AI/ML arXiv cs.AI

Agentic ERP: Multi-Agent Large Language Model Architecture for Autonomous Enterprise Resource Planning

The paper introduces Agentic ERP, a multi-agent LLM architecture that uses role-aligned agents and a graph-based orchestrator to autonomously execute complex business workflows.

AI/ML arXiv cs.AI

DeeperRadar: End-to-End MIMO Radar Design and Multi-Modal Fusion for Autonomous Vehicle Perception

DeeperRadar is an end-to-end framework that co-designs MIMO radar sensing and multi-modal fusion for autonomous vehicles to optimize receiver antenna activation.

AI/ML arXiv cs.AI

Self-Modifying Lean Proof Agents with Verifier-Grounded Benchmark Coevolution

Researchers developed a self-evolving Lean proof agent that coevolves its workflow and benchmark, achieving a 45.1% solve rate on a held-out miniF2F test split.

AI/ML arXiv cs.AI

Quantifying Diversity of Thought: A Predictive Law of Weighted LLM Ensemble Lift

This study derives a formal law to predict the performance lift of LLM ensembles based on 'diversity of thought' and accuracy-adjusted correctness correlation.

AI/ML arXiv cs.AI

Intermittent Control Is Not Diluted Control: A Switching Effect in Artificial Agency

The paper identifies a 'switching effect' in artificial agency where intermittent access to anticipatory control can reduce the long-term regulatory burden of recovery.

AI/ML arXiv cs.AI

Empirical Grounding Improves the Realism of LLM Agents Simulating Human Behavior During Disruptions

The study demonstrates that grounding LLM agents with real-world demographic and routine data significantly improves their ability to simulate human behavior during disruptions.

Software Engineering arXiv cs.AI

AEC-DS: Adaptive Erasure Coding with PDP-Triggered Reputation and QoS-Aware Migration for Decentralized Storage

AEC-DS is a closed-loop adaptive erasure coding mechanism for decentralized storage that uses PDP feedback to optimize shard placement and data durability.

AI/ML arXiv cs.AI

Panache: One-Pass Motif Discovery at Every Window Length

Panache is a new one-pass streaming algorithm for z-normalized pan matrix profile motif discovery that is significantly faster than current CPU and GPU baselines.

AI/ML arXiv cs.AI

Pailitao-MMSearch: Building Native E-Commerce Multimodal Search Foundation

Taobao introduces Pailitao-MMSearch, a native multimodal search foundation model for e-commerce that improves GMV and transaction volume through hybrid reasoning.

Software Engineering Hacker News

How to pack ternary numbers in 8-bit bytes

A discussion on the technical methods of packing ternary numbers into 8-bit bytes for memory efficiency.

Other Hacker News

VTubing: How a Japanese Phenomenon Is Going Worldwide

An exploration of the global expansion of VTubing, focusing on its cultural impact and growth from Japan.

AI/ML arXiv cs.AI

Is Your Model Thinking or Just Stagnating? PUMA: Diagnosing Reasoning Pathology via Phase-Momentum Alignment

Introduces PUMA, a training-free framework designed to diagnose and mitigate 'overthinking' in Large Reasoning Models by aligning phase-momentum.

AI/ML arXiv cs.AI

Toward Anthropomorphic Dialogue: A Closed-Loop Framework for Human-Like Chat Generation, Evaluation, and Preference Alignment

Presents AnthroDial, a closed-loop framework for generating and evaluating human-like, anthropomorphic dialogue in LLMs.

AI/ML arXiv cs.AI

A Systematic Evaluation of Trajectory Data Curation for LoRA Fine-Tuning of Code Agents

A systematic study on how the quality and quantity of trajectory data affect the LoRA fine-tuning of code agents like Qwen2.5-Coder.

AI/ML arXiv cs.AI

Constrained Path Reasoning: Measuring When Committed Stages Earn Their Cost

Discusses Constrained Path Reasoning (CPR), a method to measure the cost-effectiveness of committed stages in LLM reasoning pipelines.

AI/ML arXiv cs.AI

LenGuard-GPC: Length Guarding with Guided-Prompt Consistency for Spatial Reasoning Reinforce Learning

Introduces LenGuard-GPC, a dense reward framework for Reinforcement Learning to optimize spatial reasoning and control response length in VLMs.

AI/ML arXiv cs.AI

Coordinated Disentanglement with Iterative Mode Discovery Under Hidden Correlations

Proposes CoDID, an end-to-end framework for disentangled representation learning that discovers modes under hidden correlations.

AI/ML arXiv cs.AI

An Explicit World Model Based on Data-First Ontology: DaoQL Multimodal Storage Validation and Counterfactual Reasoning Evaluation

Introduces DaoQL, a multimodal storage system that creates an explicit world model to reduce hallucinations and improve counterfactual reasoning in LLMs.

AI/ML arXiv cs.AI

Lossless but Not Free: An Empirical Anatomy of Speculative Decoding on Consumer Hardware

An empirical study on the performance and pitfalls of speculative decoding on consumer Apple-silicon hardware.

Software Engineering Hacker News

Linux kernel will support $ORIGIN, sort of

The Linux kernel is adding limited support for $ORIGIN to improve how shared libraries are located.