# RLHF fine-tuning

Llama 3.2 is a multilingual large language model launched by Meta, offering two parameter scales of 1B and 3B, supporting various language tasks, and performing better than existing open-source and closed-source models.

The Violet series language models are specifically trained for instruction following and maintaining Russian dialogues. The fourth generation has been optimized with RLHF, offering stronger response capabilities and richer information content.

Starling-7B is an open-source large language model trained through Reinforcement Learning from AI Feedback (RLAIF), fine-tuned based on Openchat 3.5, and performs excellently in MT Bench.

A causal language model based on Pythia-2.8b, fine-tuned using the Anthropic human preference dataset

Llama 2 is an open-source large language model series developed by Meta, ranging from 7B to 70B parameters, supporting English text generation tasks.

StableVicuna-13B is a dialogue model fine-tuned via RLHF based on Vicuna-13B v0, using 4-bit GPTQ quantization format

StableVicuna-13B is a fine-tuned version of the Vicuna-13B v0 model, enhanced through Reinforcement Learning from Human Feedback (RLHF) and Proximal Policy Optimization (PPO) on various dialogue and instruction datasets.

A dialogue generation model based on OPT-1.3b, optimized through RLHF training using the DeepSpeed-Chat framework