Best AI papers explained

506 Episoade

1. The Art of Scaling Reinforcement Learning Compute for LLMs

   Publicat: 16.10.2025
2. A small number of samples can poison LLMs of any size

   Publicat: 16.10.2025
3. Dual Goal Representations

   Publicat: 14.10.2025
4. Welcome to the Era of Experience

   Publicat: 14.10.2025
5. Value Flows: Flow-Based Distributional Reinforcement Learning

   Publicat: 14.10.2025
6. Self-Adapting Language Models

   Publicat: 12.10.2025
7. The Markovian Thinker

   Publicat: 12.10.2025
8. Moloch’s Bargain: emergent misalignment when LLMs compete for audiences

   Publicat: 12.10.2025
9. Transformer Predictor Dynamics and Task Diversity

   Publicat: 11.10.2025
10. Base models know how to reason, thinking models learn when

    Publicat: 11.10.2025
11. Spectrum tuning: Post-training for distributional coverage and in-context steerability

    Publicat: 11.10.2025
12. Understanding Prompt Tuning and In-Context Learning via Meta-Learning

    Publicat: 11.10.2025
13. MLPs Learn In-Context on Regression and Classification tasks

    Publicat: 11.10.2025
14. Is Pre-Training Truly Better than Meta-Learning?

    Publicat: 11.10.2025
15. Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models

    Publicat: 11.10.2025
16. Do LLMs Recognize Your Preferences? Evaluating Personalized Preference Following in LLMs

    Publicat: 09.10.2025
17. Learning dynamics of LLM finetuning

    Publicat: 09.10.2025
18. Iterative Data Smoothing: Mitigating Reward Overfitting and Overoptimization in RLHF

    Publicat: 09.10.2025
19. OpenAI Agent Builder and n8n: Orchestrating Reasoning Versus Automating Process

    Publicat: 08.10.2025
20. Training Agents Inside of Scalable World Models

    Publicat: 08.10.2025