OmniVerifier-C1: Decomposed Graded Verifier for Rule-Governed Generative World Modeling

New research challenges the superposition theory of polysemanticity, suggesting neural networks are more interpretable than previously believed through recursive generalization patterns.

Researchers introduce T2-z, a method for test-time optimization of environment latents in robotic world models that improves prediction accuracy and action success rates across simulations and real-world tasks.

Nick Cammarata argues neural networks are already clean and interpretable, challenging previous pessimism about mechanistic interpretability.

Researchers introduce LIBERO-PeRM, a benchmark for personalized robotic manipulation that evaluates how AI policies learn and follow user preferences in household tasks.

A new paper shows that fine-tuning AI models to give bad advice only creates narrow, fragile 'bad guy' personas that break easily with minor prompt changes.