Best AI papers explained

534 Episodes

1. RLAD: Training LLMs to Discover Abstractions

   Published: 10/29/2025
2. How to Train Your Advisor: Steering Black-Box LLMs with ADVISOR MODELS

   Published: 10/29/2025
3. Self-improving LLM agents at Test-Time

   Published: 10/27/2025
4. KL-Regularized Reinforcement Learning is designed to Mode Collapse

   Published: 10/27/2025
5. How do LLMs use their depth?

   Published: 10/27/2025
6. Thought Communication in Multiagent Collaboration

   Published: 10/27/2025
7. Reasoning with Sampling: Base Models Outperform RL

   Published: 10/26/2025
8. Continual Learning via Sparse Memory Finetuning

   Published: 10/26/2025
9. Direct Preference Optimization with Unobserved Preference Heterogeneity: The Necessity of Ternary Preferences

   Published: 10/24/2025
10. The Coverage Principle: How Pre-Training Enables Post-Training

    Published: 10/24/2025
11. The Era of Real-World Human Interaction: RL from User Conversations

    Published: 10/24/2025
12. Agent Learning via Early Experience

    Published: 10/24/2025
13. Demystifying the Mechanisms Behind Emergent Exploration in Goal-conditioned RL

    Published: 10/22/2025
14. Rewriting History: A Recipe for Interventional Analyses to Study Data Effects on Model Behavior

    Published: 10/22/2025
15. A Definition of AGI

    Published: 10/22/2025
16. Provably Learning from Language Feedback

    Published: 10/21/2025
17. In-Context Learning for Pure Exploration

    Published: 10/21/2025
18. On the Role of Preference Variance in Preference Optimization

    Published: 10/20/2025
19. Training LLM Agents to Empower Humans

    Published: 10/20/2025
20. Richard Sutton Declares LLMs a Dead End

    Published: 10/20/2025