The authors designed SWE-Stream to address the bottleneck of obtaining high-quality executable tasks for training SWE agents.
Scaling executable training data for software engineering agents with progressive construction and elastic streaming
From 10,728 repositories across six languages, SWE-Stream constructs 80,874 validated tasks and achieves strong resolve rates on SWE-bench variants.
Research Lab
Chengyang Liu · Wen Yang · Shifeng Chen · Delai Qiu · Ji Yang · Junnan Zhu · +2 more
Xiamen University · Institute of Automation, Chinese Academy of Sciences · Unisound
Research Digest··2 min read
The authors present SWE-Stream, a system that scales the construction of executable training data for software engineering (SWE) agents by progressively validating and refining task candidates from development histories.
Why this paper
From Institute of Automation, Chinese Academy of Sciences and 2 others
In one line
SWE-Stream constructs 80,874 executable SWE tasks from 10,728 repositories and achieves 69.0% on SWE-bench Verified.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ✓Reports numbers on named benchmarks (3 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§