Skip to content

πŸ™Œ OpenHands: AI-Driven Development

eval: remove gemini-specific swebench template (#9623)

Xingyao Wangcff569745689
diff --git a/evaluation/benchmarks/swe_bench/run_infer.py b/evaluation/benchmarks/swe_bench/run_infer.py
index ca9dd52ec3..30a40c8e66 100644
--- a/evaluation/benchmarks/swe_bench/run_infer.py
+++ b/evaluation/benchmarks/swe_bench/run_infer.py
@@ -109,9 +109,7 @@ def get_instruction(instance: pd.Series, metadata: EvalMetadata) -> MessageActio
template_name = 'swt.j2'
elif mode == 'swe':
if 'claude' in llm_model:
- template_name = 'swe_claude.j2'
- elif 'gemini' in llm_model:
- template_name = 'swe_gemini.j2'
+ template_name = 'swe_default.j2'
elif 'gpt-4.1' in llm_model:
template_name = 'swe_gpt4.j2'
else: