Ax Konstantin Krestnikov 27d ago

Truth as a Compression Artifact in Language Model Training

Controlled experiments showing LMs prefer correct answers because error compressibility structure guides learning, not inherent truth preference.

Ax Ninghui Li, Kaiyuan Zhang, Kyle Polley, Jerry Ma 27d ago

Security Considerations for Artificial Intelligence Agents

Perplexity's recommendations on security considerations for frontier AI agents based on operating agentic systems at scale.

Ax Siddharth Srikanth, Freddie Liang, Ya-Chuan Hsu, Varun Bhatt, Shihan Zhao, Henry Chen, Bryon Tjanaka, Minjune Hwang, Akanksha Saran, Daniel Seita, Aaquib Tabrez, Stefanos Nikolaidis 27d ago

Red-Teaming Vision-Language-Action Models via Quality Diversity Prompt Generation for Robust Robot Policies

Quality diversity optimization for red-teaming vision-language-action robot models to improve robustness against prompt variations.

Ax Angelika Romanou, Mark Ibrahim, Candace Ross, Chantal Shaib, Kerem Oktar, Samuel J. Bell, Anaelia Ovalle, Jesse Dodge, Antoine Bosselut, Koustuv Sinha, Adina Williams 27d ago

Brittlebench: Quantifying LLM robustness via prompt sensitivity

Brittlebench framework quantifying LLM robustness through prompt sensitivity evaluation beyond static benchmarks.

Ax Kushal Khemani (Independent Researcher, India), Anjum Nazir Qureshi (Rajiv Gandhi College of Engineering Research,Technology) 27d ago

AI-Driven Predictive Maintenance with Environmental Context Integration for Connected Vehicles: Simulation, Benchmarking, and Field Validation

Contextual data fusion framework integrating vehicle sensors with environmental signals for predictive maintenance in connected vehicles.

Ax Shidong He, Haoyu Wang, Wenjie Luo 27d ago

Generate Then Correct: Single Shot Global Correction for Aspect Sentiment Quad Prediction

Generates then corrects predictions for aspect sentiment quad prediction in fine-grained opinion mining tasks.

Ax Zhexi Lian, Haoran Wang, Xuerun Yan, Weimeng Lin, Xianhong Zhang, Yongyu Chen, Jia Hu 27d ago

Fine-tuning is Not Enough: A Parallel Framework for Collaborative Imitation and Reinforcement Learning in End-to-end Autonomous Driving

Proposes parallel framework combining imitation and reinforcement learning for end-to-end autonomous driving instead of sequential fine-tuning.

Ax Panayiotis Panayiotou, \"Ozg\"ur \c{S}im\c{s}ek 27d ago

Causal Discovery in Action: Learning Chain-Reaction Mechanisms from Interventions

Studies causal discovery in chain-reaction dynamical systems using interventional data with identifiability guarantees.

Ax Benjamin Lange 27d ago

Unilateral Relationship Revision Power in Human-AI Companion Interaction

Philosophical analysis of moral dimensions in human-AI companion interactions and provider control structures.

Ax Dogan Urgun, Gokhan Gungor 27d ago

Large Language Model Guided Incentive Aware Reward Design for Cooperative Multi-Agent Reinforcement Learning

Framework using LLMs to automatically synthesize reward programs for cooperative multi-agent reinforcement learning systems.

Ax Xuepeng Jing, Wenhuan Lu, Hao Meng, Zhizhi Yu, Jianguo Wei 27d ago

TIGFlow-GRPO: Trajectory Forecasting via Interaction-Aware Flow Matching and Reward-Guided Optimization

Combines flow matching with reward optimization for trajectory forecasting in autonomous driving and crowd surveillance scenarios.

Ax Peiyuan Jiang, Yao Liu, Yanglei Gan, Jiaye Yang, Lu Liu, Daibing Yao, Qiao Liu 27d ago

MuDD: A Multimodal Deception Detection Dataset and GSR-Guided Progressive Distillation for Non-Contact Deception Detection

Proposes multimodal deception detection dataset using GSR-guided distillation to improve non-contact deception detection.

Ax Yoseph Berhanu Alebachew, Hunter Leary, Swanand Vaishampayan, Chris Brown 27d ago

Beyond Code Snippets: Benchmarking LLMs on Repository-Level Question Answering

Introduces StackRepoQA, a repository-level QA benchmark for evaluating LLMs on multi-file program comprehension tasks beyond isolated code snippets.

Ax Marcin Abram 27d ago

Toward Evaluation Frameworks for Multi-Agent Scientific AI Systems

Analyzes challenges in benchmarking multi-agent scientific AI systems, including reasoning vs retrieval, data contamination, ground truth, tool use, and reproducibility in evolving knowledge bases.

Ax Ousmane Tom Bechir, Ad\'an Jos\'e-Garc\'ia, Zaineb Chelly Garcia, Vincent Sobanski, Clarisse Dhaenens 27d ago

A Firefly Algorithm for Mixed-Variable Optimization Based on Hybrid Distance Modeling

Firefly Algorithm adaptation for optimization problems with mixed continuous, ordinal, and categorical variables.

Ax Robert Aufschl\"ager, Jakob Folz, Gautam Savaliya, Manjitha D Vidanalage, Michael Heigl, Martin Schramm 27d ago

Towards Context-Aware Image Anonymization with Multi-Agent Reasoning

CAIAMAR framework uses multi-agent reasoning for context-aware anonymization of personally identifiable information in street-level imagery.

Ax Yufei Xu, Fanxu Meng, Fan Jiang, Yuxuan Wang, Ruijie Zhou, Zhaohui Wang, Jiexi Wu, Zhixin Pan, Xiaojuan Tang, Wenjie Pei, Tongxuan Liu, Di Yin, Xing Sun, Muhan Zhang 27d ago

HISA: Efficient Hierarchical Indexing for Fine-Grained Sparse Attention

Hierarchical indexing method HISA optimizes sparse attention mechanisms in LLMs by reducing indexer bottlenecks in token selection.

Ax Qing Lyu, Jianxu Wang, Jeremy Hudson, Ge Wang, Chirstopher T. Whitlow 27d ago

MRI-to-CT synthesis using drifting models

Medical imaging technique using diffusion models to synthesize CT images from MRI for pelvic imaging without ionizing radiation.

Ax Ya Zhou, Tianxiang Hao, Ziyi Cai, Haojie Zhu, Kejun He, Jia Liu, Xiaohan Fan, Jing Yuan 27d ago

Detecting low left ventricular ejection fraction from ECG using an interpretable and scalable predictor-driven framework

Machine learning framework for detecting low left ventricular ejection fraction from ECG. Emphasizes interpretability and scalability over black-box models.

Ax Joonhyung Bae 27d ago

ASTRA: Mapping Art-Technology Institutions via Conceptual Axes, Text Embeddings, and Unsupervised Clustering

ASTRA taxonomy for art-technology institutions. Uses text embeddings and clustering to map global landscape of art-tech organizations.

Ax Haiyue Song, Masao Utiyama 27d ago

OptiMer: Optimal Distribution Vector Merging Is Better than Data Mixing for Continual Pre-Training

OptiMer method for continual pre-training of LLMs. Decouples data mixture ratio selection from training by optimizing distribution vectors.

Ax Ivan Pasichnyk 27d ago

Beta-Scheduling: Momentum from Critical Damping as a Diagnostic and Correction Tool for Neural Network Training

Beta-scheduling for neural network optimization. Derives time-varying momentum from critically damped harmonic oscillator physics.

Ax Zichao Wei 27d ago

On the Mirage of Long-Range Dependency, with an Application to Integer Multiplication

Analysis of long-range dependency in neural networks for integer multiplication. Argues dependency is a computational artifact, not an intrinsic problem.

Ax Kavindu Herath, Joshua Zhao, Saurabh Bagchi 27d ago

Beyond Corner Patches: Semantics-Aware Backdoor Attack in Federated Learning

Backdoor attacks on federated learning with realistic semantic triggers. Proposes SABLE method using in-distribution patterns instead of synthetic corner patches.

Ax Aiman Al Masoud, Antony Anju, Marco Arazzi, Mert Cihangiroglu, Vignesh Kumar Kembu, Serena Nicolazzo, Antonino Nocera, Vinod P., Saraga Sakthidharan 27d ago

Security in LLM-as-a-Judge: A Comprehensive SoK

Systematization of knowledge on security and reliability risks in LLM-as-a-Judge paradigm. Documents vulnerabilities where judges become targets of adversarial manipulation.

Ax Gabriel U. Talasso, Meghdad Kurmanji, Allan M. de Souza, Nicholas D. Lane, Leandro A. Villas 27d ago

Task-Centric Personalized Federated Fine-Tuning of Language Models

Personalized federated learning approach for fine-tuning language models on heterogeneous tasks. Improves performance on diverse client tasks while maintaining privacy.

Ax KrishnaSaiReddy Patil 27d ago

RAGShield: Detecting Numerical Claim Manipulation in Government RAG Systems

Security evaluation of RAG systems in government applications. Demonstrates embedding-based defenses fail to detect subtle numerical claim manipulation in tax/benefits systems.

Ax Bj\"orn Roman Kohlberger (EctoSpace, Dublin, Ireland) 27d ago

Spectral Compact Training: Pre-Training Large Language Models via Permanent Truncated SVD and Stiefel QR Retraction

Spectral Compact Training (SCT) for LLMs on consumer hardware. Uses permanent truncated SVD factors to avoid materializing dense weight matrices during training.

Ax Ken M. Nakanishi 27d ago

Screening Is Enough

Multiscreen attention mechanism for language models. Introduces absolute query-key relevance to reject irrelevant keys, addressing softmax attention limitations.

Ax Xiaofan Zhou, Huy Nguyen, Bo Yu, Chenxi Liu, Lu Cheng 27d ago

Adaptive Stopping for Multi-Turn LLM Reasoning

Adaptive stopping mechanism for multi-turn LLM reasoning. Determines optimal stopping points for agents using retrieval-augmented generation and ReAct-style interactions.

Ax Zirui Zhao, Jun Hao Liew, Yan Yang, Wenzhuo Yang, Ziyang Luo, Doyen Sahoo, Silvio Savarese, Junnan Li 27d ago

GPA: Learning GUI Process Automation from Demonstrations

Vision-based robotic process automation (RPA) using sequential Monte Carlo localization. Enables stable GUI automation from single demonstrations with improved robustness.

Ax Dun Yuan, Fuyuan Lyu, Ye Yuan, Weixu Zhang, Bowei He, Jiayi Geng, Linfeng Du, Zipeng Sun, Yankai Chen, Changjiang Han, Jikun Kang, Alex Chen, Haolun Wu, Xue Liu 27d ago

Beyond Message Passing: A Semantic View of Agent Communication Protocols

Framework for analyzing agent communication protocols across three layers: communication, syntactic, and semantic. Systematically studies 18 representative protocols for LLM systems.

Ax Carmine Valentino, Federico Pichi, Francesco Colace, Dajana Conte, Gianluigi Rozza 27d ago

Integrating Artificial Intelligence, Physics, and Internet of Things: A Framework for Cultural Heritage Conservation

Framework integrating IoT and AI with physics knowledge for monitoring and maintenance of cultural heritage conservation.

Ax Xun Sun, Baiheng Xie, Li Huang, Qiang Gao 27d ago

Scaling DPPs for RAG: Density Meets Diversity

Method scaling determinantal point processes for RAG systems to improve diversity of retrieved context while maintaining relevance.

Ax Lin Wang, Junfeng Fang, Dan Zhang, Fei Shen, Xiang Wang, Tat-Seng Chua 27d ago

DRAFT: Task Decoupled Latent Reasoning for Agent Safety

Framework for monitoring safety of tool-using LLM agents through latent reasoning that decouples safety judgment into trainable stages.

Ax Genwei Ma, Ting Luo, Ping Yang, Xing Zhao 27d ago

General Explicit Network (GEN): A novel deep learning architecture for solving partial differential equations

Novel neural network architecture for solving PDEs addressing limitations of physics-informed neural networks.

Ax Maharshi Savdhariya 27d ago

NativeTernary: A Self-Delimiting Binary Encoding with Unary Run-Length Hierarchy Markers for Ternary Neural Network Weights, Structured Data, and General Computing Infrastructure

Binary encoding scheme for ternary neural network weights enabling efficient storage and computation for compressed LLMs.

Ax AbdulQoyum A. Olowookere, Usman A. Oguntola, Ebenezer. Leke Odekanle, Maridiyah A. Madehin, Aisha A. Adesope 27d ago

Towards Intelligent Energy Security: A Unified Spatio-Temporal and Graph Learning Framework for Scalable Electricity Theft Detection in Smart Grids

AI framework combining spatio-temporal and graph learning for electricity theft detection in smart grids.

Ax Bilal Khalid, Pedro Freire, Sergei K. Turitsyn, Jaroslaw E. Prilepsky 27d ago

Hardware-Oriented Inference Complexity of Kolmogorov-Arnold Networks

Analysis of computational efficiency for Kolmogorov-Arnold Networks on hardware-constrained deployment scenarios.

Ax Paul Saves, Matthieu Mastio, Nicolas Verstaevel, Benoit Gaudou 27d ago

From Model-Based Screening to Data-Driven Surrogates: A Multi-Stage Workflow for Exploring Stochastic Agent-Based Models

Multi-stage pipeline combining experimental design and machine learning surrogates to explore agent-based models efficiently.

Ax Yaxin Xu, Yue Zhou, Tianyu Zhao, Fengwei An, Zhixiang Ren 27d ago

The limits of bio-molecular modeling with large language models : a cross-scale evaluation

Cross-scale evaluation of LLMs on biomolecular modeling tasks revealing performance gaps compared to mechanistic understanding.

Ax Haotian Xiang, Bingcong Li, Qin Lu 27d ago

Scalable Variational Bayesian Fine-Tuning of LLMs via Orthogonalized Low-Rank Adapters

Bayesian fine-tuning method for LLMs using low-rank adapters to improve uncertainty quantification in safety-critical applications.

Ax Peng Zhang, Xuefeng Li, Xiaoqi Wang, Han-Wei Shen, Yifan Hu 27d ago

Beauty in the Eye of AI: Aligning LLMs and Vision Models with Human Aesthetics in Network Visualization

Using LLMs and vision models trained on human preferences to improve network visualization aesthetics beyond traditional heuristic metrics.

Ax Mohammadreza Rostami, Solmaz S. Kia 27d ago

Adaptive Threshold-Driven Continuous Greedy Method for Scalable Submodular Optimization

Submodular maximization algorithm with improved approximation guarantees for combinatorial optimization problems in sensing and resource allocation.

Ax Sribalaji C. Anand, George J. Pappas 27d ago

Adversarial Robustness of Deep State Space Models for Forecasting

Control-theoretic analysis of state-space model robustness under adversarial perturbations, examining Spacetime SSM forecasters and Kalman filter representations.

Ax Matthew Levinson 27d ago

MetaSAEs: Joint Training with a Decomposability Penalty Produces More Atomic Sparse Autoencoder Latents

MetaSAEs: sparse autoencoder training with decomposability penalty producing more atomic, single-concept latents for safety-relevant LLM applications.

Ax William Merrill, Yanhong Li, Tyler Romero, Anej Svete, Caia Costello, Pradeep Dasigi, Dirk Groeneveld, David Heineman, Bailey Kuehl, Nathan Lambert, Chuan Li, Kyle Lo, Saumya Malik, DJ Matusz, Benjamin Minixhofer, Jacob Morrison, Luca Soldaini, Finbarr Timbers, Pete Walsh, Noah A. Smith, Hannaneh Hajishirzi, Ashish Sabharwal 27d ago

Olmo Hybrid: From Theory to Practice and Back

OLMo Hybrid: theoretical and empirical analysis of hybrid models combining linear RNNs and attention as alternatives to pure transformers with scaling benefits.

Ax David Sewell, Xingjian Li, Stepan Tretiakov, Krishna Kumar, David Fridovich-Keil 27d ago

Neural Operators for Multi-Task Control and Adaptation

Neural operator methods for multi-task optimal control problems, mapping task descriptions to control policies using permutation-invariant architectures.

Ax Wenjing Gong, Udbhav Srivastava, Yuchen Wang, Yuhao Jia, Qifan Wu, Weishan Bai, Yifan Yang, Xiao Huang, Xinyue Ye 27d ago

Earth Embeddings Reveal Diverse Urban Signals from Space

Benchmark of Earth embedding models (AlphaEarth, Prithvi, Clay) for neighborhood-scale urban monitoring from satellite imagery.

Ax Daniel Jost, Luca Paparusso, Martin Stoll, J\"org Wagner, Raghu Rajan, Joschka B\"odecker 27d ago

Super Agents and Confounders: Influence of surrounding agents on vehicle trajectory prediction

Analysis of trajectory prediction models revealing that surrounding agents often degrade accuracy due to learned confounders, using Shapley attribution.