# Reinforcement learning optimization

Polaris is an open-source post-training method that uses reinforcement learning to optimize and enhance the model and improve inference capabilities.

The LongWriter-Zero-32B quantized model is based on the THU-KEG/LongWriter-Zero-32B base model, supports both Chinese and English, and is suitable for long context scenarios such as reinforcement learning and writing.

The LongWriter-Zero-32B Quantized Model is a multilingual model that undergoes static quantization based on the original model. It is suitable for long context scenarios such as reinforcement learning and writing.

A high-performance 7B parameter language model launched by NVIDIA, focusing on mathematical and code reasoning tasks and supporting a 128k context length.

Kimi-Dev-72B is an open-source large coding language model for software engineering tasks, achieving the best results among open-source models on SWE-bench Verified.

ContentV is an efficient video generation model framework that achieves high-quality video generation with limited computing resources through a minimalist architecture, multi-stage training strategy, and cost-effective reinforcement learning framework.

QwenLong-L1 is a long-context reasoning model trained with reinforcement learning, demonstrating excellent performance across seven long-context document QA benchmarks.

VeriReason is a Verilog RTL code generation method that combines reinforcement learning with testbench feedback, significantly improving the performance of pre-trained models in the field of hardware design.

INTELLECT 2 is a large language model launched by PrimeIntellect, supporting a context length of 40960 tokens, trained using the QwQ architecture and GRPO reinforcement learning framework.

This is an experimental financial analysis model optimized for financial fundamentals prediction through the Atropos reinforcement learning framework

Fine-tuned based on the Qwen/Qwen2.5-1.5B-Instruct model, using the TinyV reward system, which can provide more accurate reward signals in the post-training of efficient reinforcement learning (RL) and significantly improve RL efficiency and the performance of the final model.

The Camel Model is a text generation model based on the transformer architecture, supporting Azerbaijani and trained using reinforcement learning.

A pre-trained language model merged from multiple Captain-Eris series models using the mergekit tool

STILL-3-1.5B-preview is a slow-thinking model enhanced with reinforcement learning technology, achieving 39.33% accuracy on the AIME benchmark

CodeT5 is a large-scale encoder-decoder model pre-trained with NTP objectives for Python language, focusing on code understanding and generation tasks

This is a reinforcement learning agent based on the PPO algorithm, specifically designed for training and evaluation in the BreakoutNoFrameskip-v4 game environment.