Open-source alternatives to AReaL

Search-R1 is a distributed training system and reinforcement learning framework designed to create search-augmented language models. It provides an architecture for scaling model workloads across head and worker nodes while optimizing how models interleave internal reasoning with external tool calls. The system focuses on refining model behavior through custom reward signals and reinforcement learning to improve tool-use formatting and information retrieval. It implements an interleaved reasoning-search loop that allows models to alternate between internal thought generation and external data

This project is an alignment framework and suite of pipelines for training language models using supervised fine-tuning and preference optimization. It provides tools for executing large-scale distributed training across multiple GPUs and compute nodes, alongside a system for measuring model helpfulness and dialogue quality through single-turn and multi-turn benchmarks. The framework includes specialized tools for direct preference optimization to refine model behavior using paired data without a separate reward model. It also supports constitutional AI alignment and the training of reward mo

EasyR1 is a distributed model training system and reinforcement learning framework for large language and vision-language models. It functions as a multimodal trainer and an implementation of a Proximal Policy Optimization pipeline designed to refine the reasoning and perception capabilities of models that process both text and images. The system specializes in distributing reinforcement learning workloads across multiple compute nodes to manage high memory requirements. It optimizes hardware utilization through padding-free training and fine-tuning to fit large models onto available graphics

SLIME is a distributed reinforcement learning framework for large language model post-training that bridges Megatron training with SGLang inference servers. It orchestrates scalable RL loops across GPU clusters, decoupling training and inference into independent processes that communicate over HTTP and NCCL for independent scaling and fault tolerance. The system supports multi-agent reinforcement learning workflows with parallel agent instances, customizable rollout strategies, and personalized agent serving that improves models from prior conversations without disrupting API serving. The fra

This is a PyTorch implementation of reinforcement learning from human feedback designed to align large language models with human values and preferences. It provides a framework for the PaLM architecture and incorporates parameter-efficient fine-tuning to adapt models while minimizing the number of updated weights. The system enables the development of reward models that act as scoring mechanisms built from human preference data. These models evaluate generative outputs to guide the alignment process. The workflow covers policy optimization using a clipped objective, reward modeling based on

ChatGLM-Efficient-Tuning is a fine-tuning framework and toolkit designed to optimize large language models using parameter-efficient fine-tuning techniques. It provides a pipeline for adjusting model behavior and reducing the memory and compute requirements necessary for training. The project features a web-based trainer and orchestration interface for configuring and executing the fine-tuning process on a single GPU. It supports quantized training in lower precision formats to enable fine-tuning on hardware with limited memory, as well as reinforcement learning from human feedback for model

rllm is an asynchronous reinforcement learning framework for training language agents. It provides a unified pipeline that runs the same agent code for both evaluation and training, automatically capturing traces for gradient computation. The framework supports distributed reinforcement learning across multiple GPUs and nodes using pluggable backends, and executes agents in isolated sandboxes—either locally or in the cloud—for safe and scalable rollout collection. It trains agents built with LangGraph, SmolAgents, OpenAI Agents SDK, or custom frameworks without requiring core logic changes. T

This project is a fine-tuning framework and training pipeline designed to optimize and adapt large language and vision models. It provides a specialized toolkit for parameter-efficient tuning and supervised learning, serving as both a trainer for multimodal models and a deployment tool for serving fine-tuned models via high-performance inference engines. The framework focuses on reducing memory and compute requirements by updating a small subset of model parameters. It supports a wide range of adaptation strategies, including vision-language model training to align text, image, video, and aud

TinyZero is a reinforcement learning framework and implementation designed to train language models to develop reasoning and self-verification abilities. It provides a training pipeline to optimize model performance on mathematical and logical tasks. The project serves as a minimal reproduction of the DeepSeek R1 architectural and training approach. It focuses on creating reasoning models that can solve structured problems through autonomous chain-of-thought discovery. The framework incorporates group relative policy optimization and reward-based self-correction to improve accuracy on logica

Agent Lightning is an optimization framework designed to refine the performance of individual AI agents within complex multi-agent systems. It provides a platform for improving decision-making and task execution by applying reinforcement learning, supervised fine-tuning, and automated prompt optimization. The framework distinguishes itself through its ability to isolate specific agents for targeted tuning, allowing developers to enhance individual behaviors while maintaining the stability of the broader system architecture. By utilizing a modular interface, it integrates with diverse agent fr

Open-Instruct is a distributed training and instruction tuning framework for large language models. It functions as a coordinator for supervised fine-tuning, reinforcement learning from human feedback pipelines, and tool-use training, providing specialized roles for dataset curation and model alignment. The project distinguishes itself through a high-performance training architecture that utilizes actor-based distributed coordination and hybrid sharding to manage large GPU clusters. It implements advanced alignment techniques including direct preference optimization, group relative policy opt

OpenClaw-RL is a reinforcement learning framework for training large language model agents. It provides a system for optimizing agent policies using natural language instructions, reward-based optimization, and a pipeline for refining model behavior through human or judge feedback. The framework utilizes low-rank adaptation to update model weights efficiently without full parameter tuning. It features a local-first infrastructure that allows the policy model, judge, and trainer to run on private hardware to maintain data locality. The system covers a range of optimization capabilities, inclu

Open CLIP is an open source framework for training and deploying Contrastive Language-Image Pre-training models. It serves as a vision-language training framework and multimodal embedding engine that maps images and text into a shared vector space for similarity searches and zero-shot classification. The project provides a toolkit for distributed training of contrastive models and includes an image-to-text generative model for producing natural language descriptions. It supports custom text encoder integration and utilizes teacher-student model distillation to transfer knowledge from large pr

trlx is a reinforcement learning library and training framework designed to align large language models using human feedback. It serves as a distributed trainer and compute orchestrator for scaling high-parameter models across multiple GPUs and nodes. The project provides tools for reinforcement learning from human feedback and model alignment. It implements reward-model-based optimization and proximal policy optimization to refine model behavior based on goal-oriented rewards or human-labeled datasets. The framework covers distributed training strategies, including model parallelism, parame

zero_nlp is a distributed framework for training and fine-tuning large language models and multimodal architectures. It provides a specialized toolkit for distributed model parallelism, allowing neural network layers and weights to be partitioned across multiple GPU devices to train models that exceed the memory capacity of a single processor. The project distinguishes itself through a combination of high-throughput data pipelines and parameter-efficient tuning. It utilizes multi-threading and memory mapping to preprocess and stream datasets exceeding 100GB and implements memory-saving adapta

Lit-llama is a PyTorch-based implementation framework for the LLaMA language model, providing a system for pre-training, fine-tuning, and high-performance inference. It includes a pre-training pipeline for creating foundational language models from scratch and tools for running pretrained weights to generate natural text and predict sequences. The project provides specialized toolkits for parameter-efficient fine-tuning using low-rank adaptation and lightweight adapters. It also includes a quantization library that reduces model memory footprints through four-bit and eight-bit precision to en

Axolotl is a distributed training orchestrator and fine-tuning framework for large language models, multimodal systems, and quantized models. It provides a structured environment for specializing pre-trained models through full parameter updates or low-rank adaptation, as well as aligning model outputs with human expectations via preference tuning pipelines and reward modeling. The system distinguishes itself through a configuration-driven pipeline that manages preprocessing and training workflows via a single file for reproducibility. It implements high-throughput optimizations such as multi

Tinker Cookbook is an open-source framework for fine-tuning large language models, supporting supervised learning, reinforcement learning, and parameter-efficient techniques like LoRA adapters. It provides a complete pipeline for aligning models with human preferences through multi-stage RLHF workflows, from supervised fine-tuning through preference optimization to reinforcement learning. The framework distinguishes itself through recipe-based training orchestration, where fine-tuning workflows are defined as composable recipe files that chain data loading, model configuration, and training l

This project provides a Chinese large language model based on the LLaMA architecture. It is an instruction-tuned model optimized for natural language processing and multi-turn conversations in Chinese. The system includes a framework for parameter-efficient fine-tuning using low-rank adaptation and quantization to reduce memory requirements. It also implements retrieval augmented generation for local document question answering and supports long-context processing for sequences up to 64K tokens. The project covers a broad set of capabilities including supervised instruction tuning, reinforce

MMF is a modular framework for building, training, and evaluating vision-and-language models. It provides a configuration-driven experiment system where model, dataset, and training parameters are defined through composable YAML files, alongside a curated model zoo of pretrained checkpoints for state-of-the-art multimodal architectures. The framework includes a multimodal dataset loader that downloads, processes, and batches vision-and-language data, and a vision-language model trainer supporting distributed training, mixed precision, and checkpoint-based resumption. The framework distinguish

This project is a distributed training infrastructure designed for aligning large language models through reinforcement learning. It functions as an end-to-end engine for complex alignment tasks, including proximal policy optimization, direct preference optimization, and iterative self-play. By providing a unified framework for multi-turn interactions and tool-use scenarios, it enables the development of models capable of reasoning and external environment engagement. The framework distinguishes itself through a decoupled architecture that separates model training from sample generation. This

OpenManus-RL is a reinforcement learning framework and distributed training pipeline designed to train large language models as agents. It serves as an agentic reasoning optimizer and reward model trainer, providing the infrastructure to improve model decision-making through reward-based policy optimization. The project distinguishes itself through a distributed architecture that supports parameter sharding across multiple compute nodes and a coordinated rollout system for collecting interaction trajectories. It incorporates advanced reasoning strategies, such as Tree-of-Thoughts and Monte Ca

This project is a framework for aligning large language models with human preferences. It provides a library for optimizing model behavior by mapping preference data directly to a policy objective, bypassing the need for a separate reward model. The framework utilizes a closed-form mathematical objective to adjust model weights by maximizing the likelihood of preferred responses relative to dispreferred ones. It includes a toolkit for supervised fine-tuning, allowing users to establish a stable baseline on demonstration data before applying preference-based optimization techniques. The libra

This project provides a comprehensive technical guide and framework for engineering large-scale machine learning systems. It covers the full lifecycle of model development, focusing on the infrastructure and computational principles required to build, train, and serve generative AI models across distributed GPU clusters. The repository distinguishes itself by offering deep-dive tutorials and implementation strategies for complex system challenges. It emphasizes high-performance architectural primitives, such as collective communication orchestration, distributed tensor sharding, and static gr

This project is an LLM research orchestrator and autonomous AI agent framework designed to automate the scientific lifecycle. It functions as an end-to-end research pipeline and model training toolkit, managing everything from initial literature reviews and hypothesis testing to the final drafting of academic papers. The system is distinguished by its ability to convert unstructured academic PDFs into machine-executable knowledge layers, allowing agents to reproduce and extend research findings. It employs a two-loop orchestration architecture and a specialized research engineering skill libr

DSPy is a declarative programming framework designed for building complex language model applications. It treats model interactions as modular, composable programs, allowing developers to define task logic through typed class schemas rather than relying on manually written prompts. By organizing workflows into hierarchical, reusable Python objects, the framework enables the construction of sophisticated AI systems that manage state and execution flow independently. The framework distinguishes itself through an automated optimization engine that iteratively refines prompt instructions and few-

Baselines is a comprehensive suite of frameworks for reinforcement learning algorithm implementation, imitation learning, and training orchestration. It provides a library of standardized learning algorithms used to benchmark and replicate research results, alongside a deep learning policy framework for constructing neural network architectures such as multi-layer perceptrons, convolutional networks, and long short-term memory networks. The project includes a specialized imitation learning toolkit that enables agents to mimic expert behavior through behavior cloning and generative adversarial