All Articles
17939 articles total
AI Economist Agent: An Agentic Framework for Model-Grounded Economic Analysis with RAG, Knowledge Graphs, and Large Language Models
A framework for an AI economist agent that uses RAG and knowledge graphs to ground economic analysis in theory and data.
See-and-Reach: Precise Vision-Language Navigation for UAVs within the Field of View
A vision-language waypoint prediction framework called 3DG-VLN for precise UAV navigation using high-resolution visual cues.
Evaluation of EEG Foundation Models for Event-Based Burst-Suppression Detection in ICU
Evaluation of EEG foundation models for detecting burst-suppression patterns in ICU patients, finding REVE-base most effective.
Variable-Length Tokenization via Learnable Global Merging for Diffusion Transformers
A new variable-length tokenizer using learnable global merging to improve the quality-compute trade-off in Diffusion Transformers.
The Hidden Evolution of Disguised Visual Context inside the VLM
Research on how visual tokens evolve and are reshaped within Vision-Language Models (VLMs) based on their integration architecture.
IHUBERT: Vector-Based Semantic Deduplication and Domain-Balanced Pretraining for Persian Resources
IHUBERT, a RoBERTa-base Persian pretrained language model trained on curated data with semantic deduplication.
MakeupMirror: Improving Facial Attribute Preservation in Diffusion Models for Makeup Transfer
MakeupMirror, a diffusion-based makeup transfer model that preserves facial identity and skin tone using geometry conditioning.
Hybrid Diffusion Transformer for Instruction-Guided Audio Editing via Rectified Flow
A hybrid diffusion transformer architecture using rectified flow for efficient, instruction-guided audio editing.
Spotify Killed the Thrill of the Hunt
A discussion on how Spotify's recommendation algorithms have reduced the joy of discovering music manually.
A New Bill Takes Aim at Government Pressure to Silence Lawful Online Speech
Coverage of a new legislative bill aimed at preventing government pressure to silence lawful online speech.
Windows 11 update broke the Recycle Bin, OneDrive, and your PC's stability
Reports of a Windows 11 update causing stability issues and breaking the Recycle Bin and OneDrive.
Is AI ruining our skills? Early results are in – and they're not good
Exploration of whether AI is degrading human skill sets based on early empirical results.
Connect the Dots: Training LLMs for Long-Lifecycle Agents with Cross-Domain Generalization Via Reinforcement Learning
Introduces the 'Connect the Dots' (CoD) framework for training LLMs to be long-lifecycle agents capable of continuous environment exploration and self-updating context.
StreamKL: Fast and Memory-Efficient KL Divergence for Boosting Attention Distillation
Presents StreamKL, a fused GPU primitive that enables memory-efficient KL divergence for attention distillation, reducing HBM footprint from quadratic to constant.
Hierarchical Control in Multi-Agent Games: LLM-based Planning and RL Execution
A hybrid architecture combining LLM-based strategic planning with RL-based execution for improved coordination in multi-agent games.
When Lower Privileges Suffice: Investigating Over-Privileged Tool Selection in LLM Agents
Investigates over-privileged tool selection in LLM agents and proposes a privilege-aware post-training defense to enforce the principle of least privilege.
Hyundai buys Boston Dynamics, Atlas humanoid to be used at vehicle plant by 2028
Hyundai has acquired Boston Dynamics, intending to integrate the Atlas humanoid robot into its vehicle manufacturing plants by 2028.
Google workspace threatening to block Firefox access
Google Workspace is reportedly threatening to block access for users utilizing the Firefox browser.
This 1976 University Experiment Spun Up the U.S. Wind Industry
An exploration of a 1976 university experiment that helped catalyze the development of the US wind energy industry.
Show HN: Pagecast – Publish Markdown/HTML Reports to Cloudflare Pages
Pagecast is a new tool that allows developers to publish Markdown or HTML reports directly to Cloudflare Pages.