ResearchResearch paperLarge Language Models · Efficiency & Inference · Retrieval, RAG & Search1 source · Oct 16, 2026

Beyond Marginal Influence: Interaction-Aware Memory Selection for LLM Agents

LLM-agent memory selection often uses retrieval rankings based on record relevance or individual influence.

We study per-decision memory sufficiency: selecting a small subset that preserves full-memory behavior.
To separate search error from model error, we introduce JointCoreBench, a benchmark with nine interaction families, six workflow domains, natural-language decisions, and exact sufficient-set oracles, and JointCore, a relevance cost-ordered group-deletion baseline that tests blocks before individual records.
Both methods preserve full-memory behavior under the deterministic primary interface.
Tests with physical deletion, UNKNOWN masks, PENDING sentinels, counterfactual values, and alternative labels identify interface dependent behavior.

[1]Beyond Marginal Influence: Interaction-Aware Memory Selection for LLM Agents
arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 16, 10:37 AM
LLM-agent memory selection often uses retrieval rankings based on record relevance or individual influence.
We study per-decision memory sufficiency: selecting a small subset that preserves full-memory behavior.

Extractive summary: sentences quoted from the sources.

Before this