O1-Journey

O1 Journey: Strategic Progress Report Part One

CommonProductProgrammingArtificial IntelligenceMachine Learning

O1-Journey is a project initiated by the GAIR research group at Shanghai Jiao Tong University, aimed at replicating and reimagining the capabilities of OpenAI's O1 model. This project introduces a novel training paradigm called 'journey learning' and has successfully built the first model that integrates search and learning in mathematical reasoning. Through processes such as trial and error, correction, backtracking, and reflection, this model has become an effective method for tackling complex reasoning tasks.

Release the Journey Thinking training dataset on the Hugging Face platform for researchers and developers to use.
Introduce a new paradigm of journey learning that emphasizes continuous progress through learning
reflection
and adaptation
enabling AI systems to adapt to the complexities of the real world.
Explore the cognitive processes of O1 by analyzing its thinking structure and long-term thinking workflow
and how to construct long-term thinking.
Build a reward model that enhances LLMs' capabilities in reflection and backtracking through fine-grained step-level evaluations.
Construct a reasoning tree using single-step reasoning strategies to generate possible reasoning steps based on the problem at hand.
Evaluate and train models using a visualization data analysis platform built with Streamlit to assess model performance and conduct iterative training.

The target audience includes artificial intelligence researchers
developers
and students
particularly those interested in large language models
machine learning

Researchers use the datasets provided by O1-Journey for model training and testing.
Developers build their own reasoning systems using codes and methodologies provided by O1-Journey.
Students learn the latest artificial intelligence technologies and research methods through the O1-Journey project.

1. Visit the O1-Journey GitHub page to learn about the project background and objectives.
2. Download and study the datasets provided by the project
including the Journey Thinking training dataset.
3. Read the project documentation to understand the paradigm and methodology of journey learning.
4. Utilize the provided code and tools to attempt building and training your own models.

Visit

O1-Journey Visit Over Time

Monthly Visits

515580771

Bounce Rate

37.20%

Page per Visit

5.8

Visit Duration

00:06:42

O1-Journey Visit Trend

O1-Journey Visit Geography

O1-Journey Traffic Sources

O1-Journey Alternatives

OLMo-2-1124-13B-DPO — High-performance English language model suitable for diverse tasks.

Programming•Language Model•Natural Language Processing

ProactiveAgent — A proactive agent based on large language models that predicts user needs and offers assistance.

Programming•Artificial Intelligence•Machine Learning

OLMo 2 — State-of-the-art fully open language model

Programming•Language Model•Natural Language Processing

Qwen Turbo 1M Demo — The Qwen Turbo 1M Demo is a Hugging Face space provided by Qwen.

Productivity•Natural Language Processing•Machine Learning

LongVU — Spatiotemporal Adaptation Compression Model for Long Video Language Understanding

Video•Video Understanding•Spatiotemporal Compression

awesome-LLM-resources — A comprehensive collection of global large language model resources.

Productivity•Large Language Models•Resource Aggregation

Llama-3.2-3B — Multilingual Large Language Model

Productivity•Artificial Intelligence•Machine Learning

192

YunHu Ai — An AI assistant that creates limitless possibilities.

ChineseSelection•Artificial Intelligence•Chat Assistant

114

SmartDialogue — Explore unlimited intelligence and build a more perfect path of aggregation.

chatting•Artificial Intelligence•Natural Language Processing

582

RAGLAB — A modular, research-oriented unified framework for retrieval-augmented generation

Programming•Natural Language Processing•Machine Learning

DeepSeek-V2.5 — An AI model integrating general and programming capabilities.

Productivity•Artificial Intelligence•Machine Learning

186

West Lake AI Model — A multimodal model with high emotional and intellectual intelligence

ChineseSelection•Artificial Intelligence•Multimodal

264

RWKV v6 Finch 14B — RWKV v6 Finch 14B is an open-source large model that efficiently processes long texts.

Productivity•Artificial Intelligence•Natural Language Processing

216

C4AI Command R+ 08-2024 — A multilingual advanced dialogue generation model with 104 billion parameters.

Productivity•Artificial Intelligence•Machine Learning

LLM Playground — An interactive platform for loading and testing large language models.

Productivity•Artificial Intelligence•Machine Learning

Models Table — A comprehensive list and information about large language models

Others•Large Language Models•Machine Learning

198

MiLM-6B — A large-scale pre-trained language model developed by Xiaomi with a parameter scale of 64 billion.

Programming•Natural Language Processing•Machine Learning

252

GPT-4o — GPT-4o, a flagship model capable of real-time processing of audio, visual, and textual data.

chatting•Artificial Intelligence•Natural Language Processing

318

Fugaku-LLM — Fugaku-LLM is an AI model specializing in text generation.

Productivity•Text Generation•Machine Learning

300

Next AI Jobs — Discover the best AI jobs and career opportunities in artificial intelligence, machine learning, natural language processing, and data science.

Business•Artificial Intelligence•Machine Learning

114

MAP-NEO — MAP-NEO is an entirely open-source large language model offering advanced natural language processing capabilities.

Programming•Natural Language Processing•Open Source

432

Amazon Titan Text Premier — An advanced language model designed to build generative AI applications based on Retrieval Augmented Generation (RAG) and agents

Productivity•Artificial Intelligence•Natural Language Processing

234

phi3-Chinese — Committed to collecting and organizing the various training variants of the phi3 model from the open source community, as well as providing tutorials for training, inference, and deployment.

Programming•Artificial Intelligence•Machine Learning

258

spRAG — A Retrieval-Augmented Generation (RAG) framework designed to address challenging queries for dense unstructured data

Programming•Artificial Intelligence•Natural Language Processing

LLaVA++ — LLaVA++ extends the LLaVA model by integrating Phi-3 and LLaMA-3, enhancing the interaction capability between visual and language models.

Programming•Artificial Intelligence•Natural Language Processing

468

Best AI Websites & Tools

O1-Journey

O1-Journey Visit Over Time

O1-Journey Visit Trend

O1-Journey Visit Geography

O1-Journey Traffic Sources

O1-Journey Alternatives

OLMo-2-1124-13B-DPO — High-performance English language model suitable for diverse tasks.

ProactiveAgent — A proactive agent based on large language models that predicts user needs and offers assistance.

OLMo 2 — State-of-the-art fully open language model

Qwen Turbo 1M Demo — The Qwen Turbo 1M Demo is a Hugging Face space provided by Qwen.

LongVU — Spatiotemporal Adaptation Compression Model for Long Video Language Understanding

awesome-LLM-resources — A comprehensive collection of global large language model resources.

Llama-3.2-3B — Multilingual Large Language Model

YunHu Ai — An AI assistant that creates limitless possibilities.

SmartDialogue — Explore unlimited intelligence and build a more perfect path of aggregation.

GRIN-MoE — High-performance, low-resource consumption hybrid expert model

Mistral-Small-Instruct-2409 — High-performance instruct-tuned AI model

xAI — AI to Accelerate Human Scientific Discovery

MemoRAG — A memory-based RAG framework for general-purpose applications

RAGLAB — A modular, research-oriented unified framework for retrieval-augmented generation

DeepSeek-V2.5 — An AI model integrating general and programming capabilities.

West Lake AI Model — A multimodal model with high emotional and intellectual intelligence

RWKV v6 Finch 14B — RWKV v6 Finch 14B is an open-source large model that efficiently processes long texts.

C4AI Command R+ 08-2024 — A multilingual advanced dialogue generation model with 104 billion parameters.

LLM Playground — An interactive platform for loading and testing large language models.

Models Table — A comprehensive list and information about large language models

MiLM-6B — A large-scale pre-trained language model developed by Xiaomi with a parameter scale of 64 billion.

GPT-4o — GPT-4o, a flagship model capable of real-time processing of audio, visual, and textual data.

Fugaku-LLM — Fugaku-LLM is an AI model specializing in text generation.

Next AI Jobs — Discover the best AI jobs and career opportunities in artificial intelligence, machine learning, natural language processing, and data science.

MAP-NEO — MAP-NEO is an entirely open-source large language model offering advanced natural language processing capabilities.

Amazon Titan Text Premier — An advanced language model designed to build generative AI applications based on Retrieval Augmented Generation (RAG) and agents

phi3-Chinese — Committed to collecting and organizing the various training variants of the phi3 model from the open source community, as well as providing tutorials for training, inference, and deployment.

spRAG — A Retrieval-Augmented Generation (RAG) framework designed to address challenging queries for dense unstructured data

LLaVA++ — LLaVA++ extends the LLaVA model by integrating Phi-3 and LLaMA-3, enhancing the interaction capability between visual and language models.