Memory-Induced Tool-Drift in LLM Agents

2026-05-24

AI Summary

This study identifies and formally names a previously unrecognized issue—“memory-induced tool drift”—where implicit personality biases (e.g., cost sensitivity, risk preference) embedded in the long-term memory of large language model agents inadvertently interfere with tool invocation in contextually irrelevant scenarios. The authors introduce MEMDRIFT, a novel adversarial benchmark comprising 105 diverse scenarios, and employ activation vector analysis, attention tracing, and real-world MCP tool scanning to systematically evaluate bias propagation across multiple domains in state-of-the-art models. Experimental results demonstrate that such biased memory can increase tool-calling deviation by up to 3.6 points on a 5-point scale and adversely affect 608 real tool parameters. Critically, existing mitigation strategies, including prompt filtering and memory curation mechanisms, prove largely ineffective in alleviating this phenomenon.

Abstract

Modern LLM agents combine long-term memory for personalization with tool-calling interfaces for taking actions in the world -- a combination underpinning contemporary production systems. We study a previously unexamined failure of this combination: when personality-driven biases stored in memory (cost-consciousness, impatience, risk tolerance, etc.) silently affect tool calls in contexts where they are not applicable. We call this memory-induced tool-drift and operationalize it through MEMDRIFT, a benchmark of 105 scenarios spanning five bias dimensions and seven professional domains, generated through an automated adversarial pipeline. Across seven frontier models -- including those with extended reasoning -- biased memories raise deflection scores (a judge-scored measure of parameter deviation from unbiased baselines) by up to $+3.6$ points on a 1--5 scale. Tool-drift persists when memory management is handled by three production memory architectures. The phenomenon affects real-world tools: scanning 6{,}062 tools across 288 verified MCP servers, we flag 608 with susceptible parameters and confirm tool-drift on a validated subset. Mechanistically, biased memories act as implicit steering vectors, pushing activations along the same latent directions as explicit behavioral instructions. They also redistribute attention from task-relevant context toward memory entries with surface-level keyword overlap to the target parameter. Standard defenses -- prompt-based relevance instructions and memory filters -- reduce drift but do not eliminate it. As agents take increasingly consequential actions on a user's behalf, memory-induced tool-drift represents a systematic vulnerability that current safeguards do not address, motivating dedicated defenses at the intersection of memory management and tool-call generation.

Problem

• memory-induced tool-drift
• LLM agents
• tool-calling
• personality bias
• memory bias

Innovation

• memory-induced tool-drift
• LLM agents
• MEMDRIFT benchmark
• behavioral bias
• tool-call vulnerability