AI/ML arXiv cs.AI

Using AI-based Learning Assistants in Higher Education: A Large-Scale Descriptive Analysis

A large-scale study analyzing the usage patterns of the Syntea AI learning assistant among over 77,000 higher education students.

AI/ML arXiv cs.AI

Ideas Have Genomes: Benchmarking Scientific Lineage Reasoning and Lineage-Grounded Idea Generation

Presentation of IG-Bench, a benchmark designed to test if LLMs can reason about the lineage and evolutionary history of scientific ideas.

AI/ML arXiv cs.AI

Towards the Explainability of Temporal Graph Networks via Memory Backtracking and Topological Attribution

A new method for explaining Temporal Graph Network (TGN) predictions using memory backtracking and topological attribution.

AI/ML arXiv cs.AI

LLT: Local Linear Transformer for PDE Operator Learning

Introduction of the Local Linear Transformer (LLT), which improves efficiency and accuracy for solving Partial Differential Equations (PDEs).

AI/ML arXiv cs.AI

ReCoLoRA: Spectrum-Aware Recursive Consolidation for Continual LLM Fine-Tuning

ReCoLoRA, a spectrum-aware recursive consolidation framework for continual fine-tuning of LLMs to prevent catastrophic forgetting.

AI/ML arXiv cs.AI

Omni-Sleep: A Sleep Foundation Model via Hierarchical Contrastive Learning of CNS--ANS Dynamic

Omni-Sleep, a foundation model for sleep physiology that utilizes hierarchical contrastive learning of CNS-ANS dynamics.

AI/ML arXiv cs.AI

Applying JEPA-Style Predictive Learning to JA4-Derived Network Fingerprints

Researchers developed JA4-JEPA, a Transformer-based model that uses JEPA-style predictive learning to create useful embeddings from JA4 network fingerprints for protocol-family classification.

AI/ML arXiv cs.AI

Drift-Aware Temporal Graph Rewiring (DATGR) for Adaptive Semantic Modeling in Biomedical Text

The DATGR framework introduces a drift-aware temporal graph rewiring method to update semantic models in biomedical text, improving link-prediction recall without retraining embeddings.

AI/ML arXiv cs.AI

AI-guided stimuli discovery and generation to optimize facial emotion perception studies in autism

A new model-guided framework uses AI to discover and generate stimuli that maximize behavioral differences in facial emotion perception studies for autistic individuals.

AI/ML arXiv cs.AI

CommuniWave:A Machine Learning Model for Quantifying the Degree of Temporary Informal Behavior in Urban Communities

CommuniWave is a machine learning model that quantifies informal behavior in urban communities using a combination of BCN, a custom YOLOv10 variant (YLX), and random forests.

AI/ML arXiv cs.AI

SHAP-Weighted Cross-Modal Expert Fusion for Emotion and Sentiment Recognition: Evidence and Limits

This paper analyzes XAI-guided adaptive fusion for multimodal emotion recognition, finding that sum-abs SHAP reduction effectively preserves attribution for high-dimensional cross-modal experts.

AI/ML arXiv cs.AI

Towards Precision Therapy in Hepatocellular Carcinoma: A Clinical-Reasoning LLM for Risk Stratification and Treatment Guidance

HCC-STAR is a clinically aligned LLM for hepatocellular carcinoma risk stratification and treatment guidance, showing superior performance over existing guidelines and other LLMs in multi-center tests.

AI/ML arXiv cs.AI

The complexities of patient-centred conversational artificial intelligence

Researchers developed a patient simulator to model diverse communication styles in health chatbots, highlighting how communication diversity affects triage outcomes in LLMs.

AI/ML arXiv cs.AI

Formal Mechanisms for Market Stability in Self-Interested Agent Societies: A Marketplace Simulation Study

A simulation study using DeepSeek-V3 agents found that a 'Mediation' mechanism provides stability and adversarial robustness in self-interested agent marketplaces.

AI/ML arXiv cs.AI

SolarChain-Eval: A Physics-Constrained Benchmark for Trustworthy Economic Agents in Decentralized Energy Markets

SolarChain-Eval is a physics-constrained benchmark for evaluating the trustworthiness and safety of autonomous economic agents in decentralized energy markets.

AI/ML arXiv cs.AI

Remember When It Matters: Proactive Memory Agent for Long-Horizon Agents

The Proactive Memory Agent is a plug-and-play module that actively injects decision-relevant reminders into long-horizon agent trajectories to prevent behavioral state decay.

AI/ML arXiv cs.AI

MobiDiff: Semantic-Aware Multi-Channel Discrete Diffusion for Human Mobility Data Generation

MobiDiff is a discrete diffusion framework designed for efficient and privacy-preserving human mobility data generation by denoising multi-channel semantic skeletons.

AI/ML arXiv cs.AI

FedOPAL: One-Shot Federated Learning via Analytic Visual Prompt Tuning

FedOPAL introduces a one-shot federated learning framework using analytic visual prompt tuning to handle heterogeneous data on the edge with zero server-side training costs.

AI/ML arXiv cs.AI

Towards Mechanistically Understanding Why Memorized Knowledge Fails to Generalize in Large Language Model Finetuning

Research into the 'Knowing--Using Gap' in LLMs reveals that memorized knowledge often fails to generalize because it is not routed to computation-effective layers.

AI/ML arXiv cs.AI

Game Theory Driven Multi-Agent Framework Mitigates Language Model Hallucination

G-Frame is a multi-agent framework using game theory to reduce LLM hallucinations in scientific domains, leading to the creation of the OmniChem model for chemistry.