rl-rag/

open-scholar-gpt-oss-120b

Deep research agent evaluation on data/drtulu_open_scholar.jsonl (normal split).

	Results




	

Metric Value

pass@1 0.0%

avg@1 0.0%

Trajectory accuracy 0.0% (0/11854)

Questions 11854

Trajectories 11854 (1 per question)

Avg tool calls 17.3

Full conversations ✅

	Model & Setup




	




	

Model gpt-oss-120b

Judge gpt-4o

Max tool calls 50

Temperature 0.7

Blocked domains None