Posts by Tags

2026

Reinforcement Learning

Curiosity-Critic: Rewarding Cumulative Progress

12 minute read

Published: May 18, 2026

Tags: 2026, Reinforcement Learning, Research, World Models

Agent trajectories at step 30,000 on a stochastic grid world. — Accepted to EIML @ ICML 2026. What is Curiosity?! Our work Curiosity-Critic separates reducible from irreducible prediction error to train world models without getting trapped by noisy-TV dynamics.

Research

Curiosity-Critic: Rewarding Cumulative Progress

12 minute read

Published: May 18, 2026

Tags: 2026, Reinforcement Learning, Research, World Models

Agent trajectories at step 30,000 on a stochastic grid world. — Accepted to EIML @ ICML 2026. What is Curiosity?! Our work Curiosity-Critic separates reducible from irreducible prediction error to train world models without getting trapped by noisy-TV dynamics.

World Models

Curiosity-Critic: Rewarding Cumulative Progress

12 minute read

Published: May 18, 2026

Tags: 2026, Reinforcement Learning, Research, World Models

Agent trajectories at step 30,000 on a stochastic grid world. — Accepted to EIML @ ICML 2026. What is Curiosity?! Our work Curiosity-Critic separates reducible from irreducible prediction error to train world models without getting trapped by noisy-TV dynamics.