AI

Meta AI Releases EUPE: A Compact Vision Encoder Family Under 100M Parameters That Rivals Specialist Models Across Image Understanding, Dense Prediction, and VLM Tasks

Automated Analysis of Global AI Safety Initiatives: A Taxonomy-Driven LLM Approach

Researchers developed an automated framework using large language models to compare AI safety policy documents across a shared taxonomy of activities. The study found that model choice significantly affects comparison outcomes, with some document pairs showing high disagreement across different LLMs, though human expert evaluation showed high inter-annotator agreement.

Towards the AI Historian: Agentic Information Extraction from Primary Sources

Researchers have introduced Chronos, an AI Historian tool that enables historians to convert image scans of primary sources into structured data through natural-language interactions. The first module is open-source and allows historians to adapt AI workflows for analyzing heterogeneous historical source materials without requiring fixed extraction pipelines.

Affording Process Auditability with QualAnalyzer: An Atomistic LLM Analysis Tool for Qualitative Research

Researchers have developed QualAnalyzer, an open-source Chrome extension that makes AI-assisted qualitative research more transparent by preserving detailed audit trails of LLM analysis processes. The tool processes data segments independently and maintains records of prompts, inputs, and outputs to enable systematic comparison between AI and human judgments.

LLM-Agent-based Social Simulation for Attitude Diffusion

Researchers have developed discourse_simulator, an open-source Python framework that combines large language models with agent-based modeling to simulate how public attitudes change over time in response to real-world events. The framework models social media interactions and opinion dynamics through AI agents in social networks, offering a new tool for social science research on attitude polarization and belief evolution.

CODE-GEN: A Human-in-the-Loop RAG-Based Agentic AI System for Multiple-Choice Question Generation

Researchers developed CODE-GEN, a human-in-the-loop AI system that uses retrieval-augmented generation to create multiple-choice programming questions for educational purposes. The system achieved 79.9% to 98.6% success rates across seven pedagogical dimensions when evaluated by subject-matter experts, demonstrating strong performance in computational verification tasks while still requiring human expertise for complex instructional design.

A Model of Understanding in Deep Learning Systems

A new research paper proposes a model for understanding in deep learning systems, arguing that contemporary AI can achieve systematic understanding through internal models that track regularities and support reliable predictions. However, the research suggests this understanding falls short of scientific ideals due to symbolic misalignment and lack of explicit reductive properties.

Same World, Differently Given: History-Dependent Perceptual Reorganization in Artificial Agents

Researchers developed a minimal AI architecture where a 'perspective latent' creates history-dependent perception in artificial agents. The system allows identical observations to be processed differently based on accumulated experience, demonstrating measurable plasticity that persists even after conditions return to normal.

BLK-Assist: A Methodological Framework for Artist-Led Co-Creation with Generative AI Models

Researchers have developed BLK-Assist, a modular framework that enables artists to fine-tune AI diffusion models using their own artwork while maintaining privacy and stylistic control. The system includes three components for concept generation, transparency-preserving assets, and high-resolution outputs, demonstrating a consent-based approach to human-AI collaboration in creative work.

Toward Artificial Intelligence Enabled Earth System Coupling

This research review explores how artificial intelligence techniques can enhance Earth system modeling by improving coupling between physical, chemical, and biological processes across Earth's spheres. The study focuses on AI's potential to strengthen cross-domain interactions and create more unified Earth system frameworks beyond traditional climate models.

TreeGaussian: Tree-Guided Cascaded Contrastive Learning for Hierarchical Consistent 3D Gaussian Scene Segmentation and Understanding

TreeGaussian introduces a new framework for 3D scene understanding that uses tree-guided cascaded contrastive learning to better capture hierarchical semantic relationships in complex 3D environments. The method addresses limitations in existing 3D Gaussian Splatting approaches by implementing structured learning across object-part hierarchies and improving segmentation consistency.

Can LLMs Reason About Attention? Towards Zero-Shot Analysis of Multimodal Classroom Behavior

Researchers developed a privacy-preserving AI system that analyzes classroom videos to understand student engagement using pose detection and gaze tracking, with data processed by the QwQ-32B-Reasoning LLM. The system deletes original video frames and retains only geometric coordinates to comply with FERPA privacy regulations.

Measuring LLM Trust Allocation Across Conflicting Software Artifacts

Researchers developed TRACE, a framework to evaluate how LLMs allocate trust between conflicting software artifacts like code, documentation, and tests. The study found that current LLMs are better at identifying natural-language specification issues than detecting subtle code-level problems, with models showing systematic blind spots when implementations drift while documentation remains plausible.

AI Appeals Processor: A Deep Learning Approach to Automated Classification of Citizen Appeals in Government Services

Researchers developed an AI Appeals Processor that uses deep learning to automatically classify government citizen appeals, achieving 78% accuracy with Word2Vec+LSTM architecture. The system reduces processing time by 54% compared to traditional manual processing that averages 20 minutes per appeal with only 67% accuracy.

Fusion and Alignment Enhancement with Large Language Models for Tail-item Sequential Recommendation

Researchers propose FAERec, a new framework that uses large language models to improve sequential recommendation systems for rarely-interacted (tail) items. The system addresses fusion and alignment challenges between collaborative signals and semantic knowledge to enhance recommendation accuracy.

When Models Know More Than They Say: Probing Analogical Reasoning in LLMs

Researchers found that large language models (LLMs) have an asymmetry between their internal knowledge and prompted responses when detecting analogies. While probing reveals models understand rhetorical analogies better than their prompted responses suggest, both methods perform poorly on narrative analogies requiring deeper abstraction.

Gram-Anchored Prompt Learning for Vision-Language Models via Second-Order Statistics

Researchers propose Gram-Anchored Prompt Learning (GAPL), a new framework that improves Vision-Language Model adaptation by incorporating second-order statistical features via Gram matrices. This approach enhances robustness against domain shifts and local noise compared to existing methods that rely solely on first-order spatial features.

Toward a Sustainable Software Architecture Community: Evaluating ICSA's Environmental Impact

A study presents the first systematic audit of carbon footprint from GenAI usage in software architecture research and IEEE ICSA conference activities. The research provides two carbon inventories examining both AI inference usage in research papers and traditional conference operations including travel and venue energy consumption.

Effects of Generative AI Errors on User Reliance Across Task Difficulty

Researchers conducted an experimental study on user reliance on AI systems with varying error rates (10%, 30%, 50%) across easy and hard diagram generation tasks. The study found that while more errors reduce AI usage, users are not significantly more averse to AI failures on easy tasks versus hard tasks, challenging assumptions about how people react to AI's 'jagged frontier' of capabilities.

Discrete Prototypical Memories for Federated Time Series Foundation Models

Researchers propose FeDPM, a federated learning framework that addresses semantic misalignment issues when using Large Language Models for time series analysis. The system uses discrete prototypical memories to better handle cross-domain time-series data while preserving privacy in distributed settings.

Paper Espresso: From Paper Overload to Research Insight

Paper Espresso is an open-source platform that uses large language models to automatically discover, summarize, and analyze trending arXiv papers to help researchers manage information overload. Over 35 months, it has processed over 13,300 papers and revealed key trends in AI research, including a surge in reinforcement learning for LLM reasoning and strong correlation between topic novelty and community engagement.

An AI Teaching Assistant for Motion Picture Engineering

Researchers at Trinity College Dublin implemented an AI Teaching Assistant using Retrieval Augmented Generation for a Motion Picture Engineering course, testing it with 43 students over 7 weeks. The study found students rated the AI-TA as beneficial (4.22/5) but preferred human tutoring, while exam performance remained unchanged when AI-TA access was allowed.

Artificial Intelligence and Cost Reduction in Public Higher Education: A Scoping Review of Emerging Evidence

A scoping review of 241 academic records found that AI applications in public higher education can reduce costs through automation, resource optimization, and personalized learning, while also identifying implementation barriers and digital divide concerns. The research analyzed 21 empirical studies to examine how AI tools like ChatGPT and predictive analytics impact educational efficiency and accessibility.

InfBaGel: Human-Object-Scene Interaction Generation with Dynamic Perception and Iterative Refinement

Strengthening Human-Centric Chain-of-Thought Reasoning Integrity in LLMs via a Structured Prompt Framework