Jorge Mendez-Mendez

I find it deeply annoying that papers (in AI and robotics) now treat the "Related Work" section as optional and often relegate it to an appendix (or omit it!). RW is part of the story: what gap is your work filling, and how does it build upon existing stuff. #AcademicWriting

If you're on the faculty job market this year, consider applying to our CS department's searches on AI, with particular attention to candidates on GenAI for programming, AI for science, and trustworthy AI. Job ad: apply.interfolio.com/175012. #AcademicJobMarket

Lol, I've used them heavily in my writing since I started my PhD. Now I have to check myself 🙃

Lots of great intuition in this thread about LLM RL.

Maybe "correcting" wasn't the right word. The TAMP system tells the LLM what the mistakes were (e.g., a precondition wasn't met or an object couldn't be placed) and asks the LLM to generate a new plan, in a loop.

The verdict? LLMs can't replace TAMP systems yet, especially for long-horizon, geometrically constrained tasks.

They are more promising as fast "idea generators", where a formal planner verifies their output.

Paper: arxiv.org/abs/2510.001...
Code & Data: github.com/jorge-a-mend...

More counter-intuitive insight: giving the LLM more information can hurt performance!

When the prompt included geometric details, the LLM made more PDDL errors. The extra info "distracts" the model from the logical constraints. Are these systems sensitive to prompt engineering!

Key finding: LLMs can solve many TAMP problems but lag behind engineered planners.

"Thinking" isn't always better! Non-reasoning LLMs outperformed reasoning ones.

Why? It's more efficient for the LLM generate plans quickly and have a formal TAMP system verify and correct them.

Can Large Language Models solve complex robotics problems with intricate geometric constraints? 🤖

Excited to share our preprint, "A Systematic Study of Large Language Models for Task and Motion Planning With PDDLStream"!

16 LLM planners using Gemini, 4950 TAMP problems.

🧵👇

I will be at @rl-conference.bsky.social this week. If you're around and would like to chat about lifelong learning, robotics, or abstractions, say hi!

I am looking forward to @collasconf.bsky.social next week, where I will be presenting my work and vision for lifelong robot learning through modularity.

Looking forward to seeing folks at @collasconf.bsky.social in Philly in a couple of weeks!

📢 2 weeks until #CoLLAs2025!

Here’s what you need to make the most of it

🗓 Program: lifelong-ml.cc/Conferences/...

✅ Papers: lifelong-ml.cc/Conferences/...

📍 Venue: lifelong-ml.cc/Conferences/...

🔗 Info & registration: lifelong-ml.cc/Conferences/...

#AI #ML #ContinualLearning #LifelongLearning

Submit your tutorial proposal or topic/presenter suggestions to tutorials@lifelong-ml.cc by April 30!

Broadcasting at its finest

I'll be at @neuripsconf.bsky.social #NeurIPS2024 next week for workshops and am looking to recruit PhD students for Fall '25. I'm generally interested in lifelong learning, robotics, and RL. If you'll be around and are interested in working with me, please send a note in Whova to meet up!

I'm curious what the gap would be for getting this running with other platforms. Your thread seems to suggest it's fairly flexible and could be used with other 6-DoF mobile manipulators?

This looks awesome! Maybe I should consider a Stretch as my starter robot for my lab. 🤔

The call for papers for RLC is now up! Abstract deadline of 2/14, submission deadline of 2/21!
Please help us spread the word.
rl-conference.cc/callforpaper...

We did spend a few weeks on deep RL. OOD and continual learning for sure will be there. What are some good reads for uncertainty quantification?

I'm wrapping up my Deep Learning course this Thursday with a lecture titled "The future of deep learning". What topics should make it in there? (I want to do very little of "the state-of-the-art of deep learning" and much more "things that don't yet work but some believe will be game-changing".)

📢 Exciting News! The Fourth Conference on Lifelong Learning Agents (CoLLAs 2025) will be held at the University of Pennsylvania (@Penn) in Philadelphia, USA 🇺🇸

🗓️ Important Dates:

Abstract Deadline: Feb 21, 2025
Submission Deadline: Feb 26, 2025
Conference Dates: Aug 11 - Aug 14, 2025 1/

I did a starter pack of people in New York (City) working on ML/AI. Please distribute and feel free to self nominate!

go.bsky.app/BoEtagz

🙋‍♂️ I'm on the Island, if you're counting that.

New feed on continual learning, lifelong learning, and related ML topics: bsky.app/profile/did:...

I hope you find it useful. Please let me know if you notice it is missing any posts, so that I can adjust the regular expression.

Scalable Continual Learning Abstract

It's almost time for #NeurIPS2024! I'll be there for workshops (and I'll be giving a talk at the Scalable Continual Learning workshop: sites.google.com/view/continu...). If you'll be around and want to chat, please reach out!

To clarify, this starter pack is for lifelong learning of the *machine* variety.

I work on embodied lifelong learning; would you mind adding me? Also, I made a related started pack for lifelong/continual learning: go.bsky.app/QXGY1iT. There may be some more folks there that are relevant!

That may be true. Normalizing high-quality non-anonymous criticism of all (not just powerful) people in our field would be a positive. I still wonder if many junior folks would shy away from writing highly critical reviews knowing that their name would be visible later.

While it may not be a prevalent issue, I don't think our community is exempt from that kind of abuse of power.