Best AI papers explained

512 Folgen

1. e3: Learning to Explore Enables Extrapolation of Test-Time Compute for LLMs

   Vom: 17.6.2025
2. Extrapolation by Association: Length Generalization Transfer in Transformers

   Vom: 17.6.2025
3. Uncovering Causal Hierarchies in Language Model Capabilities

   Vom: 17.6.2025
4. Generalization or Hallucination? Understanding Out-of-Context Reasoning in Transformers

   Vom: 17.6.2025
5. Improving Treatment Effect Estimation with LLM-Based Data Augmentation

   Vom: 17.6.2025
6. LLM Numerical Prediction Without Auto-Regression

   Vom: 17.6.2025
7. Self-Adapting Language Models

   Vom: 17.6.2025
8. Why in-context learning models are good few-shot learners?

   Vom: 17.6.2025
9. Take Caution in Using LLMs as Human Surrogates: Scylla Ex Machina∗

   Vom: 14.6.2025
10. The Logic of Machines: The AI Reasoning Debate

    Vom: 12.6.2025
11. Layer by Layer: Uncovering Hidden Representations in Language Models

    Vom: 12.6.2025
12. Causal Attribution Analysis for Continuous Outcomes

    Vom: 12.6.2025
13. Training a Generally Curious Agent

    Vom: 12.6.2025
14. Estimation of Treatment Effects Under Nonstationarity via Truncated Difference-in-Q’s

    Vom: 12.6.2025
15. Strategy Coopetition Explains the Emergence and Transience of In-Context Learning

    Vom: 12.6.2025
16. Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs

    Vom: 11.6.2025
17. Agentic Supernet for Multi-agent Architecture Search

    Vom: 11.6.2025
18. Sample Complexity and Representation Ability of Test-time Scaling Paradigms

    Vom: 11.6.2025
19. Aligning with Human Judgement: The Role of Pairwise Preference in Large Language Model Evaluators

    Vom: 10.6.2025
20. LLMs Get Lost In Multi-Turn Conversation

    Vom: 9.6.2025