OpenChat

Advancing Open Source LLMs with Mixed Quality Data through offline RL-inspired C-RLFT. ⠀⠀⠀⠀⠀⠀⠀⠀⠀ ⠀⠀⠀⠀⠀⠀⠀⠀⠀ ⠀⠀𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝗟𝗲𝗮𝗱: Guan Wang, @AlpayAriyak

[OpenChat @OpenChatDev]

🚀Introducing OpenChat 3.6

🌟Surpassed official Llama3-Instruct—with 1-2M synthetic data compared to ~10M human labels
🤫GPTs are close to limits—excel at generation but fall short at complex tasks
🎯We are training next gen—capable of deterministic reasoning and planning

🔗 Explore OpenChat-3.6 (20240522 Llama 3 Version):

HuggingFace: https://t.co/gcvR7caFVN
Live Demo: https://t.co/0ZMWnRStup
GitHub: https://t.co/rFeB7voG6V

[Guan Wang @makingAGI]

Will Sudoku become the MNIST for reasoning?

Simple rules, clear structure, unique solutions—yet surprisingly challenging for modern LLMs, often requiring explicit trial-and-error to solve.

https://t.co/BmqUaW8Qno

[Guan Wang @makingAGI]

🚀Introducing Hierarchical Reasoning Model🧠🤖

Inspired by brain's hierarchical processing, HRM delivers unprecedented reasoning power on complex tasks like ARC-AGI and expert-level Sudoku using just 1k examples, no pretraining or CoT!

Unlock next AI breakthrough with neuroscience. 🌟

📄Paper: https://t.co/Sxprojsv0c
💻Code: https://t.co/k15cUS2wlf

[OpenChat @OpenChatDev]

Thrilled to see RSP featured at AAAI'25! This pioneering concept was a key inspiration for developing OpenChat! 🚀 #AI #AAAI25

[Haoyi Niu✈️CVPR @t641769919]

🚨Recursive Skip-Step Planning (RSP)
Relying on larger, expressive models for sequential decision-making has recently become a popular choice, but are they truly necessary? Can we replace these heavy models? Yes—RSP empowers shallow MLPs to excel in long-horizon tasks!🧵(1/n)

[Alignment Lab AI @alignment_lab]

skronge bones in that one 🔍

excellent job!

a 7b model out cracking gpt4 turbo and gpt4o and claude 3 sonnet!

[OpenChat @OpenChatDev]

4)This is why I am embarking on a journey to explore new frontiers in AI, specifically targeting the current limitations of GPTs in Planning and Reasoning.

[OpenChat @OpenChatDev]

🚀Introducing OpenChat 3.6

🌟Surpassed official Llama3-Instruct—with 1-2M synthetic data compared to ~10M human labels
🤫GPTs are close to limits—excel at generation but fall short at complex tasks
🎯We are training next gen—capable of deterministic reasoning and planning

🔗 Explore OpenChat-3.6 (20240522 Llama 3 Version):

HuggingFace: https://t.co/gcvR7caFVN
Live Demo: https://t.co/0ZMWnRStup
GitHub: https://t.co/rFeB7voG6V

[OpenChat @OpenChatDev]

3) However, while training these new models, I can't help but realize the upper limit of what autoregressive models can do. They struggle to solve complex tasks such as software engineering, advanced mathematics, and creating super assistants. It is mathematically challenging for GPT models to efficiently and effectively decompose and plan for the multistep, deterministic actions necessary for AGI.

[OpenChat @OpenChatDev]

🚀Introducing OpenChat 3.6

🌟Surpassed official Llama3-Instruct—with 1-2M synthetic data compared to ~10M human labels
🤫GPTs are close to limits—excel at generation but fall short at complex tasks
🎯We are training next gen—capable of deterministic reasoning and planning

🔗 Explore OpenChat-3.6 (20240522 Llama 3 Version):

HuggingFace: https://t.co/gcvR7caFVN
Live Demo: https://t.co/0ZMWnRStup
GitHub: https://t.co/rFeB7voG6V

[OpenChat @OpenChatDev]

@burkov It's an experiment. To see if 6T tokens are the secret 🤣

[OpenChat @OpenChatDev]

🚀 The World's First Gemma fine-tune based on openchat-3.5-0106 data and method (C-RLFT). Almost the same performance as the Mistral-based version.

6T tokens = secret recipe?

HuggingFace: https://t.co/X5WIZyxAlr

[OpenChat @OpenChatDev]

@ramirosalas Thank you! We're tuning MoE hyperparams and getting a GPU cluster to train 70b 🤣

[OpenChat @OpenChatDev]

@ZahirHamroune 8192 (same as Gemma base)

[OpenChat @OpenChatDev]

@_philschmid @teknium @Mascobot Just replaced the Gemma's tokenizer with the instruction-tuned version's. Feel free to use it

https://t.co/wUNjn86t10

[OpenChat @OpenChatDev]

@o_b @LMStudioAI Thank you! Our Mixtral-based model is ongoing ⚙️

[OpenChat @OpenChatDev]

🚀Announcing OpenChat-3.5 Update 0106: 𝗪𝗼𝗿𝗹𝗱’𝘀 𝗕𝗲𝘀𝘁 𝗢𝗽𝗲𝗻 𝗦𝗼𝘂𝗿𝗰𝗲 𝟳𝗕 𝗟𝗟𝗠!

Experience ChatGPT & Grok-level AI locally 💿!

Surpassing Grok-0 (33B) across all 4 benchmarks and Grok-1 (???B) on average and 3/4 benchmarks 🔥.

🎯 This update mainly enhanced training methodology, in-context learning & coding skills, outperforming the last 1210 release on 7 out of 8 benchmarks!

🌐 The model is available on our live demo, HuggingFace, and GitHub:

HuggingFace: https://t.co/Y6XQbzB2cE
Live Demo: https://t.co/0ZMWnRStup
GitHub: https://t.co/rFeB7voG6V

🛠️ To deploy it yourself, visit our GitHub (https://t.co/rFeB7voG6V) for full instructions to serve OpenChat models with an accelerated vLLM backend, API key authentication, and more!

[OpenChat @OpenChatDev]

@aisinne

[OpenChat @OpenChatDev]

@aisinne Me, gpt4 and openchat have the same answer 🤣

"The banana is now in the living room, on the plate you took with you."