Evaluating Memory Condensation Strategies for Coding Agents in Data-Driven Scientific Discovery
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909054956208128 |
|---|---|
| author | Chintalapati, Renuka Raskar, Sid Acharya, Anurag Willard, Jared Emami, Patrick Horawalavithana, Sameera |
| author_facet | Chintalapati, Renuka Raskar, Sid Acharya, Anurag Willard, Jared Emami, Patrick Horawalavithana, Sameera |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_18854 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Evaluating Memory Condensation Strategies for Coding Agents in Data-Driven Scientific Discovery Chintalapati, Renuka Raskar, Sid Acharya, Anurag Willard, Jared Emami, Patrick Horawalavithana, Sameera Machine Learning 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. |
| title | Evaluating Memory Condensation Strategies for Coding Agents in Data-Driven Scientific Discovery |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2605.18854 |