Ax Manasi Waghe, Danish Chandargi, Mohammad Aamir Rayyan, Raviraj Joshi, A. R. Deshpande 14d ago

Structure-Preserving Document Translation via Multi-Stage LLM Pipeline: A Case Study in Marathi

Multi-stage LLM pipeline for preserving document structure and formatting in machine translation of Indian government documents to Marathi.

Ax Bingye Ni, Xiaoyu Wang, Yingli Wang, Lingjiong Zhu 14d ago

Variance Reduction for Stochastic Gradient Generalized Non-reversible Langevin Monte Carlo Algorithms

Theoretical analysis of variance reduction in stochastic gradient Langevin Monte Carlo algorithms under small-stepsize regime.

Ax Carel F. W. Peeters 14d ago

Perspectives on Latent Factor Indeterminacy and its Implications for Data Representation

Analysis of indeterminacy problem in latent factor models and relationship to autoencoders and generative networks.

Ax Junyi Zou, Avrova Donz 14d ago

Memory-Managed Long-Context Attention: A Preliminary Study of Editable Request-Local Memory

Memory-managed attention mechanism separating long-context compression from explicit memory write/edit operations in language models.

Ax Xuanting Wu, Fan Zhanga, Fei Ma, Ling Guan, Guochun Ma, Yongsheng Zhou 14d ago

A Task-Driven and Quality-Assured Agent Framework for SAR Data Generation

SAGA: Agent framework for synthetic aperture radar data augmentation with task-driven generation and quality validation.

Ax Zhenyu Liu, Huaze Tang, Shao-Lun Huang 14d ago

A Theoretical Interpretation of In-Context Learning via Probabilistic Modeling

Probabilistic theoretical framework explaining in-context learning in LLMs with rigorous performance analysis.

Ax Edward Raff, Maor Ashkenazi, Sagar Samtani, David J. Elkind, Sven Krasser 14d ago

Cybersecurity is the True Frontier for Generative AI Success or Failure

Position paper on cybersecurity as testbed for LLM-based autonomous agents handling heterogeneous tools and data.

Ax Junyan Tan, Yifan Li, Minghao Wang, Zihan Chen, Haoyu Zhang 14d ago

FinInvest-GTCN: Explainable Graph-Temporal-Causal Modeling for Risk-Aware Investment Decision Optimization

Graph-temporal-causal network for venture capital investment risk assessment combining heterogeneous data sources.

Ax Junhan Li, Yuxin Zhang, Haoran Wang, Minghao Chen 14d ago

A3M: Adaptive, Adversarial and Multi-Objective Learning for Strategic Bidding in Repeated Auctions

A3M framework combining reinforcement learning and adversarial reasoning for strategic bidding in multi-unit auctions.

Ax Aoying Zheng, Anqi Du, Zizhuang Deng, Yuxuan Chen 14d ago

FlipGuard: Defending Large Language Models Against Quantization-Conditioned Backdoor Attacks

FlipGuard defense framework protecting LLMs against backdoor attacks that activate only after quantization.

Ax Eun Cheol Choi, Youngrae Kim, Prabhu Pugalenthi, Hong-En Chen, Bo-Ruei Huang 14d ago

Beyond the Mean: Three-Axis Fidelity for Aligning LLM-Based Survey Simulators from Small Pilot Data

Methods for calibrating LLM-based survey simulators to recover statistical properties of human populations from pilot data.

Ax Francesca Pia Panaccione, Eugenio Lomurno, Francesca Fati, Carlotta Pecchiari, Marina Rosanu, Luigi De Vitis, Lucia Ribero, Gabriella Schivardi, Giovanni Damiano Aletti, Nicoletta Colombo, Maria Francesca Spadea, Francesco Multinu, Matteo Matteucci, Elena De Momi 14d ago

Evidence-Based Text-Conditioned 3D CT Synthesis for Ovarian Cancer

Text-conditioned 3D CT synthesis using generative models to create synthetic ovarian cancer imaging data.

Ax Soroosh Tayebi Arasteh 14d ago

The strength of clinical evidence is recoverable from language model representations but not from their stated grades

Analysis of whether LLM internal representations capture clinical evidence strength differently from stated confidence grades.

Ax Samuel Ahnert, Esther Lagemann, H. Jane Bae, Kunihiko Taira, Ricardo Vinuesa, Christian Lagemann, Steven L. Brunton 14d ago

Weak Dominant Balance for Robust Identification of Dynamically Consistent Fluid Flow Structure

Method for extracting physical mechanisms from spatiotemporal fluid dynamics data using weak dominant balance analysis.

Ax Fabio Ciani, Harald Schweiger, Emilia Parada-Cabaleiro, Markus Schedl 14d ago

A Comparative Study on Affective Cues in Text Embeddings Across Psychological Emotion Theories

Study examining how text embeddings capture psychological emotion theories for sentiment analysis and emotion recognition.

Ax Young-Jun Lee, Seungone Kim, Minki Kang, Alistair Cheong Liang Chuen, Zerui Chen, Seungho Han, Taehee Jung, Dongyeop Kang 14d ago

Evolution Fine-Tuning: Learning to Discover Across 371 Optimization Tasks

LLMs integrated with evolutionary search to optimize across 371 tasks including GPU kernel design and mathematical conjectures.

Ax George Coote, Matthew J. Colbrook 14d ago

Residual-Guided Dictionary Learning for Spectrally Accurate Koopman Approximation

Dictionary learning method for improving spectral accuracy in Koopman operator approximations of nonlinear dynamics.

Ax Giorgio Angelotti, Stephen Parsons, Federica Nicolardi, Youssef Nader, Sean Johnson, David Josey, Paul Henderson, Hendrik Schilling, Johannes Rudolph, Forrest McDonald, Elian Rafael Dal Pr\'a, Paul Tafforeau, Alessandro Mirone, Clifford Seth Parker, Jan Paul Posma, Benjamin Kyles, Claudio Vergara, Alessia Lavorante, Rossella Villa, Maria Chiara Robustelli, Marzia D'Angelo, Gianluca Del Mastro, Michael McOsker, Kilian Fleischer, Christy Chapman, Nat Friedman, William Brent Seales 14d ago

Complete virtual unwrapping and reading of a rolled Herculaneum papyrus

X-ray CT and virtual unwrapping technique to digitally read unopened Herculaneum papyri rolls.

Ax Jing Yang, Mayoore Jaiswal, Zian Wang, Steven Zeng, Rochelle Pereira, Yajie Zhao, Jianyuan Min 14d ago

HorizonRelight: Relighting Long-horizon Videos Consistently via Diffusion Transformers

Diffusion transformer approach for consistent relighting of long video sequences without temporal discontinuities.

Ax Stephan Goerttler, Min Wu, Fei He 14d ago

Connectivity Estimation using Stochastic Graph Heat Modelling

Methods for estimating brain connectivity from neurophysiological data using stochastic graph heat modeling.

Ax Abdullah Al Shafi, Md Kawsar Mahmud Khan Zunayed, Safin Ahmmed, Sk Imran Hossain, Engelbert Mephu Nguifo 14d ago

BTI-Net: Bidirectional Decoder-Level Task Interaction via Uncertainty-Aware Gating for Multi-Task Medical Image Analysis

BTI-Net architecture for multi-task medical image segmentation and classification with bidirectional decoder-level task interaction.

Ax Hooman Torkaman, Ellis Oti Boateng, Jignesh Solanki, Anurag Srivastava 14d ago

An Integrated Two-Stage Deep-Learning Tool for Rapid Post-Hurricane Damage Identification and Repair Scheduling

Two-stage deep learning tool for post-hurricane damage assessment and repair scheduling using synthetic datasets and optimization.

Ax Brian Keith-Norambuena 14d ago

An Information-Geometric Justification for Composite Coherence in Event-Based Narrative Extraction

Information-theoretic analysis of coherence metrics for graph-based narrative extraction from text using embedding similarity and topic proximity.

Ax Yi Shi, Yifeng Jiang, Chen Tessler, Xue Bin Peng 14d ago

GPC: Large-Scale Generative Pretraining for Transferable Motor Control

Generative Pretrained Controllers for physics-based character animation using tokenization and next-token modeling on motion datasets.

Ax Pranath Reddy 14d ago

Selective Memory Retention for Long-Horizon LLM Agents

TraceRetain framework for memory management in frozen LLM agents. Scores memory entries by interpretable features and evicts lowest-scoring ones at capacity. Tested on ALFWorld with gpt-4-mini.

HN tartoran 14d ago

Why AI is like a (Clever Hans) Horse [video]

Video comparing AI behavior to Clever Hans phenomenon. Minimal content provided.

HN danieltanfh95 14d ago

Moonshot AI (kimi) launches a credit card

Moonshot AI launches credit card with AI membership benefits integrated. Promotional content with data collection terms.

HN Kiyo-Lynn 14d ago

AI Does Not Have to Kill Humans to End the Human Future

Philosophy-first AGI safety research lab discusses existential risks from AI. Incomplete excerpt from founder statement.

HN dodizzle 14d ago

Shipwright Harness – open-source autonomous delivery agent for Claude Code (MIT)

Open-source autonomous coding agent for Claude Code that plans, builds, reviews and ships tasks through pipeline.

HN ninglz 14d ago

Show HN: Sell Lemons Info – a source-checked Roblox guide site

Fan-maintained game guide for Roblox tycoon idle game with verified source-checked information.

HN enamakel 14d ago

Show HN: TinyAgents – a Rust based recursive LLM harness

Rust-based recursive LLM harness for orchestrating sub-agents, inspired by LangGraph and LangChain with Rust transpiler.

HN ivewonyoung 14d ago

Super Micro Office Raided as Taiwan Expands Nvidia Chip Smuggling Probe

Taiwan authorities raid Super Micro Computer investigating Nvidia chip smuggling to China. Supply chain enforcement news.

HN type0 14d ago

Why Won't Europe Build AI Data Centers in Iceland?

Europe plans to triple AI data centers but overlooks Iceland's geothermal infrastructure. Explores EU tech sovereignty and infrastructure challenges.

HN ivrr 14d ago

Show HN: Agentic Orchestrator, a TUI for long-running coding agents

Terminal UI for orchestrating long-running coding agents through engineering workflow phases: requirements, design, implementation, review.

HN honghe 14d ago

Daytona sandbox is closed-source

Notice that Daytona sandbox moved development to private codebase, repo no longer maintained.

HN swq115 14d ago

Show HN: Let an AI manage your homelab over MCP without giving it SSH

Tool allowing AI agents to manage homelab infrastructure via MCP protocol without SSH access. Open source developer tool for AI agent orchestration.

HN lemming 14d ago

Tau – Learn how coding agents are built

Educational Python coding agent project demonstrating agent architecture with readable layers, tool calling, and session management.

HN jacoblav 14d ago

Show HN: PDF Insight – local-first AI that sorts your PDFs on-device

PDF Insight: local-first on-device AI tool that classifies, sorts, and merges PDF documents without cloud upload. Privacy-focused developer tool.

HN foray-ai 14d ago

Foray – AI that generates a complete business blueprint and stress-tests ur idea

Foray generates business plans and financial models using AI, stress-tests assumptions. Consumer AI product with promotional focus.

HN ejhooooon 14d ago

Show HN: AMA2, messenger built for AI agent

AMA2: messaging runtime specifically designed for AI agents to manage context and multi-step tasks. Open source Show HN project addressing agent-tool integration gaps.

HN soji_mathew 14d ago

Open Memory Protocol – One Memory Store for Claude, ChatGPT, Curso

Protocol enabling shared memory storage across Claude, ChatGPT, and Curso. Standardizes LLM context persistence.

BL 14d ago

Introducing GeneBench-Pro

GeneBench-Pro benchmark evaluates AI agent judgment in computational biology: interpreting patterns, assessing data quality, decision-making.

BL 14d ago

Core dump epidemiology: fixing an 18-year-old bug

OpenAI infrastructure debugging blog post; incomplete excerpt.

BL 14d ago

Inside Genebench-Pro

GeneBench-Pro: research-level benchmark measuring AI agent performance on computational biology tasks with ambiguous scientific data.

HN Tanxsinxlnx 15d ago

How to Optimize Tools for Agents

Guide on optimizing tools for AI agents. Limited detail without content provided.

HN enricodeleo 15d ago

Show HN: Crudio – Turn an OpenAPI spec into a stateful mock back end

Crudio: CLI tool converts OpenAPI 3.x specifications into stateful mock API backends with schema validation and persistence. Developer testing tool.

HN AlgerianSam 15d ago

Show HN: OpenSfM v1.0

OpenSfM v1.0 revived: Structure-from-Motion engine for aerial/GIS imagery processing. Faster, more scalable open source computer vision tool.

HN chainshieldai 15d ago

Open-source AI agent workflow for auditing Solidity smart contracts

Open-source workflow using AI agents to audit Solidity smart contracts. Headline only, limited details.

HN blackhawk094 15d ago

Zeus – a local AI agent orchestrator with web and phone UI (open source)

Local open-source AI agent orchestrator with web/phone UI. Runs agents for file ops, browser control, web search, task management.

HN jnord 15d ago

South Korea to spend $1T on more memory chip production and humanoid robots

South Korea government announces $1T investment in memory chips, AI data centers, and humanoid robots by 2028.