ORACLE-SWE: Quantifying the Contribution of Oracle Information Signals on SWE Agents

TL;DR

This paper introduces Oracle-SWE, a method to isolate and quantify the impact of different oracle information signals on the performance of language model agents in software engineering tasks.

Contribution

It provides a unified approach to measure the individual contribution of various contextual signals in SWE agents, guiding future research priorities.

Findings

Quantifies the impact of individual signals on agent performance.

Evaluates the performance gain when signals are provided by strong language models.

Guides research prioritization for autonomous coding systems.

Abstract

Recent advances in language model (LM) agents have significantly improved automated software engineering (SWE). Prior work has proposed various agentic workflows and training strategies as well as analyzed failure modes of agentic systems on SWE tasks, focusing on several contextual information signals: Reproduction Test, Regression Test, Edit Location, Execution Context, and API Usage. However, the individual contribution of each signal to overall success remains underexplored, particularly their ideal contribution when intermediate information is perfectly obtained. To address this gap, we introduce Oracle-SWE, a unified method to isolate and extract oracle information signals from SWE benchmarks and quantify the impact of each signal on agent performance. To further validate the pattern, we evaluate the performance gain of signals extracted by strong LMs when provided to a base agent, approximating real-world task-resolution settings. These evaluations aim to guide research prioritization for autonomous coding systems.