Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Premise Selection on Premise Selection (val)
Loading...
26.4
Recall
GPT-5.5
3.104
9.152
15.2
21.248
Jun 24, 2026
Recall
Mean Retrieved Candidates
Updated 1mo ago
Evaluation Results
Method
Method
Links
Recall
Mean Retrieved Candidates
GPT-5.5
k=10, H=6
2026.06
26.4
60
Qwen3-8B SDPO + same-paper graph augmentation
k=10, H=6, Training St...
2026.06
22.7
140
Gemini 3.1 Pro
k=10, H=6
2026.06
17
60
Qwen3-8B SDPO + cross-paper graph augmentation
k=10, H=6, Training St...
2026.06
13.4
115
Qwen3-8B SDPO
k=10, H=6, Training St...
2026.06
7.9
60
Qwen3-8B base
k=10, H=6, Training St...
2026.06
4
60
Feedback
Search any
task
Search any
task