artificial intelligence

LightCRS: Lightweight Multi-Domain Conversational Recommendation System

This work proposes LightCRS, a lightweight, query-conditioned framework for unified multi-domain multimodal conversational recommendation and achieves state-of-the-art average HR@1, NDCG@5, and MRR, improving HR@1 by 35.4\% over the strongest baseline while reducing per-query latency by 266 times.

MAMath-Bench: Process-Aware and Robust Evaluation of Multimodal Multi-Agent Mathematical Reasoning

MAMath-Bench is introduced, a process-aware benchmark designed to evaluate the multimodal mathematical reasoning capabilities of MASs, and finds that multi-agent collaboration does not consistently improve final-answer accuracy over strong single LLMs, and current systems remain vulnerable to adversarial reasoning erro...

Fix, Not Just Resample: Bounded Compiler-Guided Repair in Lean 4

This work introduces a bounded, online step-level repair module that feeds the compiler's error message to a pretrained repair model to fix individual rejected steps in place and improves the solve rate over a sampling-only baseline across all tested candidate budgets.

Industrial Agentic AI: LLM-Driven New-Generation Agents for Smart Manufacturing

A structured reference for the design, development, and application of industrial agents is provided and a four-layer system architecture is proposed and detail its core modules and operational mechanisms for closed-loop task execution.

DataPraxis: Scaling Analytics Tasks via Verified Workspace Evolution

This work constructs DataPraxis-5K, a dataset of 5,000 verified trajectories spanning diverse domains and output formats, to train DataPraxis-27B and DataPraxis-35B-A3B, an analytics task synthesis framework based on verified workspace evolution.

MedPAO-Assist: Edge-Deployable, Modality-Aware Agentic AI for Radiology Report Structuring

MedPAO-Assist is presented, an edge-deployable agentic framework that generates structured radiology reports directly from radiologist dictation and voice commands, requiring only $\approx3.3$ GB peak RAM and achieving an average end-to-end latency of $\approx31$ seconds on a standard laptop CPU through an asynchronous...

Pareto-Optimal RTL Code Generation via Multi-Objective Reinforcement Learning with Large Language Models

This work is the first to propose a learning method for LLM-based circuit generation that optimizes, not the quality of a single circuit, but the quality and diversity of the entire circuit set obtained through multiple generations in the PPA objective space.

Large Language Model and Reinforcement Learning Based Autonomous Construction and Progress Updates

An enhanced soft actor-critic reinforcement learning algorithm is developed to plan actions for a mobile crane according to construction tasks and a large language model, ChatGPT 5, is used to update construction progress, provide safety information to control system, and generate construction progress reports.

Navalia: Advancing Cowork Agents through Verifiable Task Synthesis and Long-Horizon Post-Training

Navalia is introduced, a framework for Cowork Agents that combines verifiable task synthesis with long-horizon post-training post-training capabilities and generalizes to broader Cowork settings, including general office tasks.

The Geography of Artificial Intelligence Innovation in China

Artificial Human: Genuine Need as a Precondition for Genuine AGI

We argue that a system with substantial processing capacity can be a genuine artificial general intelligence (AGI) — an independent agent rather than a computing system — only if its processing is coupled to a genuine need. We call a system that satisfies this condition an artificial human. Our claims are three. First,...

ARTIFICIAL INTELLIGENCE IN GYNECOLOGICAL ULTRASONOGRAPHY: DIAGNOSTIC CAPABILITIES, CLINICAL VALIDATION, AND LIMITATIONS

Artificial intelligence (AI) is increasingly proposed as an adjunct to gynecological ultrasonography, including assessment of adnexal masses, endometrial lesions, adenomyosis, uterine leiomyomas, and reproductive imaging. However, model discrimination reported in curated datasets does not automatically establish safety...