Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Mathematical Reasoning on AIME24, AIME25, AMC23, MATH, Olympiad (Aggregate)
Loading...
47.3
Average Score
GRPO + coupled rhythm credit
36.692
39.446
42.2
44.954
Oct 15, 2025
Average Score
Updated 1mo ago
Evaluation Results
Method
Method
Links
Average Score
GRPO + coupled rhythm credit
Backbone=Qwen3-4B-Base...
2025.10
47.3
GRPO + local-chunk credit
Backbone=Qwen3-4B-Base...
2025.10
47.2
GRPO + global-anchor credit
Backbone=Qwen3-4B-Base...
2025.10
47.1
GRPO
Backbone=Qwen3-4B-Base...
2025.10
44.8
GRPO + random credit
Backbone=Qwen3-4B-Base...
2025.10
44.7
GRPO + high-entropy credit
Backbone=Qwen3-4B-Base...
2025.10
44.5
GRPO + coupled rhythm credit
Backbone=Qwen3-8B-Base...
2025.10
43.2
GRPO + global-anchor credit
Backbone=Qwen3-8B-Base...
2025.10
42.1
GRPO + local-chunk credit
Backbone=Qwen3-8B-Base...
2025.10
41.3
GRPO + high-entropy credit
Backbone=Qwen3-8B-Base...
2025.10
40.1
GRPO + random credit
Backbone=Qwen3-8B-Base...
2025.10
39.6
GRPO
Backbone=Qwen3-8B-Base...
2025.10
39.4
GRPO + coupled rhythm credit
Backbone=Qwen3-4B-Base...
2025.10
39.2
GRPO + local-chunk credit
Backbone=Qwen3-4B-Base...
2025.10
38.6
GRPO + global-anchor credit
Backbone=Qwen3-4B-Base...
2025.10
38.2
GRPO + high-entropy credit
Backbone=Qwen3-4B-Base...
2025.10
37.2
GRPO
Backbone=Qwen3-4B-Base...
2025.10
37.1
GRPO + random credit
Backbone=Qwen3-4B-Base...
2025.10
37.1
Feedback
Search any
task
Search any
task