Artanic30
/

NoisyGRPO-3B

Reinforcement Learning

Model card Files Files and versions

This is the official checkpoint released for the NeurIPS 2025 paper NoisyGRPO.

For model usage, please follow the instructions in qwen2.5-vl

References

Model Paper

Downloads last month: 5

Safetensors

Model size

4B params

Tensor type

BF16

·

Video Preview

Reinforcement Learning

loading

Model tree for Artanic30/NoisyGRPO-3B

Base model

Qwen/Qwen2.5-VL-3B-Instruct

Finetuned

(694)

this model

Quantizations

Dataset used to train Artanic30/NoisyGRPO-3B

Collection including Artanic30/NoisyGRPO-3B

NoisyGRPO

This is the collection for the NeurIPS paper NoisyGRPO. Project Page: https://artanic30.github.io/project_pages/NoisyGRPO/ • 2 items • Updated Nov 28, 2025 • 2

Paper for Artanic30/NoisyGRPO-3B

NoisyGRPO: Incentivizing Multimodal CoT Reasoning via Noise Injection and Bayesian Estimation

Paper • 2510.21122 • Published Oct 24, 2025