← sources

[Paper] REMORY: Learning Residual Memory for Context Compaction

source: reddit (reddit:LocalLLaMA)published 2026-10-20 13:50ingested 2026-10-21 02:41

retained text

>Long-horizon agents compact their history to continue within a finite context window, but a textual summary alone may not support every subsequent decision. We introduce REMORY, a neural memory network that supplements the summary with a bounded sequence of soft memory tokens. Given the history and summary, the network learns to generate tokens that help a frozen LLM approximate the continuation it would produce with the full history. The tokens are conditioned on the summary and appended after it, forming an analogue of a residual connection along the sequence dimension. On SummHay, REMORY improves source attribution at nearly unchanged insight coverage and approaches the full-context joint score using only 5.2% of the input positions. Across long-horizon agent benchmarks, Qwen3.8-27B and GLM-5.3-Flash show consistent gains with residual memory. Both models also exhibit substantially fewer repeated tool outputs and tool errors on BrowseComp and Terminal-Bench 2.1.

* **arXiv** : [https://arxiv.org/abs/2610.11287](https://arxiv.org/abs/2610.11287)
* **PDF** : [https://arxiv.org/pdf/2610.11287](https://arxiv.org/pdf/2610.11287)
* **GitHub** : [https://github.com/fudan-generative-vision/Remory](https://github.com/fudan-generative-vision/Remory)
* **HuggingFace** : [https://huggingface.co/collections/fudan-generative-ai/remory](https://huggingface.co/collections/fudan-generative-ai/remory)

document fetched 2026-10-21 00:41 · canonical https://reddit.com/r/LocalLLaMA/comments/1xasp90/paper_remory_learning_residual_memory_for_context