AI/ML Hacker News

Matrix Orthogonalization Improves Memory in Recurrent Models

A new research paper proposes using matrix orthogonalization to improve memory retention and stability in recurrent neural network models.

Tech Business/VC Hacker News

Firms that adopt AI grow headcount 10% over the two years following adoption

Study finds that firms adopting AI technologies tend to increase their total headcount by approximately 10% over the subsequent two years.

Software Engineering Hacker News

Pystd, similar-ish functionality with a fraction of the compile time

Introduction of Pystd, a library providing similar functionality to standard libraries but with significantly reduced compile times.

AI/ML arXiv cs.AI

MultiUAV-Plat: An LLM-Oriented Platform, Benchmark and Framework for Multi-UAV Collaborative Task Planning

The MultiUAV-Plat platform and Agent4Drone framework enable LLM-driven collaborative task planning for multiple UAVs.

AI/ML arXiv cs.AI

DDIAgents: Mechanism-Conditioned Context Flow for Drug-Drug Interaction Prediction

DDIAgents is a multi-agent framework that improves drug-drug interaction prediction by dynamically orchestrating specialized expert agents.

AI/ML arXiv cs.AI

Revealing Safety-Critical Scenarios for UTM via Transformer

A new transformer-based RL approach is used to discover safety-critical vulnerabilities in Unmanned Traffic Management (UTM) systems.

AI/ML arXiv cs.AI

The Past Is Prologue: A Plug-in Controller for Selective Updates in Sequentially Evolving LLM Memory

Janus is a plug-in memory controller for LLMs that selectively accepts memory updates to prevent knowledge overwriting and bias.

AI/ML arXiv cs.AI

Scenario Generation for Testing of Autonomous Driving Systems Using Real-World Failure Records

A pipeline using LLMs to generate synthetic testing scenarios for autonomous driving systems based on real-world failure records.

AI/ML arXiv cs.AI

Beyond the Library: An Agentic Framework for Autoformalizing Research Mathematics

A multi-agent framework using general coding LLMs to autoformalize research-level mathematics into Lean 4 proofs.

Other Hacker News

Redeploying Fable 5

A Hacker News thread discussing the redeployment of Fable 5.

AI/ML arXiv cs.AI

AgRefactor: Self-Evolving Agentic Workflow for HLS Compatibility and Performance

Introduces AgRefactor, an LLM-based multi-agent workflow that refactors software into HLS-compatible programs for better hardware synthesis performance.

Cybersecurity arXiv cs.AI

Neuro-Bayesian-Symbolic Residual Attention Shallow Network: Explainable Deep Learning for Cybersecurity Risk Assessment

Presents the Neuro-Bayesian-Symbolic Residual Attention Shallow Network (NBS-RASN) for explainable cybersecurity risk assessment in open-source ecosystems.

AI/ML arXiv cs.AI

HyPOLE: Hyperproperty-Guided Multi-Agent Reinforcement Learning under Partial Observation

Introduces HyPOLE, a framework for Multi-Agent Reinforcement Learning under partial observation guided by temporal logic HyperLTL.

AI/ML arXiv cs.AI

AgentBound: Verifiable Behavioral Governance for Autonomous AI Agents

Presents AgentBound, a runtime governance framework for autonomous AI agents that provides verifiable behavioral oversight using cryptographically verifiable receipts.

AI/ML arXiv cs.AI

When Regulation Has Memory: Hysteresis and Control Burden in Artificial Agency

Explores the concept of hysteresis and control burden in artificial agents, suggesting that adaptive agents should be evaluated by the control effort required for stability.

AI/ML arXiv cs.AI

A Three-Phase Foundation Model for Tax-Aware Personalized Portfolio Management

A three-phase deep reinforcement learning system for tax-aware personalized portfolio management using time series foundation models and Mixture of Experts (MoE).

AI/ML arXiv cs.AI

Beyond Compilation: Evaluating Faithful Natural-Language-to-Lean Statement Formalization

Analyzes the gap between compilation and faithful formalization of natural language to Lean theorem proving, proposing a more rigorous evaluation protocol.

AI/ML arXiv cs.AI

LabGuard: Grounding Natural-Language Laboratory Rules into Runtime Guards for Embodied Laboratory Agents

Introduces LabGuard, a safety suite that transforms natural-language laboratory rules into executable runtime monitors for embodied AI agents in scientific labs.

AI/ML arXiv cs.AI

OpenLife: Toward Open-World Artificial Life with Autonomous LLM Agents

Proposes OpenLife, a paradigm for open-world Artificial Life (ALIFE) using LLM agents with persistent memory and budget-based metabolism in the real world.

AI/ML Hacker News

Segmenting Robot Video into Actionable Subtasks

Discussion regarding the segmentation of robot videos into actionable subtasks for better task execution.