# Rare, Relevant, and Still Wrong: How Memory Fails LLM Group Decisions

Source: [arXiv](https://arxiv.org/abs/2610.37624v1)  
Feed7 permalink: https://feed7.dev/p/2610-37624v1-1modgvk  
Published: 2026-10-27T17:49:39.000Z  
Trust: Needs Review (needs_review)

## Why Included

A benchmark study finds that relevance-based and rarest-first memory selection both fail in multi-agent decisions, while attribution tags can amplify social-proof effects. Memory quality depends on the evidence retained, not just the selection heuristic.

## Source Summary

CrossProfile tests memory selection on **480 tasks** across 60 templates. Relevance-based selection can keep on-topic decoys, while rarest-first performs no better than random on average; removing or randomizing agent attribution tags raised accuracy by **13–24 points** in the classic hidden-profile setup.

## Practical Implication

Builders compressing multi-agent discussions should test whether summaries preserve the balance of supporting and opposing evidence, and whether attribution labels sway the final decision. A relevance filter alone may not protect against decoys.

## Agent-Ready Context

CrossProfile tests memory selection on **480 tasks** across 60 templates. Relevance-based selection can keep on-topic decoys, while rarest-first performs no better than random on average; removing or randomizing agent attribution tags raised accuracy by **13–24 points** in the classic hidden-profile setup.

Builders compressing multi-agent discussions should test whether summaries preserve the balance of supporting and opposing evidence, and whether attribution labels sway the final decision. A relevance filter alone may not protect against decoys.

The authors also report that full transcripts in natural discussions performed worse than no notes. These findings come from specific models and diagnostic tasks, so they do not establish how every memory system will behave.

## Connected Context

Feed7 judgment across 944 accumulated Signals:

This narrows confidence in memory-selection heuristics for group decisions: topical relevance can retain decoys, rarity performs no better than random on average, and attribution tags can shift accuracy substantially in a diagnostic setup. It also reports that full transcripts can hurt in natural discussions, so preserving more discussion is not by itself a reliable memory strategy.

- [Bayesian Belief Layer for Controllable Opinion Dynamics in LLM Agents](https://feed7.dev/p/2609-21997v1-0yth274) — The Bayesian belief layer makes agent state explicit and auditable; this study shows that attribution attached to remembered content can itself alter group decisions, making the treatment of source labels a relevant design choice.
- [What LLM Agents Say When No One Is Watching: Social Structure and Latent Objective Emergence in Multi-Agent Debates](https://feed7.dev/p/2607-02507v1-1ctgeey) — The debate study finds that social structure can affect agents' public and private positions, while this study measures attribution tags affecting group accuracy; both make social context a potential influence on multi-agent outcomes.
- [HypoEvolve: Genetic Algorithms Enable Multi-Agent LLMs to Discover Scientific Hypotheses](https://feed7.dev/p/2609-15938v1-0oezpu6) — HypoEvolve formalizes how candidate hypotheses are retained and selected; this study cautions that selection based on relevance or rarity alone can preserve decoys or fail to outperform random choice.

## Context Map

- Layer: agent
- Domains: research
- Topics: agent-memory, multi-agent, agent-evals

## Uncertainty

- The authors also report that full transcripts in natural discussions performed worse than no notes. These findings come from specific models and diagnostic tasks, so they do not establish how every memory system will behave.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.