The Past Is Prologue: A Plug-in Controller for Selective Updates in Sequentially Evolving LLM Memory
Quick Answer
Janus is a plug-in memory controller for LLMs that selectively updates memory, improving accuracy by 2.7 to 4.6 points across various datasets.
Quick Take
Janus is a plug-in memory controller for LLMs that selectively updates memory, improving accuracy by 2.7 to 4.6 points across various datasets. It uses a Memory Momentum Trigger to evaluate updates efficiently, preventing the loss of useful knowledge. This method is agnostic to existing updaters, enhancing performance without altering their rules.
Key Points
- Janus improves LLM memory updates without altering existing update rules.
- Utilizes a Memory Momentum Trigger to assess memory update effectiveness.
- Achieves accuracy improvements of 2.7 to 4.6 points across six datasets.
- Method is agnostic, compatible with various LLM architectures and updaters.
- Prevents overwriting valuable knowledge and reduces bias towards recent examples.
Paper Resources
Article Excerpt
From source RSS / original summaryarXiv:2606. 31121v1 Announce Type: new Abstract: Sequentially evolving LLM memory enables agents to reuse past experience, but existing systems usually deploy each locally generated memory update without checking whether it improves future behavior. As a result, updates that help the current task may overwrite useful knowledge, introduce over-specific rules, or bias the final memory toward recent examples.
We propose Janus, a plug-in memory controller that decides whether to accept a candidate memory update or retain the previous memory. To make this decision efficient, Janus uses a Memory Momentum Trigger to identify suspicious deviations in the memory-update trajectory, and compares old and new memories on a compact hybrid evaluation set of coverage, boundary, and fresh tasks instead of replaying the full history. Janus is method-agnostic and wraps existing updaters without changing their update rules.
Across six datasets, two backbone LLMs, and two memory updaters, Janus improves average accuracy by +2. 7 to +4. 6 points over the corresponding base updaters.
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →The Verification Horizon: No Silver Bullet for Coding Agent Rewards
As coding agents evolve, verifying solutions becomes more challenging than generating them, necessitating a focus on scalable, faithful, and robust verification methods. The study reveals that no fixed reward function can sustain effectiveness as model capabilities advance, emphasizing the need for verification to evolve alongside solution generation.