AI News

MulRobBench: A Decision-Level Benchmark for Safe and Security-Policy-Compliant Multimodal UAV Agents (arxiv.org)

Where Quality Breaks in Compressed Short-Text Generation: Staged Bottleneck Localization (arxiv.org)

DomainPilot: Domain-Level Loss-Guided Two-Stage Data Mixture Optimization for Efficient Language Model Fine-Tuning (arxiv.org)

Context Is King: How In-Context Specification Shapes the Geometry of Concepts (arxiv.org)

Probabilistic Symbolic Regression for Equation Discovery via Operator-induced and Regularized Symbolic Forests (arxiv.org)

Real-Time Semantic Segmentation with Optimized RetinaNet Architectures for Embedded Automotive Systems (arxiv.org)

Imprompt: A Language Framework for Prompt Programming (arxiv.org)

Adapter Merging Reactivates Latent Reasoning Traces: A Mechanism Analysis (arxiv.org)

DOSA: A Tree-Guided, Self-Regressive Framework for Long Document Structure Analysis (arxiv.org)

Parallel Swin Transformer-Enhanced 3D MRI-to-CT Synthesis for MRI-Only Radiotherapy Planning (arxiv.org)

When Rates Are Geometric: Rate-Certificate Transfer for Contact Splittings in Optimization (arxiv.org)

DECAF: De-Clustering for Adaptive Representational Unlearning (arxiv.org)

Low-Rank Dependence Decomposition via Accelerated Symmetric Non-negative Matrix Factorization (arxiv.org)

Learning to Access Computation: Accessibility Plasticity as a Principle of Adaptive Intelligence (arxiv.org)

Spatial Reasoning in LLM Game Agents: Impact of Causal Context and Multi-Step Planning (arxiv.org)

Recovery of latent inner products from an anisotropic Gaussian random geometric graph (arxiv.org)

WorldDiT: A Unified Diffusion Architecture for World and Action Modeling (arxiv.org)

SeT-Diff: Towards Semantic Foundation Models for HPC Telemetry and Time-Series (arxiv.org)

IndicTalk: A Large-Scale Persona-Based Multilingual Conversational Corpus for Indic Languages (arxiv.org)

Greedy dynamical meta-learning (arxiv.org)

SMART: LLM-Augmented Hybrid Retrieval for Dynamic Product Ads (arxiv.org)

Physics-Informed Neural Networks for Predicting Nitrous Oxide Flux (arxiv.org)

X-Stage: An Overlooked Pipeline Stage for Communication-Computation Overlap in DiT Inference (arxiv.org)

Bayesian Feature Extraction using Gaussian and Diffused-gamma Priors for High Dimensional Spatio-Temporal Data (arxiv.org)

Attention-Guided Layer Selection for Contrastive Decoding in Large Language Models (arxiv.org)

FMOPF: Latent Flow Matching with Constraint-Aware Interaction Priors for AC Optimal Power Flow (arxiv.org)

DraftExpert: Expansion-Aware Self-Speculative Decoding for End-Device MoE Inference (arxiv.org)

Physics Transformer: Tailoring Transformer for General PDE Prediction (arxiv.org)

Co-Harness: Co-Evolving Harnesses and Model Weights for LLM Agents (arxiv.org)

Harnessing X-ray Absorption Spectroscopy Data through Multimodal Mining of Battery Literature (arxiv.org)

Variational Boosting for Physics-Informed Neural Networks (arxiv.org)

What CLIP Knows but Cannot Say: Recovering Negation from Frozen Intermediate Features (arxiv.org)

Beyond Local Inspection: Global, Guideline-Grounded Evaluation of Post-hoc XAI Methods for ECG Classification (arxiv.org)

Domain-Prior-Regularized Graph Modeling for Anomaly Detection in Cyber-Physical Systems (arxiv.org)

Smooth Learning with Hard Constraints via Legendre-Regularized Policies (arxiv.org)

Breaking the Total Variance Barrier: Sharp Sample Complexity for Linear Heteroscedastic Bandits with Fixed Action Set (arxiv.org)

Progress-conditioned Group Policy Optimization for Long-Horizon Agentic Tasks (arxiv.org)

Escaping the Euclidean Void: Manifold-Informed Flow Matching for Sequential Recommendation (arxiv.org)

Stress-testing large language model agents in a robotic chemistry laboratory (arxiv.org)

DeepFaith: Evidence-Grounded LLMs for Faithful Incident Reporting in Multi-Stage APT Defense (arxiv.org)

An Integrated Deep Learning and Statistical Framework for Whole-Network Gene--Environment Association with Leaf Vascular Architecture (arxiv.org)

Exploration of the generative capabilities of Boltzmann machines applied to social systems under the majority rule (arxiv.org)

The Intruder Threshold: A Spectral Law for LoRA Fine-Tuning (arxiv.org)

Kalypso: Relational LLM Serving (arxiv.org)

A Multi-stage Constrained Optimization Framework for Data-driven Problems (arxiv.org)

Semalith v1.4: A Calibrated 184M Safety Classifier Achieving State-of-the-Art Prompt-Injection Detection at 44x Fewer Parameters than Llama-Guard-3-8B (arxiv.org)

Beyond a Global Norm: Personalizing Toxicity Sensitivity in Language Models Without Retraining (arxiv.org)

A Statistical Difference between Single-Layer Learning and Hierarchical Learning in Wide Neural Networks (arxiv.org)

RoleMix: Unifying Sequential and Non-Sequential Features via Semantic Tokenization for Post-Click Conversion Rate Prediction (arxiv.org)

DuoAD: Leveraging [CLS] Dual Characteristics for Training-Free Few-Shot Anomaly Detection (arxiv.org)