Scope Before You Persist: Preventing Cross-Family Interference in Agent Memory

2026-09-24

AI Summary

This study addresses the performance degradation caused by cross-task interference in the persistent memory of agents. To mitigate this issue, we propose Scope Matching as a supplementary control mechanism. By aligning the retrieval scope with the certification scope and introducing Orthogonal Regression Control (ORC) alongside a verification gating strategy, the proposed method achieves reliable repeated adaptation. We further construct a new benchmark, ProcStream-RSI, to evaluate our approach. Experimental results demonstrate that the proposed method improves trajectory utility to 0.816 and completely eliminates harmful deployments, significantly outperforming global memory baselines.

Abstract

Persistent memory lets language-model agents improve prompts and skills without updating model weights. We show that matching retrieval scope to certification scope enables these edits to support reliable repeated adaptation across recurring task families. We study frozen-model agents on ProcStream-RSI, a 12-round code-repair stream, using Orthogonal Regression Control (ORC), an execution-grounded gate for persistent skill edits. In an intervention that holds proposals and gate decisions fixed, retrieving each accepted skill only for its originating family raises mean hidden trajectory utility from 0.713 under global memory to 0.816 and changes harmful deployments from six of eight to none. In 27 paired randomized-order streams, Scoped-ORC improves mean trajectory utility by 0.063 [0.037, 0.094] over Global-ORC, accepts 63 rather than 12 updates, and produces multiple accepted updates in 19/27 streams, with 0/63 harmful acceptances. The global control reaches 0.713, below the static agent's 0.775, because locally valid edits can interfere with unrelated families. These results establish scope matching as a complementary control for persistent agent memory: certification determines whether an edit is supported, while retrieval scope determines where that evidence authorizes its use.

Problem

• persistent memory
• cross-family interference
• language-model agents
• retrieval scope

Innovation

• Persistent Memory
• Scope Matching
• Orthogonal Regression Control (ORC)
• Cross-Family Interference
• Language Model Agents