arXiv cs.CL· Tianxiang Fei, Jiazheng Kang, Mao Zheng, Xingyan He, Xiang Yu·· 2 天前AI 评分34
Hy-Memory-30B-A3B：面向长期记忆写入的归因强化学习
Hy-Memory-30B-A3B: Attributed Reinforcement Learning for Long-Term Memory Writing
AI 导读
研究团队提出 Hy-Memory-30B-A3B，一个跨 Mem0、MemOS 和 Hy-Memory 训练的紧凑型长期记忆写入模型，通过 Memory On-Policy Distillation 学习各框架写入协议，再用 Memory RL 结合分块探针、跨块归因奖励与 chunk 级 GRPO 优化下游可答性。
正文 · 原文
Abstract:External-memory systems extend language-model applications beyond a context window, but their quality depends critically on the model executing the write path: deciding what to retain, how precisely to represent it, and how new evidence should update persistent state. Existing frameworks typically place a general-purpose large language model behind framework-specific prompts. This raises serving cost, and a writer specialized for one protocol does not transfer directly to another; a unified model that can operate across established memory frameworks remains largely unexplored. We present Hy-Memory-30B-A3B, a compact long-term-memory writer trained across Mem0, MemOS, and Hy-Memory. Memory On-Policy Distillation teaches framework-specific write protocols on student-generated states; Memory RL then optimizes downstream answerability using dense per-chunk probes, attributed cross-chunk rewards, over-storage controls, and chunk-level GRPO. Carrying policy-written banks across windows keeps training aligned with deployment. On fixed 20% benchmark subsamples, Hy-Memory-30B-A3B ranks first on all four evaluated harnesses, including held-out EverMemOS, with a 55.49% macro average versus 54.62% for Claude Opus 5.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.22196 [cs.CL] |
| (or arXiv:2610.22196v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.22196 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Tianxiang Fei [view email]
[v1]
Thu, 15 Oct 2026 16:46:47 UTC (699 KB)
来源：arXiv cs.CL · arxiv.org