Memory Transfer Learning: How Memories are Transferred Across Domains in Coding Agents

TL;DR

This paper introduces Memory Transfer Learning (MTL), enabling coding agents to transfer shared memory across diverse domains, improving performance by leveraging high-level insights while highlighting the importance of abstraction.

Contribution

It pioneers the use of heterogeneous domain memory pools in coding agents, demonstrating improved transferability and establishing empirical principles for cross-domain memory utilization.

Findings

Cross-domain memory improves average performance by 3.7%.

High-level insights transfer well, low-level traces can cause negative transfer.

Memory transfer effectiveness scales with memory size and can be model-agnostic.

Abstract

Memory-based self-evolution has emerged as a promising paradigm for coding agents. However, existing approaches typically restrict memory utilization to homogeneous task domains, failing to leverage the shared infrastructural foundations, such as runtime environments and programming languages, that exist across diverse real-world coding problems. To address this limitation, we investigate \textbf{Memory Transfer Learning} (MTL) by harnessing a unified memory pool from heterogeneous domains. We evaluate performance across 6 coding benchmarks using four memory representations, ranging from concrete traces to abstract insights. Our experiments demonstrate that cross-domain memory improves average performance by 3.7\%, primarily by transferring meta-knowledge, such as validation routines, rather than task-specific code. Importantly, we find that abstraction dictates transferability; high-level insights generalize well, whereas low-level traces often induce negative transfer due to excessive specificity. Furthermore, we show that transfer effectiveness scales with the size of the memory pool, and memory can be transferred even between different models. Our work establishes empirical design principles for expanding memory utilization beyond single-domain silos. Project page: https://memorytransfer.github.io/