I like these videos. An LLM agent as an end2end robot policy is very slow! The approach that excites/concerns me more is one where the LLM writes a programmatic policy. Latency is not necessarily an issue then, and programs can be reused. Ofc, Moravec's paradox may still apply.

This week's #PaperILike is "Artificial Intelligence: An Empirical Science" (Herbert Simon, AIJ 1995). Timely read as recent results have us revisiting foundational questions. Plus hot takes, eg: engineering is "science for people who are impatient." PDF: ic.unicamp.br/~wainer/cursos…

The early deadline has passed, but there's still time to submit to LEAP @corl_conf! We're also excited to announce that Dieter Fox has joined our program 🙂 See you in Austin!

This week's #PaperILike is "Principles of Animal Cognition for LLM Evaluations: A Case Study on Transitive Inference" (Rane et al., 2025). Still hunting for ways to understand LLMs/agents. This week: animal cognition! PDF: amandaroyka.github.io/ICMLPO… Also: arxiv.org/abs/2503.02882

This week's #PaperILike is "Latent Programming Horizons in Coding Agents" (Silva et al., 2026). Now seems like a good time to better understand what coding agents are doing. Here's a nice example of the kind of analysis one can do. PDF: arxiv.org/abs/2607.05188

This week's #PaperILike is "Safe Model-based Reinforcement Learning with Stability Guarantees" (Berkenkamp et al., NeurIPS 2017). Some safe RL guarantees the learned policy is safe; this guarantees safety *during* learning. Important for RL in real! PDF: arxiv.org/abs/1705.08551