Instruct. Retrieval uses all-MiniLM-L6-v2 top ten. Twelve native settings remain separate. MemAgent/Mem-alpha/MemSkill are explicitly fine-tuned on 423 local examples; PMR is weight-trained; these unpublished resulting checkpoints are held rather than attributed to base models. Inference decoding not reported.
Metric: Native setting score (%). Source: arxiv.org. Saturation forecast: Around February 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 7B Instruct | 22.38 | #2583 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=personalized-memory-rubrics-baselines-personabench-noise-0-3 · How It Works · Data refreshed daily, snapshot 2026-11-12.