Sudhanshu1985

slm-125m-rlaif

README

License: apache-2.0

slm-125m-rlaif

RLAIF (Reinforcement Learning from AI Feedback) alignment of Sudhanshu1985/slm-125m-sft. A scalar reward model (Bradley-Terry on gpt-4.1-mini preference pairs) scores on-policy samples; the policy is optimized with RLOO policy gradient + KL to the frozen SFT. Mean reward -0.158 -> 0.155. This is the RL-based counterpart to Sudhanshu1985/slm-125m-dpo. Uses the SFT chat template: <|bos|><|system|>SYS<|user|>USER<|assistant|>ANSWER<|eos|>.

Available on FriendliAI

Dedicated Endpoints

Run this model inference on single tenant GPU with unmatched speed and reliability at scale.

Learn more

Container

Run this model inference with full control and performance in your environment.

Learn more

Model Details

Model Provider

Sudhanshu1985

Model Tree

Base