Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Multi-token generation on HMathA
Loading...
60.4
Accuracy
ICL
19.424
30.062
40.7
51.338
Sep 26, 2025
Accuracy
Updated 1mo ago
Evaluation Results
Method
Method
Links
Accuracy
ICL
Backbone=Qwen3-4B-Base...
2025.09
60.4
IA2 → SFT
Backbone=Qwen3-4B-Base...
2025.09
55.3
SFT only
Backbone=Qwen3-4B-Base...
2025.09
50.4
IA2 only
Backbone=Qwen3-4B-Base...
2025.09
47.8
w/o ICL
Backbone=Qwen3-4B-Base...
2025.09
21
Feedback
Search any
task
Search any
task