Teach Robots to Learn Like People

Kari Jaaskelainen

21 Nov 2025 — 1 min read

Teach Robots to Learn Like People

Building versatile embodied AI (agents that see, read, and act) is hard because real-world data is scarce and training is expensive.

We introduce Deliberate Practice Policy Optimization (DPPO), a metacognitive "metaloop" that alternates between:

Supervised fine-tuning to expand general competence
Reinforcement learning to refine skills through practice

This loop spots weaknesses automatically and focuses compute where it matters, squeezing more learning from limited data. Theoretically, DPPO unifies preference learning; in practice, it powers Pelican-VL 1.0, a vision-language embodied model.

Results: +20.3% over the base model and +10.6% over open-source models around the 100B-parameter scale—using a systematic, data- and resource-efficient training recipe.

We’re open-sourcing the models and code so the community can build stronger embodied agents with fewer resources.

Paper: https://arxiv.org/abs/2511.16602v1

Paper: https://arxiv.org/abs/2511.16602v1

Register: https://www.AiFeta.com

AI Robotics EmbodiedAI ReinforcementLearning VisionLanguageModels MachineLearning OpenSource DataEfficiency

Meet GRACE: a moral governor for safer, more transparent AI

AI agents are getting powerful—so how do we make sure they do the right thing, not just the effective thing? Meet GRACE, a reason-based moral governor that keeps AI behavior aligned with human norms by separating moral reasoning from goal-driven decision-making. * Moral Module: uses deontic logic and explicit reasons

Why some fine-tuned LLMs miss phishing—and how to fix it

Not all fine-tuned LLMs spot phishing equally. A new study tests Llama 3.1 8B, Gemma 2 9B, and Mistral on high-stakes phishing detection—and uses SHAP and mechanistic interpretability to reveal why models do (or don’t) generalize. * Architecture × data diversity matters: Gemma 2 9B hits state-of-the-art performance (F1

AI that explains itself—by following the science

AI that explains itself—by following the science Black-box AI can be powerful, but it often can’t tell us why it made a decision. Concept Bottleneck Models (CBMs) try to fix that by predicting human-understandable concepts first, then the final answer. The catch: standard CBMs ignore domain-specific cause-and-effect and

Meet GenomAgent: A Team of AI Specialists for Smarter Genomics Q&A

TL;DR: Finding reliable answers in genomics is hard. GenomAgent turns one big AI into a coordinating team of specialists—and beats the current leader by 12% on a key benchmark. Genomic facts live across many databases. Standard chatbots struggle because they can’t flexibly query those sources. GeneGPT added

Teach Robots to Learn Like People

Read more

Meet GRACE: a moral governor for safer, more transparent AI

Why some fine-tuned LLMs miss phishing—and how to fix it

AI that explains itself—by following the science

Meet GenomAgent: A Team of AI Specialists for Smarter Genomics Q&A