Coding agents accumulate extensive context during long-running tasks, yet fixed context windows force practitioners to choose between truncation and task failure. While numerous memory condensation strategies have been proposed, from simple sliding windows to LLM-generated summaries, no systematic comparison exists to guide strategy selection, especially in scientific discovery tasks. We evaluate eight memory condensation strategies using GPT-4o on sixty DiscoveryBench tasks spanning six scientific domains (480 total evaluations). We find that no condenser significantly alters hypothesis quality, while LLM-based condensers increase token costs by 24-94 percent, and masking tool-call outputs achieves an 8.6 percent net savings. We also observe that the optimal condenser for data-driven scientific discovery varies by scientific domain and task length.
@article{arxiv.2605.18854,
title = {Evaluating Memory Condensation Strategies for Coding Agents in Data-Driven Scientific Discovery},
author = {Renuka Chintalapati and Sid Raskar and Anurag Acharya and Jared Willard and Patrick Emami and Sameera Horawalavithana},
journal= {arXiv preprint arXiv:2605.18854},
year = {2026}
}