Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Mathematical Reasoning on AMC 23 (Accuracy, Avg)
Loading...
65.4
Accuracy
GRPO + coupled rhythm credit
58.12
60.01
61.9
63.79
Oct 15, 2025
Accuracy
Average Score (AMC 23)
Updated 1mo ago
Evaluation Results
Method
Method
Links
Accuracy
Average Score (AMC 23)
GRPO + coupled rhythm credit
Base Model=Qwen3-8B-Ba...
2025.10
65.4
43.2
GRPO + global-anchor credit
Base Model=Qwen3-8B-Ba...
2025.10
63.1
42.1
GRPO + local-chunk credit
Base Model=Qwen3-8B-Ba...
2025.10
62.9
41.3
GRPO + Reweight+Lopti
Base Model=Qwen3-8B-Ba...
2025.10
62.5
41.1
GRPO + CAPO
Base Model=Qwen3-8B-Ba...
2025.10
61.1
41.4
GRPO + ThinkPRM-1.5B
Base Model=Qwen3-8B-Ba...
2025.10
61
41
GRPO + high-entropy selection
Base Model=Qwen3-8B-Ba...
2025.10
60.5
40.9
GRPO + high-entropy credit
Base Model=Qwen3-8B-Ba...
2025.10
60.4
40.1
GRPO + random credit
Base Model=Qwen3-8B-Ba...
2025.10
60.1
39.6
GRPO + AsyPPO
Base Model=Qwen3-8B-Ba...
2025.10
59.5
40
GRPO + path-aggregation credit
Base Model=Qwen3-8B-Ba...
2025.10
59.5
40
GRPO
Base Model=Qwen3-8B-Ba...
2025.10
59.1
39.4
GRPO + token-correlation credit
Base Model=Qwen3-8B-Ba...
2025.10
59
37.4
GRPO + gradient-based credit
Base Model=Qwen3-8B-Ba...
2025.10
58.4
40
Feedback
Search any
task
Search any
task