Large Language Models

Karpathy dropped a 200-line GPT, so I used the math to turn pandas DataFrames into searchable context windows and open sourced it (and automated my stats pipeline). [P]

TL;DR: I got tired of manually running Shapiro-Wilk tests and copy-pasting p-values at 2 AM. I built an open-source, async Python pipelin...

I built a solo AI platform from Algeria with no funding, no team and no ad spend - here's what's inside it after 2 months

Hello, 20 years old here just got into the Ai platform and launched this last two weeks and here is what I have on it so far. - Latest Ai...

USF murder suspect accused of using ChatGPT to research cover-up, prosecutors say

Days after the remains of one of the two missing University of South Florida doctoral students were found, prosecutors say the suspect ma...

[2603.04818] LLM-Grounded Explainability for Port Congestion Prediction via Temporal Graph Attention Networks

Abstract page for arXiv paper 2603.04818: LLM-Grounded Explainability for Port Congestion Prediction via Temporal Graph Attention Networks

[2603.04791] Timer-S1: A Billion-Scale Time Series Foundation Model with Serial Scaling

Abstract page for arXiv paper 2603.04791: Timer-S1: A Billion-Scale Time Series Foundation Model with Serial Scaling

[2603.04783] Breaking Contextual Inertia: Reinforcement Learning with Single-Turn Anchors for Stable Multi-Turn Interaction

Abstract page for arXiv paper 2603.04783: Breaking Contextual Inertia: Reinforcement Learning with Single-Turn Anchors for Stable Multi-T...

[2603.04751] Evaluating the Search Agent in a Parallel World

Abstract page for arXiv paper 2603.04751: Evaluating the Search Agent in a Parallel World

[2603.04750] HiMAP-Travel: Hierarchical Multi-Agent Planning for Long-Horizon Constrained Travel

Abstract page for arXiv paper 2603.04750: HiMAP-Travel: Hierarchical Multi-Agent Planning for Long-Horizon Constrained Travel

[2603.04741] CONE: Embeddings for Complex Numerical Data Preserving Unit and Variable Semantics

Abstract page for arXiv paper 2603.04741: CONE: Embeddings for Complex Numerical Data Preserving Unit and Variable Semantics

[2603.04735] Solving an Open Problem in Theoretical Physics using AI-Assisted Discovery

Abstract page for arXiv paper 2603.04735: Solving an Open Problem in Theoretical Physics using AI-Assisted Discovery

[2603.04670] Using Vision + Language Models to Predict Item Difficulty

Abstract page for arXiv paper 2603.04670: Using Vision + Language Models to Predict Item Difficulty

[2603.04636] When Agents Persuade: Propaganda Generation and Mitigation in LLMs

Abstract page for arXiv paper 2603.04636: When Agents Persuade: Propaganda Generation and Mitigation in LLMs

[2603.04631] Towards automated data analysis: A guided framework for LLM-based risk estimation

Abstract page for arXiv paper 2603.04631: Towards automated data analysis: A guided framework for LLM-based risk estimation

[2603.04589] ECG-MoE: Mixture-of-Expert Electrocardiogram Foundation Model

Abstract page for arXiv paper 2603.04589: ECG-MoE: Mixture-of-Expert Electrocardiogram Foundation Model

[2603.04582] Self-Attribution Bias: When AI Monitors Go Easy on Themselves

Abstract page for arXiv paper 2603.04582: Self-Attribution Bias: When AI Monitors Go Easy on Themselves

[2603.04549] Adaptive Memory Admission Control for LLM Agents

Abstract page for arXiv paper 2603.04549: Adaptive Memory Admission Control for LLM Agents

[2603.04514] Progressive Refinement Regulation for Accelerating Diffusion Language Model Decoding

Abstract page for arXiv paper 2603.04514: Progressive Refinement Regulation for Accelerating Diffusion Language Model Decoding

[R] Anyone experimenting with heterogeneous (different base LLMs) multi-agent systems for open-ended scientific reasoning or hypothesis generation?

Quick question — has anyone tried multi-agent setups where agents use genuinely different underlying LLMs (not just roles on the same mod...

OpenAI’s new GPT-5.4 model is a big step toward autonomous agents

OpenAI is launching GPT-5.4, the latest version of its AI model that the company says combines advancements in reasoning, coding, and pro...

Nearly Half Of UK Adults Happy To Use ChatGPT As A Counsellor

Anthropic officially told by DOD that it's a supply chain risk even as Claude used in Iran

How AI and ChatGPT Are Transforming the Wedding Industry

Labor market impacts of AI: A new measure and early evidence