Best AI papers explained

550 Episodes

1. Front-Loading Reasoning: The Synergy between Pretraining and Post-Training Data

   Published: 10/18/2025
2. Representation-Based Exploration for Language Models: From Test-Time to Post-Training

   Published: 10/18/2025
3. The attacker moves second: stronger adaptive attacks bypass defenses against LLM jail- Breaks and prompt injections

   Published: 10/18/2025
4. When can in-context learning generalize out of task distribution?

   Published: 10/16/2025
5. The Art of Scaling Reinforcement Learning Compute for LLMs

   Published: 10/16/2025
6. A small number of samples can poison LLMs of any size

   Published: 10/16/2025
7. Dual Goal Representations

   Published: 10/14/2025
8. Welcome to the Era of Experience

   Published: 10/14/2025
9. Value Flows: Flow-Based Distributional Reinforcement Learning

   Published: 10/14/2025
10. Self-Adapting Language Models

    Published: 10/12/2025
11. The Markovian Thinker

    Published: 10/12/2025
12. Moloch’s Bargain: emergent misalignment when LLMs compete for audiences

    Published: 10/12/2025
13. Transformer Predictor Dynamics and Task Diversity

    Published: 10/11/2025
14. Base models know how to reason, thinking models learn when

    Published: 10/11/2025
15. Spectrum tuning: Post-training for distributional coverage and in-context steerability

    Published: 10/11/2025
16. Understanding Prompt Tuning and In-Context Learning via Meta-Learning

    Published: 10/11/2025
17. MLPs Learn In-Context on Regression and Classification tasks

    Published: 10/11/2025
18. Is Pre-Training Truly Better than Meta-Learning?

    Published: 10/11/2025
19. Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models

    Published: 10/11/2025
20. Do LLMs Recognize Your Preferences? Evaluating Personalized Preference Following in LLMs

    Published: 10/9/2025