| title | REMORY: Learning Residual Memory for Context Compaction |
|---|---|
| authors | Hanchen Xia, Baoyou Chen, Yutang Ge, Naihao Deng, Senqiao Yang, Zilong Dong, Weihao Yuan, Siyu Zhu |
| date | 2026-10-08 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2610.11287 |
| affiliation | Fusion Lab: Generative Vision Lab of Fudan University and SII |
| upvotes | 54 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | 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. |
cite
@misc{xia2026remory,
title={REMORY: Learning Residual Memory for Context Compaction},
author={Hanchen Xia and Baoyou Chen and Yutang Ge and Naihao Deng and Senqiao Yang and Zilong Dong and Weihao Yuan and Siyu Zhu},
year={2026},
eprint={2610.11287},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2610.11287}
}