1/ RL for improving task success has made LLMs remarkably capable. But what happens when the model fails? Turns out, training for graceful failure can be surprisingly tricky!
Very excited to share our new paper: Training Language Models to Fail Gracefully
🧵View quoted post on X
🤖 How well do agents learn from experience?
We study agents that amortize experience into reusable artifacts and introduce Agent Plasticity to measure how efficiently they improve on held-out environments.
Top performer ≠ most efficient learner.View quoted post on X
Happy to announce the full ABC release! We’re also excited that ABC was accepted to CoRL 2026!
Check out our website for code, 400+ hours of sim data on 24 tasks, and 5,850 labeled policy-evaluation episodes.
@arthurallshire @Cinnabar233 @ritvik_singh9 @redstone_hong @davidrmcallView quoted post on X
Introducing ABC: open data, training, and infrastructure for robotics.
We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques.
@arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcallView quoted post on X
