Model Card for Model ID

PPO-C (PPO with Calibrated Reward Calculation) is an RLHF algorithm to mitigate verbalized overconfidence in RLHF-trained Large Language Models. PPO-C adjusts standard reward model scores during PPO training. It maintains a running average of past reward scores as a dynamic threshold to classify responses, and adjusts the reward scores based on model expressed verbalized confidence. Please refer to our preprint (Taming Overconfidence in LLMs: Reward Calibration in RLHF) and repo for more details.

Model Details

Model Description

We train OpenRLHF/Llama-3-8b-sft-mixture on our HINT-lab/prompt-collections-final-v0.3 with a vanilla reward model OpenRLHF/Llama-3-8b-rm-mixture.

• Developed by: Jixuan Leng, Chengsong Huang, Banghua Zhu, Jiaxin Huang
• Finetuned from model: OpenRLHF/Llama-3-8b-sft-mixture

Model Sources [optional]

• Repository: Our repo
• Paper: Taming Overconfidence in LLMs: Reward Calibration in RLHF