MemCompiler: Compile, Don't Inject -- State-Conditioned Memory for Embodied Agents

TL;DR

MemCompiler introduces a dynamic, state-conditioned memory compilation approach for embodied agents, significantly improving effectiveness and efficiency over static memory injection methods across multiple benchmarks.

Contribution

It presents a novel memory utilization paradigm that dynamically compiles relevant memory based on current agent state, outperforming static injection methods.

Findings

Up to +129% improvement over no-memory baselines.

Reduces per-step latency by 60%.

Matches or approaches performance of frontier closed-source systems.

Abstract

Existing memory systems for embodied agents typically inject retrieved memory as static context at episode start, a paradigm we term Ahead-of-time Monolithic Memory Injection (AMMI). However, this static design quickly becomes misaligned with the agent's evolving state and may degrade lightweight executors below the no-memory baseline. To address this, we propose MemCompiler, which reframes memory utilization as State-Conditioned Memory Compilation. A learned Memory Compiler reads a structured Brief State capturing the agent's current execution state and dynamically selects and compiles only relevant memory into executable guidance. This guidance is delivered through a text channel and a latent Soft-Mem channel that preserves perceptual information not expressible in text. Across Alf World, EmbodiedBench, and ScienceWorld, MemCompiler consistently improves over no-memory across open-source backbones (up to +129%), matches or approaches frontier closed-source systems, and reduces per-step latency by 60%, demonstrating that state-aware memory compilation improves both effectiveness and efficiency.