Drawing on Memory: Dual-Trace Encoding Improves Cross-Session Recall in LLM Agents

TL;DR

This paper introduces dual-trace memory encoding for LLM agents, pairing facts with scene reconstructions to enhance cross-session recall and temporal reasoning.

Contribution

It presents a novel dual-trace encoding method that significantly improves cross-session recall in LLM agents by creating richer memory traces.

Findings

Dual-trace encoding achieves 73.7% accuracy versus 53.5% for fact-only.

Gains are concentrated in temporal reasoning, knowledge-update tracking, and multi-session aggregation.

The method improves recall without additional token costs.

Abstract

LLM agents with persistent memory store information as flat factual records, providing little context for temporal reasoning, change tracking, or cross-session aggregation. Inspired by the drawing effect [3], we introduce dual-trace memory encoding. In this method, each stored fact is paired with a concrete scene trace, a narrative reconstruction of the moment and context in which the information was learned. The agent is forced to commit to specific contextual details during encoding, creating richer, more distinctive memory traces. Using the LongMemEval-S benchmark (4,575 sessions, 100 recall questions), we compare dual-trace encoding against a fact-only control with matched coverage and format over 99 shared questions. Dual-trace achieves 73.7% overall accuracy versus 53.5%, a +20.2 percentage point (pp) gain (95% CI: [+12.1, +29.3], bootstrap p < 0.0001). Gains concentrate in temporal reasoning (+40pp), knowledge-update tracking (+25pp), and multi-session aggregation (+30pp), with no benefit for single-session retrieval, consistent with encoding specificity theory [8]. Token analysis shows dual-trace encoding achieves this gain at no additional cost. We additionally sketch an architectural design for adapting dual-trace encoding to coding agents, with preliminary pilot validation.