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

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Kenan, Jin, Qirui, Zhu, Liao, Huang, Xiaosong, Wu, Yijia, Zhang, Yikai, Zhang, Xin, Jin, Zijian, Huang, Yufan, Nallipogu, Elsie, Zhang, Chaoyun, Kang, Yu, Rajmohan, Saravan, Lin, Qingwei, Lee, Wenke, Zhang, Dongmei
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911725853343744
author Li, Kenan
Jin, Qirui
Zhu, Liao
Huang, Xiaosong
Wu, Yijia
Zhang, Yikai
Zhang, Xin
Jin, Zijian
Huang, Yufan
Nallipogu, Elsie
Zhang, Chaoyun
Kang, Yu
Rajmohan, Saravan
Lin, Qingwei
Lee, Wenke
Zhang, Dongmei
author_facet Li, Kenan
Jin, Qirui
Zhu, Liao
Huang, Xiaosong
Wu, Yijia
Zhang, Yikai
Zhang, Xin
Jin, Zijian
Huang, Yufan
Nallipogu, Elsie
Zhang, Chaoyun
Kang, Yu
Rajmohan, Saravan
Lin, Qingwei
Lee, Wenke
Zhang, Dongmei
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2604_07789
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle ORACLE-SWE: Quantifying the Contribution of Oracle Information Signals on SWE Agents
Li, Kenan
Jin, Qirui
Zhu, Liao
Huang, Xiaosong
Wu, Yijia
Zhang, Yikai
Zhang, Xin
Jin, Zijian
Huang, Yufan
Nallipogu, Elsie
Zhang, Chaoyun
Kang, Yu
Rajmohan, Saravan
Lin, Qingwei
Lee, Wenke
Zhang, Dongmei
Multiagent Systems
Computation and Language
Software Engineering
I.2.7; I.2.5
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.
title ORACLE-SWE: Quantifying the Contribution of Oracle Information Signals on SWE Agents
topic Multiagent Systems
Computation and Language
Software Engineering
I.2.7; I.2.5
url https://arxiv.org/abs/2604.07789