Ax Murad Dawood, Usama Ahmed Siddiquie, Shahram Khorshidi, Maren Bennewitz 4/3/2026

Constraint-Aware Reinforcement Learning via Adaptive Action Scaling

Safe reinforcement learning approach using adaptive action scaling to reduce constraint violations during training.

Ax Hongyu Zhu, Lin Chen, Xin Jin, Mingsheng Shang 4/3/2026

MS-Mix: Sentiment-Guided Adaptive Augmentation for Multimodal Sentiment Analysis

Sentiment-guided augmentation technique for multimodal sentiment analysis addressing data scarcity in video, audio, and text.

Ax Dan Jacobellis, Mateen Ulhaq, Fabien Racap\'e, Hyomin Choi, Neeraja J. Yadwadkar 4/3/2026

DeDelayed: Deleting Remote Inference Delay via On-Device Correction

System for reducing remote video inference latency through on-device correction with lightweight models for robotics and edge devices.

Ax Elias Hossain, Mehrdad Shoeibi, Ivan Garibay, Niloofar Yousefi 4/3/2026

BIOGEN: Evidence-Grounded Multi-Agent Reasoning Framework for Transcriptomic Interpretation in Antimicrobial Resistance

Multi-agent framework using LLMs for interpreting gene clusters from RNA-seq data in antimicrobial resistance research.

Ax Dachuan Zhao, Weiyue Li, Zhenda Shen, Yushu Qiu, Bowen Xu, Haoyu Chen, Yongchao Chen 4/3/2026

Bias Is a Subspace, Not a Coordinate: A Geometric Rethinking of Post-hoc Debiasing in Vision-Language Models

Geometric approach to post-hoc debiasing in vision-language models by treating bias as subspace rather than individual coordinates.

Ax Sonia Laguna, Andrea Agostini, Alain Ryser, Samuel Ruiperez-Campillo, Irene Cannistraci, Moritz Vandenhirtz, Stephan Mandt, Nicolas Deperrois, Farhad Nooralahzadeh, Michael Krauthammer, Thomas M. Sutter, Julia E. Vogt 4/3/2026

Structure is Supervision: Multiview Masked Autoencoders for Radiology

Self-supervised learning framework using masked autoencoders to learn view-invariant representations from multi-view radiology data.

Ax Yuang Geng, Zhuoyang Zhou, Zhongzheng Zhang, Siyuan Pan, Hoang-Dung Tran, Ivan Ruchkin 4/3/2026

Deterministic World Models for Verification of Closed-loop Vision-based Systems

Deterministic world models for verification of vision-based control systems avoiding stochastic latent variable overapproximation.

Ax Irched Chafaa, Giacomo Bacci, Luca Sanguinetti 4/3/2026

Linear Attention for Joint Power Optimization and User-Centric Clustering in Cell-Free Networks

Lightweight transformer model for joint AP clustering and power allocation in cell-free massive MIMO networks.

Ax Safal Thapaliya, Zehong Wang, Jiazheng Li, Ziming Li, Yanfang Ye, Chuxu Zhang 4/3/2026

Semantic Refinement with LLMs for Graph Representations

Framework using LLMs to refine semantic information in graph representations for improved learning across diverse graph domains.

Ax Ben Xue, Dan Liu, Lixiang Wang, Mingjie Sun, Peng Wang, Pengfei Zhang, Shaoyun Shi, Tianyu Xu, Yunhao Sha, Zhiqiang Liu, Bo Kong, Bo Wang, Hang Yang, Jieting Xue, Junhao Wang, Shengyu Wang, Shuping Hui, Wencai Ye, Xiao Lin, Yongzhi Li, Yuhang Chen, Zhihui Yin, Quan Chen, Shiyang Wen, Wenjin Wu, Han Li, Guorui Zhou, Changcheng Li, Peng Jiang, Kun Gai 4/3/2026

Generative Recommendation for Large-Scale Advertising

GR4AD production-oriented generative recommender system for large-scale advertising with custom architecture and serving.

Ax Jiacheng Wang, Yucheng Sheng, Le Liang, Hao Ye, Shi Jin 4/3/2026

Wireless Power Control Based on Large Language Models

PC-LLM applies pre-trained LLMs as relational reasoning backbone for wireless power control optimization.

Ax Yulin Li, Tengyao Tu, Li Ding, Junjie Wang, Huiling Zhen, Yixin Chen, Yong Li, Zhuotao Tian 4/3/2026

Efficient Reasoning with Balanced Thinking

Balanced thinking method addressing overthinking and underthinking in Large Reasoning Models for efficient deployment.

Ax Amos Goldman (NVIDIA Corporation), Nimrod Boker (NVIDIA Corporation), Maayan Sheraizin (NVIDIA Corporation), Nimrod Admoni (NVIDIA Corporation), Artem Polyakov (NVIDIA Corporation), Subhadeep Bhattacharya (NVIDIA Corporation), Fan Yu (NVIDIA Corporation), Kai Sun (NVIDIA Corporation), Georgios Theodorakis (NVIDIA Corporation), Hsin-Chun Yin (NVIDIA Corporation), Peter-Jan Gootzen (NVIDIA Corporation), Aamir Shafi (NVIDIA Corporation), Assaf Ravid (NVIDIA Corporation), Salvatore Di Girolamo (NVIDIA Corporation), James Dinan (NVIDIA Corporation), Xiaofan Li (NVIDIA Corporation), Manjunath Gorentla Venkata (NVIDIA Corporation), Gil Bloch (NVIDIA Corporation) 4/3/2026

NCCL EP: Towards a Unified Expert Parallel Communication API for NCCL

NCCL EP library for unified Expert Parallelism communication in Mixture-of-Experts LLM training and inference.

Ax Jun Liu, Pu Zhao, Zhenglun Kong, Xuan Shen, Peiyan Dong, Fan Yang, Lin Cui, Hao Tang, Geng Yuan, Wei Niu, Wenbin Zhang, Xue Lin, Gaowen Liu, Yanzhi Wang, Dong Huang 4/3/2026

When Should a Robot Think? Resource-Aware Reasoning via Reinforcement Learning for Embodied Robotic Decision-Making

Resource-aware RL framework for embodied robotic agents to optimize when LLM reasoning is invoked during task execution.

Ax Guoan Wang, Shihao Yang, Jun-en Ding, Hao Zhu, Feng Liu 4/3/2026

NeuroNarrator: A Generalist EEG-to-Text Foundation Model for Clinical Interpretation via Spectro-Spatial Grounding and Temporal State-Space Reasoning

NeuroNarrator foundation model for EEG-to-text clinical interpretation using spectro-spatial grounding.

Ax Edoardo Calvello, Elizabeth Carlson, Nikola Kovachki, Michael N. Manta, Andrew M. Stuart 4/3/2026

Operator Learning for Smoothing and Forecasting

Theoretical framework for data-driven smoothing and forecasting algorithms in dynamical systems.

Ax Lisan Al Amin, Md Ismail Hossain, Rupak Kumar Das, Mahbubul Islam, Abdulaziz Tabbakh 4/3/2026

Democratizing AI: A Comparative Study in Deep Learning Efficiency and Future Trends in Computational Processing

Benchmark study comparing deep learning efficiency across models and discussing GPU resource accessibility trends.

Ax Jingzhi Fang, Xiong Gao, Renwei Zhang, Zichun Ye, Lei Chen, Jie Zhao, Chengnuo Huang, Hui Xu, Xuefeng Jin 4/3/2026

DVM: A Bytecode Virtual Machine Approach for Dynamic Tensor Computation

Bytecode virtual machine approach for efficient dynamic tensor computation with flexible shape and control flow handling.

Ax Mingmeng Geng, Yuhang Dong, Thierry Poibeau 4/3/2026

Beyond Via: Analysis and Estimation of the Impact of Large Language Models in Academic Papers

Analysis of linguistic shifts in arXiv papers attributable to LLM usage patterns and limitations in model classification.

Ax Yu Zhang, Karl Mason 4/3/2026

D-SPEAR: Dual-Stream Prioritized Experience Adaptive Replay for Stable Reinforcement Learning in Robotic Manipulation

D-SPEAR algorithm improves reinforcement learning stability in robotic manipulation through prioritized experience replay.

Ax Zhiyuan Peng, Xuyang Wu, Huaixiao Tou, Yi Fang, Yu Gong 4/3/2026

MemRerank: Preference Memory for Personalized Product Reranking

MemRerank framework distills user purchase history into preference signals for personalized product reranking in LLM shopping agents.

Ax Xue Jiang, Tianyu Zhang, Ge Li, Mengyang Liu, Taozhi Chen, Zhenhua Xu, Binhua Li, Wenpin Jiao, Zhi Jin, Yongbin Li, Yihong Dong 4/3/2026

Think Anywhere in Code Generation

Adaptive reasoning approach for LLM code generation that allocates thinking throughout implementation rather than upfront.

Ax Simone Betteti, Luca Laurenti 4/3/2026

Hybrid Energy-Based Models for Physical AI: Provably Stable Identification of Port-Hamiltonian Dynamics

Energy-based models for stable system identification in physical dynamics using Lyapunov functions.

Ax Zhengyang Tang, Ke Ji, Xidong Wang, Zihan Ye, Xinyuan Wang, Yiduo Guo, Ziniu Li, Chenxin Li, Jingyuan Hu, Shunian Chen, Tongxu Luo, Jiaxi Bi, Zeyu Qin, Shaobo Wang, Xin Lai, Pengyuan Lyu, Junyi Li, Can Xu, Chengquan Zhang, Han Hu, Ming Yan, Benyou Wang 4/3/2026

Do Phone-Use Agents Respect Your Privacy?

MyPhoneBench framework for evaluating privacy behavior of mobile phone-use AI agents during task execution.

HN dotvignesh 4/3/2026

When a reasoning LLM chooses, which comes first: thought or decision?

arXiv abstract about reasoning order in LLM decision-making processes.

HN srinivasvaddi 4/3/2026

Framecraft – Generate demo videos from a prompt using HTML Canvas/Skill/MCP/LLM

Framecraft: tool enabling LLMs to generate demo videos via HTML Canvas, MCP servers, and automated rendering from prompts.

HN ieuanking 4/3/2026

Show HN: Ubik – Analyze local files like NotebookLM, with Cursor-like power

Ubik Studio: local file analysis tool combining notebook-style interaction with multi-hop reasoning and approval workflows for knowledge workers.

HN Sottilde 4/3/2026

Cc-plugin-audit: Defensive supply-chain security analysis on plugin auto-update

Supply-chain security tool detecting and surfacing diffs for Claude Code plugin auto-updates. Security analysis for plugin ecosystems.

HN gmays 4/3/2026

AI Applications and Vertical Integration

Newsletter discussing vertical integration pattern in AI application companies.

HN dropsmith 4/3/2026

DropSmith – MCP server for structured NPC dialogue generation

MCP server for structured NPC dialogue generation with emotion tags. Works with Ollama locally. Game dev tool for LLM agents.

HN cochranblock 4/3/2026

Any-GPU: my AMD 5700 XT OC can EARN IT's rent using Debian Trixie+Vulkan

any-gpu tensor engine running on AMD/Intel/Apple GPUs using WGSL compute shaders. Cross-vendor GPU ML framework.

HN teleforce 4/3/2026

Fine-Tuning Large Language Models LLMs with a Production-Grade Pipeline (2023)

Production-grade ML pipeline for fine-tuning LLMs using SkyPilot and DVC for resource management and experiment tracking.

HN mooreds 4/3/2026

Run a Local LLM, and discover why LLMs are unpredictable

Guide to running local LLMs on personal hardware for cost reduction and privacy, covering unpredictability challenges.

HN olivergrabner 4/3/2026

Show HN: Composer – AI architect / MCP for software architecture diagrams

Tool that converts ideas and existing codebases into architecture diagrams using MCP protocol and Claude/LLM integration.

HN menggg 4/3/2026

I simulated a 19th-century utopian commune with AI agents

Simulation of Brook Farm utopian commune using AI agents with historical personas, exploring cooperation and resource management.

HN hkdb 4/3/2026

Show HN: LM Gate – Auth and access-control gateway for self-hosted LLM back ends

Open-source authentication and access-control gateway for self-hosted LLM backends like Ollama. Adds identity, authz, rate limiting.

HN didgeoridoo 4/3/2026

Show HN: Claudebar, the missing interactive menu bar for Claude Code

Claudebar wraps Claude Code in tmux with status bar, task management, and agent team panels. Developer tool enhancement.

HN mosiddi 4/3/2026

Open-source runtime security toolkit for autonomous AI agents covering OWASP Top

Open-source runtime security toolkit for autonomous AI agents, covering OWASP Top security frameworks for agent governance and safety.

HN bneb-dev 4/2/2026

Show HN: Trytet – Deterministic WASM substrate for stateful AI agents

Trytet: Embeddable WebAssembly substrate for stateful AI agents with zero-trust execution, sub-millisecond latency, and state persistence.

HN reddome 4/2/2026

AI benchmarks are broken. Here's what we need instead

Critique of isolated AI benchmarks proposing human-centered, context-specific evaluation methods for measuring real-world AI impact.

HN AmarDahmani 4/2/2026

Lisa Core – semantic compression for AI conversations (80:1 ratio, 100% local)

LISA Core browser extension providing semantic compression for AI conversations across platforms with local processing (80:1 ratio).

HN DREDREG 4/2/2026

Show HN: MicroSafe-RL – Sub-microsecond safety layer for Edge AI 1.18µs latency

MicroSafe-RL: Sub-microsecond safety layer for edge RL agents with 1.18µs latency preventing hardware damage during sim-to-real transfer.

HN Garrus800 4/2/2026

Genesis Agent – A self-modifying AI agent that runs local (Electron, Ollama)

Genesis Agent: Self-modifying local AI agent running on Electron/Ollama that reads, fixes, and evolves its own source code autonomously.

HN matt_d 4/2/2026

KernelEvolve: Meta's Ranking Engineer Agent Optimizes AI Infrastructure

KernelEvolve: Meta's agentic kernel authoring system that autonomously optimizes low-level AI infrastructure for ranking model efficiency at scale.

HN adinhitlore 4/2/2026

Ask HN: Not much interest in offline CPU LLMs that may kill SaaS?

Discussion of competitive landscape for offline CPU-based LLMs versus SaaS models, addressing privacy and connectivity concerns.

HN tanelpoder 4/2/2026

rpg: A modern psql-compatible Postgres terminal and TUI written in Rust

rpg: psql-compatible Postgres terminal in Rust with built-in DBA diagnostics and AI assistant, single binary, cross-platform.

HN navbuildz 4/2/2026

I built an Open-source Gmail MCP server: multi-account support, read/write

Open-source Gmail MCP server enabling AI agents multi-account email access with read/write capabilities, compatible with Claude and other MCP clients.

HN Kalki-M 4/2/2026

BlackSwanX,174 AI agents predict the future by fighting each other,run on Ollama

BlackSwanX: 174 AI agents running on Ollama predicting market movements via disagreement, seeks cognitive dissonance gaps rather than consensus.

HN ThatGuyRaion 4/2/2026

New IRIX Emulator (I am not the author)

SGI Indy MIPS emulator built in Rust with AI assistance, boots IRIX with networking and X11.

HN tetsuto 4/2/2026

Show HN: Turbo1Bit – Run Bonsai-8B at 65K context in 3.9 GB RAM

Turbo1Bit: KV cache compression enabling Bonsai-8B LLM with 65K context to run on 3.9GB RAM via quantization and Flash Attention.