Baptiste Colle (@_BaptisteColle) header
Baptiste Colle (@_BaptisteColle) profile picture

Baptiste Colle

@_BaptisteColle

Founder @ https://teapawt.org/t.co/VpkpMbr83M 👀 | Building the next generation of AI assistants ✨ | Ex-HuggingFace researcher 🤗 NeurIPS author

 Ireland
baptistecolle.com
Joined February 2017
  • Tweets64
  • Following252
  • Followers435
12Media
Photo by Baptiste Colle (@_BaptisteColle)Photo by Baptiste Colle (@_BaptisteColle)Photo by Baptiste Colle (@_BaptisteColle)Photo by Baptiste Colle (@_BaptisteColle)Photo by Baptiste Colle (@_BaptisteColle)Photo by Baptiste Colle (@_BaptisteColle)Photo by Baptiste Colle (@_BaptisteColle)Photo by Baptiste Colle (@_BaptisteColle)Photo by Baptiste Colle (@_BaptisteColle)Photo by Baptiste Colle (@_BaptisteColle)
I’ve officially wrapped up my time at Hugging Face. Working with such brilliant talent has been a highlight of my career. I learned a ton, grew fast, and loved contributing to work that pushes the field forward 🚀🤗
Image from Baptiste Colle (@_BaptisteColle)
2
7
370
✨ What’s next I’m taking the leap and starting my own company to finally scratch the entrepreneurship itch. One big lesson from HF: the value of building in public. So that’s exactly what I’ll be doing here for this next chapter 💡🔥
1
3
164
I’m building ViralPilot ✈️ viralpilot.co For teams managing tens, hundreds, or thousands of TikTok accounts. If you know, you know. More soon.
3
136
Proud to share our work MindForge with @MirceaTudor4, @ossjs & @chiragraman accepted at NeurIPS 2025 🎉 We explore social lifelong learning agents 🤝that learn and grow together 🧠 Super relevant as agent-agent collaboration will be key to the future of intelligence 🔮
Chirag Raman (@chiragraman) profile pictureChirag Raman@chiragraman26 Oct 2025
Lifelong learning meets social intelligence: Human intelligence evolved in social environments; we learn by observing, teaching, and adapting to others. Recent lifelong-learning LLM agents, like Voyager, explore open worlds such as Minecraft, acquiring and reusing skills across tasks. Yet they still learn in isolation, and when powered by open-weight LLMs often fail at even the simplest tasks. Our NeurIPS '25 paper MindForge introduces social lifelong learning: agents that continually improve by reasoning about one another's beliefs, goals, and actions through perspective-taking and interaction. 📄 arxiv.org/abs/2411.12977 🧵1/7
Figure illustrating the differences between Voyager and MindForge agents. On the left, the Voyager agent powered by an open-weight LLM fails a simple task, saying “I want to collect dirt but I don’t have tools,” and cannot proceed. On the right, two MindForge agents exchange natural-language messages in the in-game chat: one asks for help obtaining a dirt block, and the other infers its partner’s situation and explains that dirt can be mined by hand. The dialogue leads to task success, highlighting how MindForge uses perspective-taking and communication to coordinate. Caption: “MindForge enables agents to collaborate through structured Theory-of-Mind representations, improving task success compared to Voyager.”

ALT Figure illustrating the differences between Voyager and MindForge agents. On the left, the Voyager agent powered by an open-weight LLM fails a simple task, saying “I want to collect dirt but I don’t have tools,” and cannot proceed. On the right, two MindForge agents exchange natural-language messages in the in-game chat: one asks for help obtaining a dirt block, and the other infers its partner’s situation and explains that dirt can be mined by hand. The dialogue leads to task success, highlighting how MindForge uses perspective-taking and communication to coordinate. Caption: “MindForge enables agents to collaborate through structured Theory-of-Mind representations, improving task success compared to Voyager.”

1
3
403
🚀 We just dropped the codebase for our Jupyter Agent Dataset! 📊 Downloaded 7TB of Kaggle Notebooks & Datasets, creating 0.2B agentic training traces to empower LLMs to craft/edit Jupyter notebooks. 🧠 Check out our blog for the full scoop! 🔍
Image from Baptiste Colle (@_BaptisteColle)
4
20
3
76
8,275
more replies
📖 Don’t forget to read our blogpost for a deep dive into our work and detailed analysis!
Image from Baptiste Colle (@_BaptisteColle)
1
4
248
Here are the links to the blog post and codebase 📚💻 - huggingface.co/blog/jupyter-… - github.com/huggingface/jupyt…
Jupyter Agents: training LLMs to reason with notebooks

Jupyter Agents: training LLMs to reason with notebooks

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

huggingface.co
2
185
Building a Jupyter notebook agent was a blast! 🎉 Watching it create notebooks to answer dataset questions is pure magic. Notebooks are key to science, and we hope this sparks better agents for the scientific community! 🌟
1
1
91
📖 Don’t forget to read our blogpost for a deep dive into our work and detailed analysis!
Image from Baptiste Colle (@_BaptisteColle)
1
74
Load more