Share your thoughts, 1 month free Claude Pro on usSee more

Mathematical Reasoning on AIME24 (Pass@k)

63.3Pass@k

Qwen2.5-7B + GRPO w/ VERL.

Updated 3mo ago

Evaluation Results

Method	Links
Qwen2.5-7B + GRPO w/ VERL. 2025.09		63.3
Qwen2.5-7B 2025.09		60
Qwen2.5-3B 2025.09		56.7
Qwen2.5-7B + GRPO 2025.09		56.7
Qwen2.5-7B + PPO w/ VERL. 2025.09		56.7
Llama-3.1-8B-Instruct + GRPO 2025.09		53.3
Llama-3.1-8B-Instruct + PPO w/ VERL. 2025.09		53.3
Qwen2.5-3B + GRPO w/ VERL. 2025.09		53.3
Qwen2.5-3B + PPO w/ VERL. 2025.09		53.3
Qwen2.5-7B + PPO 2025.09		53.3
Mathstral-7B-v0.1 + PPO 2025.09		53.3
Mathstral-7B-v0.1 + PPO w/ VERL. 2025.09		53.3
Llama-3.2-3B-Instruct + GRPO w/ VERL. 2025.09		50
Llama-3.1-8B-Instruct + GRPO w/ VERL. 2025.09		50
Mathstral-7B-v0.1 + GRPO w/ VERL. 2025.09		50
Llama-3.2-3B-Instruct + PPO 2025.09		46.7
Llama-3.2-3B-Instruct + PPO w/ VERL. 2025.09		46.7
Llama-3.1-8B-Instruct 2025.09		46.7
Llama-3.1-8B-Instruct + PPO 2025.09		46.7
Qwen2.5-3B + GRPO 2025.09		46.7
Qwen2.5-3B + PPO 2025.09		43.3
Mathstral-7B-v0.1 2025.09		43.3
Llama-3.2-3B-Instruct 2025.09		40
Llama-3.2-3B-Instruct + GRPO 2025.09		40
Mathstral-7B-v0.1 + GRPO 2025.09		36.7
Mistral-7B-v0.3 2025.09		20
Mistral-7B-v0.3 + GRPO w/ VERL. 2025.09		16.7
Mistral-7B-v0.3 + GRPO 2025.09		10
Mistral-7B-v0.3 + PPO 2025.09		6.7
Mistral-7B-v0.3 + PPO w/ VERL. 2025.09		3.3