| title | δ-mem: Efficient Online Memory for Large Language Models |
|---|---|
| authors | Jingdi Lei, Di Zhang, Junxian Li, Weida Wang, Kaixuan Fan, Xiang Liu, Qihan Liu, Xiaoteng Ma, Baian Chen, Soujanya Poria |
| date | 2026-05-12 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2605.12357 |
| upvotes | 109 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | Large language models increasingly need to accumulate and reuse historical information in long-term assistants and agent systems. Simply expanding the context window is costly and often fails to ensure effective context utilization. We propose δ-mem, a lightweight memory mechanism that augments a frozen full-attention backbone with a compact online state of associative memory. δ-mem compresses past information into a fixed-size state matrix updated by delta-rule learning, and uses its readout to generate low-rank corrections to the backbone's attention computation during generation. With only an 8times8 online memory state, δ-mem improves the average score to 1.10times that of the frozen backbone and 1.15times that of the strongest non-δ-mem memory baseline. It achieves larger gains on memory-heavy benchmarks, reaching 1.31times on MemoryAgentBench and 1.20times on LoCoMo, while largely preserving general capabilities. These results show that effective memory can be realized through a compact online state directly coupled with attention computation, without full fine-tuning, backbone replacement, or explicit context extension. |
cite
@misc{lei2026mem,
title={δ-mem: Efficient Online Memory for Large Language Models},
author={Jingdi Lei and Di Zhang and Junxian Li and Weida Wang and Kaixuan Fan and Xiang Liu and Qihan Liu and Xiaoteng Ma and Baian Chen and Soujanya Poria},
year={2026},
eprint={2605.12357},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2605.12357}
}