AutMn Lab

AutMn Lab AI Safety || Robotics
Science & Technology
Intuitive explanation of ongoing latest research
All views/opinions are personal

09/27/2026

What happens when you compare few-shot PEFT against few-shot ICL? In this 2022 study, T-Few's 11B model achieved 72.4% accuracy vs. 66.6% for GPT-3 175B ICL, while the authors estimated roughly three orders of magnitude fewer FLOPs per inference.

Liu et al., Few-Shot Parameter-Efficient Fine-Tuning is Better and Cheaper than In-Context Learning, NeurIPS 2022. NeurIPS-2022-few-shot-parameter…

09/26/2026

Instead of modifying huge weight matrices, (IA)³ learns tiny vectors that rescale internal Transformer activations

Liu et al., Few-Shot Parameter-Efficient Fine-Tuning is Better and Cheaper than In-Context Learning, NeurIPS 2022. NeurIPS-2022-few-shot-parameter…

Thrilled & honored to receive the NSF REU student funding under my ongoing NSF award ( #2525849)!If you're a CS undergra...
05/14/2026

Thrilled & honored to receive the NSF REU student funding under my ongoing NSF award ( #2525849)!

If you're a CS undergraduate at UA interested in designing safe controllers for robotic systems, with opportunities to work directly on real hardware platforms, feel free to reach out. Students with good mathematical maturity and programming experience are encouraged to apply; prior exposure to hardware platforms is a plus.

Please send your CV and transcript via email (Subject: REU Opportunity).

Details: bineet.cs.ua.edu

Lab Details: autmn.ua.edu

[ NSF REU, Undergraduate Research, University of Alabama, Robotics Research, Safe Autonomous Systems, Computer Science Students, Robotics Hardware, Formal Methods, AutMn Lab ]

A new look for AutMn.Same mission—building systems you can trust.Design. Verify. Ensure safety.
05/06/2026

A new look for AutMn.
Same mission—building systems you can trust.
Design. Verify. Ensure safety.

Ever see AI agents learn to hold the door for each other? A new research teaches robot teams to master complex, multi-st...
05/02/2026

Ever see AI agents learn to hold the door for each other? A new research teaches robot teams to master complex, multi-step sequences and work together like a dream!

Comment to know more.

Paper: Yalcinkaya, B., et al. (2025). Automata-Conditioned Cooperative Multi-Agent Reinforcement Learning.

Keywords: MARL, AI Research, Reinforcement Learning, Robotics, Computer Science

A new look for AutMn.Same mission—building systems you can trust.Design. Verify. Ensure safety.
05/02/2026

A new look for AutMn.

Same mission—building systems you can trust.
Design. Verify. Ensure safety.

Address

Tuscaloosa, AL
35401

Website

Alerts

Be the first to know and let us send you an email when AutMn Lab posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Shortcuts

Share