Coding agents struggle to continue imperfect early work in software tasks
SWE-Residual: Evaluating Starting-State Robustness of Coding Agents
Software EngineeringComputation and Language
Summary
Software tools that write or fix code usually start a task with a clean slate, but in real life they often have to continue work that someone else left unfinished. The authors created a new test called SWE-Residual to see how well coding agents handle starting with partially done and still incorrect work. They found that many current agents perform much worse when they have to pick up from these imperfect starting points, often failing to recognize what still needs fixing. This reveals a weakness in current tests and coding agents, showing they need better ways to deal with leftover code work.
What this means in practice
- •For software development teams: Evaluate and improve coding assistants’ ability to handle ongoing projects that contain previous incomplete or incorrect work using SWE-Residual scenarios.
- •For continuous integration engineers: Detect weaknesses in automated code reviewers or fixers when dealing with partial fixes inherited from earlier runs in codebases or environments.
Authors
Jinze Li, Weijie Liu, Jinfeng Xu, Shuo Yang, Xin Yang, Yang Zhang, Yijin Wang, Zifei Shan, Edith C. H. Ngai
Abstract
Software engineering (SWE) benchmarks are the primary way to measure the capabilities of coding agents: an agent is placed in a code repository or machine environment, given a request, and judged by hidden tests. These benchmarks almost always start the agent from a clean state that contains no earlier work on the task. In practice, however, agents often continue work that a developer or another agent has already done and reported as complete, with no indication of what remains incorrect or missing. We introduce SWE-Residual, a benchmark for Starting-State Robustness: the ability to maintain performance on a task when the starting state already contains earlier work on that task. SWE-Residual contains 198 Residual States drawn from 111 tasks in DeepSWE and Terminal-Bench, where the earlier work modifies a code repository or a whole machine environment. Each Residual State is the state a predecessor agent left behind after submitting its work as complete, although the hidden tests still fail. The evaluated agent receives only the original request and this state, without the predecessor's trajectory or any sign that the attempt failed; because the request and hidden tests are unchanged, the starting state is the only difference from the source benchmark. Across five frontier agents, success drops by 20.6--53.5 percentage points on DeepSWE and 20.6--34.9 points on Terminal-Bench relative to a Clean Start, and agent rankings on DeepSWE reverse. Failed continuations show a pattern we call acceptance bias: agents treat the inherited work as the definition of what remains to be done instead of checking it against the request. On DeepSWE, in 58.9--88.2% of failed runs, they submit with exactly the same tests failing, although they pass most of these tests from a Clean Start. SWE-Residual thus exposes a capability gap that clean-start evaluation does not capture.