Files
SWE-bench_Verified/eval.yaml
T
Ofir PressandNiels Rogge 880e32832a Add eval.yaml (#2)
- Add eval.yaml (8db12b107a1abfd10808f023ed233890b4d7cbae)


Co-authored-by: Niels Rogge <nielsr@users.noreply.huggingface.co>
2026-02-23 17:42:36 +00:00

11 lines
532 B
YAML

name: SWE-Bench Verified
description: SWE-bench Verified is a curated, human-annotated subset of 500 issues from the original SWE-bench, designed to accurately evaluate AI models on real-world software engineering tasks. It fixes issues in the original dataset by ensuring all tasks are solvable, well-defined, and properly tested by human developers.
evaluation_framework: swe-bench-verified
tasks:
- id: swe_bench_%_resolved
config: default
split: test
- id: swe_bench_average_cost
config: default
split: test