Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Deep search QA on Frames
Loading...
46.42
Accuracy
ProCeedRL
15.8232
23.7666
31.71
39.6534
Apr 2, 2026
Apr 16, 2026
May 1, 2026
May 16, 2026
May 31, 2026
Jun 15, 2026
Jun 30, 2026
Accuracy
Updated 25d ago
Evaluation Results
Method
Method
Links
Accuracy
ProCeedRL
Backbone=Qwen3-8B
2026.04
46.42
DAPO/Search-R1
Backbone=Qwen3-8B
2026.04
43.59
RFT
Backbone=Qwen3-8B
2026.04
40.65
Rewinding More
Backbone=Qwen3-8B, Typ...
2026.04
40.13
ECHO
Backbone=Qwen3-32B-Ins...
2026.06
39.1
Qwen3-8B-v3-SFT
Backbone=Qwen3-8B
2026.04
37.5
ECHO w/ Top-K retrieval
Backbone=Qwen3-32B-Ins...
2026.06
37.3
ECHO w/o Traceable CA & Turn-level IS
Backbone=Qwen3-32B-Ins...
2026.06
32.2
ReAct Prompting
Backbone=Qwen3-8B
2026.04
31.92
ECHO w/ Top-K retrieval & w/o turn summary
Backbone=Qwen3-32B-Ins...
2026.06
31.3
ECHO w/o traceable CA
Backbone=Qwen3-32B-Ins...
2026.06
30.8
SUPO
Backbone=Qwen3-32B-Ins...
2026.06
26.8
ECHO
Backbone=Qwen3-30B-A3B...
2026.06
25
GRPO
Backbone=Qwen3-32B-Ins...
2026.06
24.8
GRPO
Backbone=Qwen3-30B-A3B...
2026.06
19.1
SUPO
Backbone=Qwen3-30B-A3B...
2026.06
17
Feedback
Search any
task
Search any
task