Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
General Question Answering on NQ (Natural Questions) (OSR%)
Loading...
25.2
OSR (%)
StepSearch-base
-1.008
5.796
12.6
19.404
Apr 19, 2026
OSR (%)
Updated 1mo ago
Evaluation Results
Method
Method
Links
OSR (%)
StepSearch-base
Backbone=Qwen2.5-3B
2026.04
25.2
StepSearch-instruct
Backbone=Qwen2.5-7B
2026.04
20.8
Search-R1-base
Backbone=Qwen2.5-7B
2026.04
13
StepSearch-base
Backbone=Qwen2.5-7B
2026.04
12.2
Search-R1-instruct
Backbone=Qwen2.5-7B
2026.04
9.3
Search-R1-instruct
Backbone=Qwen2.5-3B
2026.04
8.7
StepSearch-instruct
Backbone=Qwen2.5-3B
2026.04
5.7
HIPRAG-instruct
Backbone=Qwen2.5-7B
2026.04
4.5
Search-R1-base
Backbone=Qwen2.5-3B
2026.04
3.5
HIPRAG-base
Backbone=Qwen2.5-3B
2026.04
2.7
AutoSearch
Backbone=Qwen2.5-7B
2026.04
1.83
HIPRAG-instruct
Backbone=Qwen2.5-3B
2026.04
0.7
AutoSearch
Backbone=Qwen2.5-3B
2026.04
0
Feedback
Search any
task
Search any
task