Ax Erjian Zhang, Yatong Hao, Liejun Wang, Zhiqing Guo 5/25/2026

The Double Dilemma in Multi-Task Radiology Report Generation: A Gradient Dynamics Analysis and Solution

Analysis and solution for gradient balancing problems in multi-task learning for automatic radiology report generation.

Ax Santo M. A. R. Thies, Hubert Baniecki, R. Teal Witter, Eyke H\"ullermeier, Maximilian Muschalik, Fabian Fumagalli 5/25/2026

Proxy-Based Approximation of Shapley and Banzhaf Interactions

ProxySHAP method for efficiently computing Shapley and Banzhaf feature interactions in machine learning models.

Ax Samar Fares, Klea Ziu, Toluwani Aremu, Nikita Durasov, Martin Tak\'a\v{c}, Pascal Fua, Karthik Nandakumar, Ivan Laptev 5/25/2026

MirrorCheck: Efficient Adversarial Defense for Vision-Language Models

MirrorCheck framework for detecting adversarial attacks on vision-language models using text-to-image regeneration.

Ax Jiaqi Yao, Tianjian Huang, Zipeng Cai, Ding Liu 5/25/2026

Universal Matrix Multiplication on Quantum Computer

Research on optimizing matrix multiplication for neural networks using quantum computing parallel processing.

Ax Thea Br\"usch, Mikkel N. Schmidt, Tommy S. Alstr{\o}m 5/25/2026

A comprehensive evaluation of pretraining strategies for channel-agnostic contrastive self-supervision of biosignals

Evaluation of pretraining strategies for contrastive self-supervised learning on multivariate biosignal time series data.

Ax Zeyu Bian, Lan Wang, Zhengling Qi 5/25/2026

A Tale of Two Cities: Pessimism and Opportunism in Offline Dynamic Pricing

Offline reinforcement learning approach for dynamic pricing with incomplete price space coverage.

Ax Daniel Daza, Alberto Bernardi, Luca Costabello, Christophe Gueret, Masoud Mansoury, Michael Cochez, Martijn Schut 5/25/2026

Interactive Query Answering on Knowledge Graphs with Soft Entity Constraints

Method for query answering over incomplete knowledge graphs using soft entity constraints for vague real-world queries.

Ax Gaotang Li, Ruizhong Qiu, Xiusi Chen, Heng Ji, Hanghang Tong 5/25/2026

Beyond Log Likelihood: Probability-Based Objectives for Supervised Fine-Tuning across the Model Capability Continuum

Alternative probability-based objectives for LLM supervised fine-tuning beyond negative log likelihood with improved generalization across model scales.

Ax Peiming Yang, Sankeerth Durvasula, Ivan Fernandez, Mohammad Sadrosadati, Onur Mutlu, Gennady Pekhimenko, Christina Giannoula 5/25/2026

DCC: Data-Centric Compilation of Machine Learning Kernels for Processing-In-Memory Architectures

DCC: Data-centric compilation framework optimizing ML kernels including LLMs for processing-in-memory hardware architectures.

Ax Yuchen Yang, Yaru Zhao, Pu Yang, Shaowei Wang, Zhi-Hua Zhou 5/25/2026

ZipMoE: Efficient On-Device MoE Serving via Lossless Compression and Cache-Affinity Scheduling

ZipMoE: Efficient on-device MoE serving system using lossless compression and cache-affinity scheduling for resource-constrained edge deployment.

Ax Hayeong Lee, JunHyeok Oh, Byung-Jun Lee 5/25/2026

TABX: A High-Throughput Sandbox Battle Simulator for Multi-Agent Reinforcement Learning

TABX: High-throughput JAX-based sandbox for multi-agent reinforcement learning with modular environment design and custom evaluation scenarios.

Ax Zihao Jing, Qiuhao Zeng, Ruiyi Fang, Yan Yi Li, Yan Sun, Boyu Wang, Pingzhao Hu 5/25/2026

Scaling-Aware Adapter for Structure-Grounded LLM Reasoning

Scaling-aware adapter architecture enabling LLMs to reason over 2D/3D structures with geometric grounding and flexible modality fusion.

Ax Liren Yu, Caiyuan Li, Feiyi Dong, Tao Zhang, Zhixuan Zhang, Dan Ou, Haihong Tang, Bo Zheng 5/25/2026

Joint Model Parameter Scaling and Universal-Domain Data Integration for E-commerce Search Ranking

E-commerce search ranking system combining model parameter scaling with cross-domain data integration for improved performance.

Ax Jun Wang, Ziyin Zhang, Rui Wang, Hang Yu, Peng Di, Rui Wang 5/25/2026

TingIS: Real-time Risk Event Discovery from Noisy Customer Incidents at Enterprise Scale

Real-time anomaly detection system for cloud services discovering risk events from noisy customer incident data at scale.

Ax Sultan Zavrak 5/25/2026

Content-Aware Attack Detection in LLM Agent Tool-Call Traffic: An Empirical Study of Features, Architectures, and Evaluation Protocols

Attack detection framework for LLM agent Model Context Protocol tool-call traffic using graph encoding and embeddings.

Ax Tristan Kirscher (ICube, Institut Strauss, DKFZ), Markus Bujotzek (DKFZ), Yannick Kirchhoff (DKFZ), Maximilian Rokuss (DKFZ), Fabian Isensee (DKFZ), Kim-Celine Kahl (DKFZ), Balint Kovacs (DKFZ), Klaus Maier-Hein (DKFZ) 5/25/2026

Lost in the Folds: When Cross-Validation Is Not a Deep Ensemble for Uncertainty Estimation

Audit of K-fold cross-validation vs deep ensembles for uncertainty estimation in medical image segmentation tasks.

Ax Miao Li, Irina Saparina, Alexander Gurung, Mirella Lapata 5/25/2026

Long-Context Reasoning Through Proxy-Based Chain-of-Thought Tuning

Method for improving long-context LLM reasoning through proxy-based chain-of-thought tuning to handle 10M token inputs.

Ax Scott Freitas, Amir Gharib 5/25/2026

GenAI-Driven Threat Detection with Microsoft Security Copilot

Dynamic Threat Detection Agent using GenAI for adaptive cybersecurity incident investigation and detection logic generation.

Ax Yang Wu, Qiang Meng, Zhaojiang Liu, Youquan Liu, Jian Yang, Jin Xie 5/25/2026

Distill to Think, Foresee to Act: Cognitive-Physical Reinforcement Learning for Autonomous Driving

Reinforcement learning approach for autonomous driving combining cognitive foundation with foresighted physical environment prediction.

Ax Hai-Ling Lu, Yu-Yang Li, Yin-Bi Li, Cun-Shi Wang, A-Li Luo, Jun-Chao Liang, Shuo Li 5/25/2026

Spectra as Language: Large Language Models for Scalable Stellar Parameter and Abundance Inference

Large language models applied to extract stellar parameters and chemical composition from spectroscopic survey data.

HN Danny6969 5/25/2026

Lynote Humanize Text – Open-source AI text humanization toolkit

Open-source toolkit for AI text humanization using DeepSeek rewrites and cross-engine translation to evade detection.

HN AlazarManakelew 5/25/2026

MetalBench – Benchmark for Apple Silicon's Metal Shading Lang

MetalBench: benchmarking suite for Apple Silicon Metal GPU kernels against MLX reference implementations.

HN Damien_220 5/25/2026

Code-mapper: Free CLI tool to reduce LLM token usage on any codebases

code-mapper: CLI tool generating compact PROJECT_CONTEXT.md to reduce LLM token usage on codebases by 78%.

HN nonlooped 5/24/2026

Built a small PR guardrail for token bloat, worth maintaining?

ContextLevy: PR guardrail tool that comments on context bloat for AI coding agents.

HN Zick9510 5/24/2026

Arcana: Shape the Language. Control the Machine

Arcana: metaprogramming-oriented systems language with malleable syntax and LLVM backend.

HN rivernova 5/24/2026

Hosomaki. A local tool for understanding Linux system output.contributors wanted

Hosomaki: local Linux system analysis tool with optional offline AI layer for understanding system output.

HN Brajeshwar 5/24/2026

How to Tame AI's Voracious Appetite for Energy

Explores methods to reduce AI's energy consumption through better algorithms, hardware design, and computing approaches.

HN oinoom 5/24/2026

Careers After AI

Career guidance tool matching job titles to AI-era opportunities with skills and resources; OpenAI launches finance feature in ChatGPT.

HN victorbuilds 5/24/2026

Migrating a dozen side projects from Azure to one VPS with an LLM agent

Developer migrated 12+ side projects from Azure cloud services to a single VPS using an LLM agent, reducing complexity and costs.

HN synapsehub_ai 5/24/2026

OpenAI co-founder Andrej Karpathy joins Anthropic

Andrej Karpathy, OpenAI co-founder and prominent AI researcher, joins Anthropic as a major hire in AI talent competition.

HN wek 5/24/2026

Constraint Decay: The Fragility of LLM Agents in Back End Code Generation

arXiv research on constraint decay vulnerability in LLM agents for code generation tasks.

HN truetotosse 5/24/2026

Show HN: A picker that maps local LLMs to hardware, hardware to LLMs

Interactive tool mapping local LLMs to hardware compatibility with benchmarks, budget comparisons, and 2-day updated data on 69 builds.

HN kunalsin9h 5/24/2026

The guide to RL environments: building and scaling them in the LLM era

Guide to reinforcement learning environments optimized for LLM era; limited content visible.

HN amund 5/24/2026

Pi-Mojo – A Mojo Port of Pi AI Agent Toolkit

Pi-Mojo: Native Mojo port of Pi agentic toolkit providing compiled reference implementation for systems-level agent architectures and type-safe integrations.

HN akashtndn 5/24/2026

Autotrader – paper trading AI agent for Indian equities

Claude-powered paper trading agent autonomously manages portfolio on Indian equities (Nifty 500), achieved 8.05% return over two weeks with real transaction costs.

HN hydra-f 5/24/2026

Show HN: Kanban CLI (A local-first, agent-first task manager for the terminal)

Kanban CLI: Rust-based terminal task manager designed as coordination layer for AI agent workflows with git integration and local-first architecture.

HN sermakarevich 5/24/2026

Show HN: Fleet – Python supervisor for running coding agents in parallel

Fleet: Python supervisor tool for running multiple Claude Code agent sessions in parallel, inspired by AMD's use case of managing 50+ concurrent coding agents.

HN alansaber 5/24/2026

Orchestration problems hurt legal AI

Analysis of how AI agents struggle with legal document understanding due to lack of persistent context models and runtime reconstruction limitations.

HN networked 5/24/2026

Google adds llms.txt check to Chrome Lighthouse

Minimal stub entry on Google adding llms.txt file validation to Chrome Lighthouse tool.

HN shannon-data-ai 5/24/2026

ShannonBase: The Lightweight Semantic Layer for Enterprise AI SQL

ShannonBase: semantic layer framework for enterprise AI agents enabling Natural Language to SQL with constraint enforcement beyond basic code generation.

HN swolpers 5/24/2026

"Write as Little Code as Possible"

Analysis arguing agentic coding tools reduce implementation friction but don't solve product discovery, validation, and systems fit problems.

HN bignet 5/24/2026

Structured LLM Learning Path, from Zero to AI Researcher, 8-Phase Curriculum

Open-source 8-phase curriculum for learning LLMs and AI agents from fundamentals to research, covering transformers, pre-training, fine-tuning, alignment, and agents over ~5 months.

HN MaykonMan 5/24/2026

From source code 2 LLM constraints:a semantic extractor for Python, SwiftUI, Lua

Semantic Extractor tool that compiles framework source code into structured IR bundles capturing usage constraints, served over MCP protocol for LLM-assisted code generation in Python, SwiftUI, and Lua.

HN prakashqwerty 5/24/2026

Greg Brockman: Inside the 72 Hours That Almost Killed OpenAI

Podcast interview with OpenAI co-founder Greg Brockman discussing company history, AGI, and organizational moments.

HN EnXu 5/24/2026

Why AI Hardware Is a Chip Layer Problem

Personal essay on hardware manufacturing challenges in China's electronics ecosystem and AI infrastructure requirements.

HN 01-_- 5/24/2026

China behind in LLM race but it can still win in AI, ex-Tencent AI lead says

Analysis of China's LLM development lag and alternative AI competitive strategies by former Tencent AI leader.

HN boundless88 5/24/2026

For developers without design skills, how do you leverage AI for front end dev?

Discussion thread asking developers how to use AI tools for front-end development without design skills.

HN matt_d 5/24/2026

SSV: Sparse Speculative Verification for Efficient LLM Inference

Research paper evaluating machine learning compilers (PyTorch compile, TensorRT, XLA, ONN) for efficient LLM inference on NVIDIA GPUs.

HN matt_d 5/24/2026

Characterization of machine learning compilers for LLM inference on NVIDIA GPUs

Anatomia: AI agent tool using four agents to scope, plan, build, and verify code changes with typed contracts and proof chains.

HN stealthyAF 5/24/2026

Scan any codebase in 3s, then verify what your AI builds

Java-based AI coding agent using LangChain4j and Mistral API. Generates functional code like calculator apps autonomously.