List your company
ResearchFormerly WizWand

Mathematical Reasoning on MATH500 (Pass@k)

91.4Pass@kQwen2.5-7B + GRPO w/ VERL.
16.31235.80655.374.794Sep 28, 2025
Updated 6mo ago

Evaluation results

MethodLinks
Qwen2.5-7B + GRPO w/ VERL.k=16, Base Model=Qwen2.5-7B, RL Method=GRPO, Framework=VERL2025.0991.4
Qwen2.5-7B + PPO w/ VERL.k=16, Base Model=Qwen2.5-7B, RL Method=PPO, Framework=VERL2025.0991.4
Qwen2.5-7B + PPOk=16, Base Model=Qwen2.5-7B, RL Method=PPO2025.0991.2
Qwen2.5-7B + GRPOk=16, Base Model=Qwen2.5-7B, RL Method=GRPO2025.0990.8
Qwen2.5-7Bk=16, Base Model=Qwen2.5-7B2025.0990.6
Qwen2.5-3B + PPO w/ VERL.k=16, Base Model=Qwen2.5-3B, RL Method=PPO, Framework=VERL2025.0988.2
Qwen2.5-3B + PPOk=16, Base Model=Qwen2.5-3B, RL Method=PPO2025.0987.8
Qwen2.5-3B + GRPO w/ VERL.k=16, Base Model=Qwen2.5-3B, RL Method=GRPO, Framework=VERL2025.0987.6
Mathstral-7B-v0.1 + GRPO w/ VERL.k=16, Base Model=Mathstral-7B-v0.1, RL Method=GRPO, Framework=VERL2025.0987
Qwen2.5-3B + GRPOk=16, Base Model=Qwen2.5-3B, RL Method=GRPO2025.0986.6
Qwen2.5-3Bk=16, Base Model=Qwen2.5-3B2025.0986
Mathstral-7B-v0.1 + GRPOk=16, Base Model=Mathstral-7B-v0.1, RL Method=GRPO2025.0984.8
Mathstral-7B-v0.1 + PPO w/ VERL.k=16, Base Model=Mathstral-7B-v0.1, RL Method=PPO, Framework=VERL2025.0984.8
Llama-3.1-8B-Instruct + GRPOk=16, Base Model=Llama-3.1-8B-Instruct, RL Method=GRPO2025.0983.4
Llama-3.1-8B-Instruct + GRPO w/ VERL.k=16, Base Model=Llama-3.1-8B-Instruct, RL Method=GRPO, Framework=VERL2025.0983.4
Llama-3.2-3B-Instruct + PPO w/ VERL.k=16, Base Model=Llama-3.2-3B-Instruct, RL Method=PPO, Framework=VERL2025.0982.4
Llama-3.1-8B-Instruct + PPO w/ VERL.k=16, Base Model=Llama-3.1-8B-Instruct, RL Method=PPO, Framework=VERL2025.0982.4
Mathstral-7B-v0.1 + PPOk=16, Base Model=Mathstral-7B-v0.1, RL Method=PPO2025.0982.4
Llama-3.2-3B-Instruct + PPOk=16, Base Model=Llama-3.2-3B-Instruct, RL Method=PPO2025.0982.2
Llama-3.2-3B-Instruct + GRPO w/ VERL.k=16, Base Model=Llama-3.2-3B-Instruct, RL Method=GRPO, Framework=VERL2025.0980.6
Mathstral-7B-v0.1k=16, Base Model=Mathstral-7B-v0.12025.0980.4
Llama-3.2-3B-Instruct + GRPOk=16, Base Model=Llama-3.2-3B-Instruct, RL Method=GRPO2025.0980.2
Llama-3.2-3B-Instructk=16, Base Model=Llama-3.2-3B-Instruct2025.0979.8
Llama-3.1-8B-Instructk=16, Base Model=Llama-3.1-8B-Instruct2025.0979.8
Llama-3.1-8B-Instruct + PPOk=16, Base Model=Llama-3.1-8B-Instruct, RL Method=PPO2025.0979.2
Mistral-7B-v0.3k=16, Base Model=Mistral-7B-v0.32025.0936
Mistral-7B-v0.3 + GRPO w/ VERL.k=16, Base Model=Mistral-7B-v0.3, RL Method=GRPO, Framework=VERL2025.0934.4
Mistral-7B-v0.3 + GRPOk=16, Base Model=Mistral-7B-v0.3, RL Method=GRPO2025.0933
Mistral-7B-v0.3 + PPOk=16, Base Model=Mistral-7B-v0.3, RL Method=PPO2025.0921.8
Mistral-7B-v0.3 + PPO w/ VERL.k=16, Base Model=Mistral-7B-v0.3, RL Method=PPO, Framework=VERL2025.0919.2