Best AI papers explained

430 Episodes

1. The Alternative Annotator Test for LLM-as-a-Judge: How to Statistically Justify Replacing Human Annotators with LLMs

   Published: 5/9/2025
2. Limits to scalable evaluation at the frontier: LLM as Judge won’t beat twice the data

   Published: 5/9/2025
3. Stratified Prediction-Powered Inference for Hybrid Language Model Evaluation

   Published: 5/9/2025
4. Accelerating Unbiased LLM Evaluation via Synthetic Feedback

   Published: 5/9/2025
5. Prediction-Powered Statistical Inference Framework

   Published: 5/9/2025
6. Optimizing Chain-of-Thought Reasoners via Gradient Variance Minimization in Rejection Sampling and RL

   Published: 5/9/2025
7. RM-R1: Reward Modeling as Reasoning

   Published: 5/9/2025
8. Reexamining the Aleatoric and Epistemic Uncertainty Dichotomy

   Published: 5/8/2025
9. Decoding Claude Code: Terminal Agent for Developers

   Published: 5/7/2025
10. Emergent Strategic AI Equilibrium from Pre-trained Reasoning

    Published: 5/7/2025
11. Benefiting from Proprietary Data with Siloed Training

    Published: 5/6/2025
12. Advantage Alignment Algorithms

    Published: 5/6/2025
13. Asymptotic Safety Guarantees Based On Scalable Oversight

    Published: 5/6/2025
14. What Makes a Reward Model a Good Teacher? An Optimization Perspective

    Published: 5/6/2025
15. Towards Guaranteed Safe AI: A Framework for Ensuring Robust and Reliable AI Systems

    Published: 5/6/2025
16. Identifiable Steering via Sparse Autoencoding of Multi-Concept Shifts

    Published: 5/6/2025
17. You Are What You Eat - AI Alignment Requires Understanding How Data Shapes Structure and Generalisation

    Published: 5/6/2025
18. Interplay of LLMs in Information Retrieval Evaluation

    Published: 5/3/2025
19. Trade-Offs Between Tasks Induced by Capacity Constraints Bound the Scope of Intelligence

    Published: 5/3/2025
20. Toward Efficient Exploration by Large Language Model Agents

    Published: 5/3/2025