OpenChat @OpenChatDev

Advancing Open Source LLMs with Mixed Quality Data through offline RL-inspired C-RLFT. ⠀⠀⠀⠀⠀⠀⠀⠀⠀ ⠀⠀⠀⠀⠀⠀⠀⠀⠀ ⠀⠀𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝗟𝗲𝗮𝗱: Guan Wang, @AlpayAriyak huggingface.co/openchat Joined July 2023

Tweets

60
Followers

2K
Following

42
Likes

141

Guan Wang @makingAGI

11 months ago

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. huggingface.co/datasets/sapie…

4 9 68 8K 21

View Details

Guan Wang @makingAGI

11 months ago

🚀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: arxiv.org/abs/2506.21734 💻Code: github.com/sapientinc/HRM

227 625 4K 1.3M 3K

View Details

OpenChat @OpenChatDev

2 years ago

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

Haoyi Niu✈️CVPR @t641769919

2 years ago

🚨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)

1 3 11 3K 2

1 0 6 764 3

View Details

Alignment Lab AI @alignment_lab

2 years ago

skronge bones in that one 🔍 excellent job! a 7b model out cracking gpt4 turbo and gpt4o and claude 3 sonnet!

Jie Liu @jie_liu1

2 years ago

🚀Excited to share our Storm-7B🌪️. This model achieves a 50.5% length-controlled win rate against GPT-4 Preview, making it the first open-source model to match GPT-4 Preview on AlpacaEval 2.0. 📄arxiv.org/pdf/2406.11817 🤗huggingface.co/jieliu/Storm-7B

11 18 109 13K 51

0 3 7 2K 6

View Details

OpenChat @OpenChatDev

2 years ago

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.

2 0 22 2K 1

View Details

OpenChat @OpenChatDev

2 years ago

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.

2 0 11 2K 0

View Details

OpenChat @OpenChatDev

2 years ago

🚀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: huggingface.co/openchat/openc… Live Demo: openchat.team GitHub: github.com/imoneoi/opench…

9 68 288 33K 128

View Details

OpenChat @OpenChatDev

2 years ago

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

2 0 3 332 1

View Details

OpenChat @OpenChatDev

2 years ago

🚀 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: huggingface.co/openchat/openc…

11 29 177 27K 72

View Details

OpenChat @OpenChatDev

2 years ago

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

1 0 7 306 1

View Details

OpenChat @OpenChatDev

2 years ago

@ZahirHamroune 8192 (same as Gemma base)

0 0 1 379 0

View Details

OpenChat @OpenChatDev

2 years ago

@_philschmid @teknium @Mascobot Just replaced the Gemma's tokenizer with the instruction-tuned version's. Feel free to use it huggingface.co/imone/gemma-7b…

0 0 3 599 1

View Details

OpenChat @OpenChatDev

2 years ago

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

1 0 5 557 0

View Details

OpenChat @OpenChatDev

2 years ago

🚀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: huggingface.co/openchat/openc… Live Demo: openchat.team GitHub: github.com/imoneoi/opench… 🛠️ To deploy it yourself, visit our GitHub (github.com/imoneoi/opench…) for full instructions to serve OpenChat models with an accelerated vLLM backend, API key authentication, and more!