This model is a text generation model fine-tuned based on Meta's Llama-3.1 architecture. It is trained using reinforcement learning with the TRL library and the GRPO (Group Relative Policy Optimization) method, and is specifically optimized for argument generation tasks.

Natural Language ProcessingTransformers

Emu3.5 is a native multimodal model developed by the Beijing Academy of Artificial Intelligence (BAAI). It can jointly predict the next state across vision and language, enabling coherent world modeling and generation. After end-to-end pre-training and large-scale reinforcement learning post-training, it demonstrates excellent performance in multimodal tasks.

MultimodalTransformers

Kimi Linear is a hybrid linear attention architecture that outperforms traditional full attention methods in various scenarios, including short and long contexts, as well as reinforcement learning extension mechanisms. It can effectively solve the problem of low efficiency of traditional attention mechanisms in long context tasks and bring more efficient solutions to fields such as natural language processing.

Natural Language ProcessingTransformers

Kimi Linear is an efficient hybrid linear attention architecture that outperforms traditional full attention methods in short context, long context, and reinforcement learning scenarios. It optimizes attention computation through the Kimi Delta Attention (KDA) mechanism, significantly improving performance and hardware efficiency, and is particularly good at handling long context tasks up to 1 million tokens.

Natural Language ProcessingTransformers

PokeeResearch - 7B is a 7 - billion - parameter in - depth research agent model developed by Pokee AI. It combines Reinforcement Learning from AI Feedback (RLAIF) and a powerful reasoning framework, enabling reliable, aligned, and scalable research - level reasoning in tool - enhanced large language models. It is suitable for complex multi - step research workflows.

Natural Language ProcessingTransformersEnglish

Granite-4.0-1B is a lightweight instruction model developed by IBM, fine-tuned based on Granite-4.0-1B-Base. This model combines open-source instruction datasets and internal synthetic datasets, and is developed using techniques such as supervised fine-tuning, reinforcement learning, and model merging. It is suitable for device-side deployment and research use cases.

Natural Language ProcessingTransformers

Granite-4.0-350M is a lightweight instruction model developed by IBM, fine-tuned based on Granite-4.0-350M-Base. This model combines open-source instruction datasets and internal synthetic datasets, and is developed using technologies such as supervised fine-tuning, reinforcement learning, and model fusion. It has powerful instruction-following capabilities and is particularly suitable for device-side deployment and research scenarios.

Natural Language ProcessingTransformers

Nanbeige4-3B-Thinking is a 3-billion parameter inference model in the fourth-generation Nanbeige large language model family. It achieves advanced inference capabilities by improving data quality and training methods. This model performs excellently in multiple fields such as mathematics, science, creative writing, and tool usage, and supports multi-stage curriculum learning and reinforcement learning training.

Natural Language ProcessingTransformersMultiple Languages

OLGA is an online reinforcement learning positioning agent built on Qwen3-VL-30B-A3B-Instruct, using a mixture-of-experts model with 3.3 billion active parameters. It is trained through a new data recipe that combines existing datasets, new data collection, automatic filtering, and online reinforcement learning, achieving advanced positioning performance among open-source models.

MultimodalTransformersEnglish

PokeeResearch-7B is a 7-billion-parameter deep research intelligent agent developed by Pokee AI. It combines reinforcement learning based on AI feedback (RLAIF) with an inference framework and can execute complex multi-step research workflows, including self-correction, verification, and comprehensive analysis.

Natural Language ProcessingTransformersEnglish

Qwen3-Nemotron-32B-RLBFF is a large language model fine-tuned based on Qwen/Qwen3-32B. The quality of the model's generated responses in the default thinking mode has been significantly improved through reinforcement learning feedback technology. This model performs excellently in multiple benchmark tests while maintaining low inference costs.

Natural Language ProcessingTransformersEnglish

KAT-Dev-72B-Exp is an open-source large language model with 72 billion parameters for software engineering tasks, achieving an accuracy of 74.6% in the SWE-Bench Verified benchmark test. It is an experimental reinforcement learning version of the KAT-Coder model.

Natural Language ProcessingTransformers

Granite-4.0-350M is a lightweight instruction model developed by IBM, fine-tuned based on Granite-4.0-350M-Base. This model combines open-source instruction datasets and internal synthetic datasets, and is developed using supervised fine-tuning, reinforcement learning, and model merging techniques. It has powerful instruction-following capabilities and tool invocation functions.

Natural Language ProcessingTransformers

Granite-4.0-1B is a lightweight instruction model developed by IBM. It is fine-tuned based on Granite-4.0-1B-Base, combining open-source instruction datasets and internal synthetic datasets, and developed using supervised fine-tuning, reinforcement learning, and model merging techniques.

Natural Language ProcessingTransformers

Granite-4.0-H-350M is a lightweight instruction model developed by IBM, fine-tuned based on Granite-4.0-H-350M-Base. This model combines open-source instruction datasets and internal synthetic datasets, and is developed using various technologies such as supervised fine-tuning, reinforcement learning, and model merging. It has powerful instruction-following capabilities and multilingual support.

Natural Language ProcessingTransformers

Granite-4.0-H-1B is a lightweight instruction model developed by the IBM Granite team. It is optimized based on the Granite-4.0-H-1B-Base model through technologies such as supervised fine-tuning, reinforcement learning, and model fusion. This model has a strong ability to follow instructions and is suitable for device-side deployment and research use cases.

Natural Language ProcessingTransformers

Tongyi Deep Research 30B is a large language model with 30 billion parameters, designed specifically for long-cycle, deep information search tasks. This model performs excellently in multiple intelligent search benchmark tests, uses innovative quantization methods to improve performance, and supports intelligent pre-training, supervised fine-tuning, and reinforcement learning.

Natural Language ProcessingTransformersEnglish

Qwen3-4B-SafeRL is a safety-aligned version of the Qwen3-4B model. It enhances the model's robustness against harmful or adversarial prompts through reinforcement learning training. This version uses a hybrid reward function for optimization, balancing three objectives: safety, usefulness, and minimizing unnecessary rejections.

Natural Language ProcessingTransformersEnglish

This project distills the inference ability of GPT into the Llama-3.1-8B model through an innovative two-stage training process. First, knowledge distillation and format alignment are performed through supervised fine-tuning, and then reinforcement learning is used to encourage the model to independently explore and optimize inference strategies, focusing on breakthroughs in the field of mathematical reasoning.

Natural Language ProcessingTransformersMultiple Languages

G²RPO is a novel reinforcement learning framework specifically designed for preference alignment in flow models, significantly improving the generation quality through a granular reward evaluation mechanism.

MultimodalDiffusers