Ax Tianyi Zhang, Mousumi Das, Abrar Anwar, Jesse Thomason, David Traum 29d ago

DiPS: Dialogue Policy Selection for High-Stakes Persuasion Agents

DiPS: Q-learning framework for dialogue policy selection in high-stakes persuasion agents, dynamically adapting strategies based on personality.

Ax Tingting Yu, Pei-Cing Huang, Chan Hsu, Chan-Tung Ku, Yihuang Kang 29d ago

ADVENT: LLM-Driven Automatic Predicate Invention for ILP

ADVENT: LLM-driven predicate invention mechanism for Inductive Logic Programming combining LLM abduction with Prolog deductive verification.

Ax Guoyang Xia, Fengfa Li, Hongjin Ji, Lei Ren, Fangxiang Feng, Kun Zhan, Yan Xie 29d ago

VLAFlow: A Unified Training Framework for Vision-Language-Action Models via Co-training and Future Latent Alignment

VLAFlow: Unified flow-matching framework for training vision-language-action models enabling controlled comparison of VLA objectives with robot manipulation datasets.

Ax Zhaoyan Sun, Shan Zhong, Daizhou Wen, Jiaxing Han, Guoliang Li, Ying Yan, Peng Zhang, Yu Su, Xiang Qi, Baolin Sun, Chengyuan Yang, Tao Fang, Huaiyu Ruan 29d ago

AgenticDataBench: A Comprehensive Benchmark for Data Agents

Comprehensive benchmark for LLM-based data agents automating data science workflows. Evaluates agent capabilities on heterogeneous datasets.

Ax Long Minh Bui, Tuan Anh Le Van, Tung Phi Duc, Phi Le Nguyen, Jana Doppa, Trong Nghia Hoang 29d ago

Model Merging as Probabilistic Inference in Fine-Tuning Parameter Space

Probabilistic framework for merging task-specific models into multitask solutions. Scores statistical utility of parameter updates across tasks.

Ax Nikolai Smolyanskiy 29d ago

Predicting Closed-Loop Performance of Latent World Models: Offline Checkpoint Selection for MPC and Model-Based RL Under Non-Markovian Rewards in LunarLander

Diagnostic framework for predicting closed-loop performance of world models in model-based RL without explicit validation metrics.

Ax Stefano Masini, Cecilia Viscardi, Michela Baccini 29d ago

Full Bayesian Reinforcement Learning via LF-IBIS

Bayesian reinforcement learning method addressing data scarcity through prior knowledge and belief updates in sequential decision-making.

Ax Jae-Ryung Hong, Ho-Joong Kim, Seong-Whan Lee 29d ago

ProCal: Inference-Time Proposal Calibration for Open-Vocabulary Object Detection

Open-vocabulary object detection calibration using frozen VLMs. Vision-language model application with limited novelty.

Ax Ahin Lee, Sehyun Yun, Taesik Gong 29d ago

EPnG: Adaptive Expert Prune-and-Grow for Parameter-Efficient MoE Fine-tuning

Adaptive expert pruning and growing for efficient MoE fine-tuning using LoRA. Parameter-efficient training for large models.

Ax Rodrigo Mendoza-Smith 29d ago

Expander Sparse Autoencoders: Parameter-Efficient Dictionaries for Mechanistic Interpretability

Expander sparse autoencoders for mechanistic interpretability with reduced parameters. ML research on interpretability and efficient dictionaries.

Ax Weiwei Xu, Xuanning Cui, Hengzhi Ye, Minghui Zhou 29d ago

Decoupling Code Complexity from Newcomer Participation: A Causal Study of AI Coding Agent Adoption in OSS

Causal study of AI coding agent adoption on open-source projects, analyzing impact on newcomer participation. Developer tools and OSS ecosystem.

Ax Yuanzhi Liu, Shousheng Zhao, Bo Zhou, Kongming Liang, Zhanyu Ma 29d ago

MMBench-Live: A Continuously Evolving Benchmark for Multimodal Models

Continuously evolving multimodal benchmark using multi-agent pipeline for VLM evaluation. Developer tools and evaluation frameworks.

Ax Xuan-Phi Nguyen, Shrey Pandit, Yiran Zhao, Semih Yavuz, Silvio Savarese, Shafiq Joty 29d ago

Mixture-of-Parallelisms: Towards Memory-Efficient Training Stack for Mixture-of-Experts Models

Memory-efficient training stack combining parallelism techniques for MoE models. ML research on scalable training infrastructure.

Ax Valentin J. J. Kreileder, Johannes Reisinger, Andreas Fischer 29d ago

Evaluating Chunking Strategies for Retrieval-Augmented Generation on Academic Texts

Evaluation of chunking strategies for RAG systems on academic texts using RAGAS framework. Directly relevant to LLM applications and retrieval techniques.

Ax Yongyi Ji, Jiaji Wang, Yi Zhou, Fuxiang Chen, Hongji Yang 29d ago

An Exploratory Study on LLM-Generated Code and Comments in Code Repositories

Empirical study of LLM-generated code and comments in real repositories, analyzing prevalence and quality concerns. Developer tools and LLM applications.

Ax Qi Lyu, Jiahua Dong, Baichen Liu, Xudong Wang, Mingfei Han, Yulun Zhang, Fahad Shahbaz Khan, Salman Khan, Lianqing Liu, Zhi Han 29d ago

SAB-LVLM: Significance-Aware Binarization for Large Vision-Language Models

Binarization technique for vision-language models reducing memory/latency for deployment. Model optimization for efficient inference.

Ax Yuriy Maksyuta, George Bredis, Ruslan Rakhimov, Daniil Gavrilov 29d ago

Rank-Then-Act: Reward-Free Control from Frame-Order Progress

Reward-free reinforcement learning from video using VLM as progress scorer and GRPO objective. AI agents and novel training approach.

Ax Yidan Xu, Xiangmin Han, Rundong Xue, Huihui Ye 29d ago

SABER: A Semantic-Aligned Brain Network Analysis Framework via Multi-scale Hypergraphs

Brain disease diagnosis framework integrating LLM semantics with brain connectivity analysis via hypergraphs. LLM application with healthcare focus.

Ax Baran Bingol, Bahaeddin Turkoglu 29d ago

TUDUM: A Turkish-Thinking Reasoning Pipeline for Qwen3.5-27B

Pipeline for adapting Qwen 27B model to perform reasoning in Turkish rather than English, addressing multilingual LLM reasoning.

Ax Javier Irigoyen, Roberto Daza, Francisco Jurado, Julian Fierrez, Ruben Tolosana, Alvaro Ortigosa, Enrique Blas, Aythami Morales 29d ago

AIriskEval-edu: New Dataset for Risk Assessment in AI-mediated K-12 Educational Explanations

Dataset of 1,639 K-12 science explanations with human and LLM-generated alternatives for training risk assessment auditors.

Ax Nicholas Tagliapietra, Gian Lorenzo Marchioni, Moritz Willig, Juergen Luettin, Lavdim Halilaj, Kristian Kersting 29d ago

CausalSteward: An Agentic Divide-Conquer-Combine Copilot for Causal Discovery

CausalSTeward agentic divide-conquer-combine system for causal discovery integrating prior knowledge to identify causal models from high-dimensional data.

Ax Peng Yun, Shouwang Huang, Hao Li, Jinxi Li, Jianan Wang, Bo Yang 29d ago

PhysMani: Physics-principled 3D World Model for Dynamic Object Manipulation

PhysMani framework combines physics-principled 3D Gaussian world model with action policy for dynamic object manipulation in embodied AI.

Ax Zhiren Gong, Zihao Zeng, Chau Yuen, Wei Yang Bryan Lim 29d ago

Conditional Co-Ablation: Recovering Self-Repair Backups in Transformer Circuits

Conditional co-ablation technique reveals transformer self-repair mechanisms where dormant backups activate after primary component ablation.

Ax Minjong Cheon 29d ago

Robust for the Wrong Reasons: The Representational Geometry of LLM Robustness to Science Skepticism

Analyzes representational geometry showing LLMs become robust to science skepticism through problematic mechanisms rather than genuine understanding.

Ax Jan Drchal 29d ago

Object Aligner: A Configurable JSON Schema Similarity Score for Graphs, Applied to LLM Prompt Optimization

Object Aligner provides configurable JSON schema similarity scoring for measuring LLM output structure alignment in tool calling and agentic systems.

Ax Sofiane Ouaari, Kevin Vorwalder, Nico Pfeifer 29d ago

Assessing VLM Reliability for Medical Image Quality Evaluation Under Corruption and Bias

Evaluates Vision-Language Model reliability for medical image quality assessment under corruption and bias conditions.

Ax Wenda Wang, Yihan Tong, Yuwei Hu, Zhewei Wei 29d ago

MolSight: A Graph-Aware Vision-Language Model for Unified Chemical Image Understanding

MolSight vision-language model combines molecular LLMs with graph-aware visual understanding for molecular structure and drug discovery tasks.

Ax Tasnim Shahriar 29d ago

Do Newer Lightweight CNNs Perform Better Under Resource Constraints? A Controlled Multigenerational Study of Architecture, Initialization, Training Budget, and Efficiency

Controlled study comparing nine lightweight CNN architectures across multiple datasets and hardware to assess efficiency claims.

Ax Shrikara Arun, Anjaly Parayil, Srikant Bharadwaj, Renee St. Amant, Victor R\"uhle 29d ago

Towards Load-Aware Prefill Deflection for Disaggregated LLM Serving

Proposes load-aware prefill deflection technique to improve disaggregated LLM serving efficiency by balancing prefill and decode GPU pools.

Ax Rheeya Uppaal, Seungwoo Lyu, Selina Sung, Junjie Hu 29d ago

OpenSafeIntent: Evaluating Intent-Calibrated Safe Completion Across Dual-Use Prompt Sets

OpenSafeIntent benchmark evaluates whether LLMs calibrate assistance appropriately across benign, dual-use, and malicious intent variants.

Ax Anna Chorna 29d ago

SPLIT: Cross-Lingual Empathy and Cultural Grounding in English and Ukrainian LLM Responses

SPLIT benchmark evaluates LLM cross-lingual empathy and cultural grounding in emotional-support contexts across English and Ukrainian.

Ax Prathamesh Patil, Arpit Jain, Aswanth Krishnan 29d ago

Beyond the Performance Illusion: Structure-Aware Stratified Partitioning and Curriculum Distributionally Robust Optimization for Spatially Correlated Domains

Demonstrates performance evaluation failures in spatiotemporally correlated domains due to data leakage from non-i.i.d. splits.

Ax Florian Tambon, Michael Konstantinou, Cedric Richter, Charles Chenouard, Mark Harman, Mike Papadakis 29d ago

Prompt Coverage Adequacy

Proposes prompt coverage adequacy testing framework to guide LLM and autonomous agent testing when prompts replace traditional code.

Ax Mahmoud Abdelfattah, Hamid Nasiri, Peter Garraghan 29d ago

kNNGuard: Turning LLM Hidden Activations into a Training-Free Configurable Guardrail

kNNGuard presents training-free guardrail for LLMs using activation space of off-the-shelf models to detect unsafe/adversarial prompts with minimal labeled data.

Ax Tien-Huy Nguyen, Minh-Nhat Nguyen, Nguyen Nhat Huy, Hung Viet Nguyen, Huy Nguyen Minh Nhat, Thanh-Huy Nguyen, Cuong Tuan Nguyen, Hoang M. Le, Dat Nguyen, Phat Kim Huynh, Min Xu, Ulas Bagci 29d ago

ESC: Emotional Self-Correction for Reliable Vision-Language Models

Introduces emotional self-correction mechanism for vision-language models to improve reasoning reliability without post-training or engineered feedback.

Ax Liuhaichen Yang, Zhuang Jiang, Chenchao Sheng, Zezhi Tang 29d ago

Guided Action Flow: Q-Guided Inference for Flow-Matching Vision-Language-Action Policies

Proposes test-time guidance framework for vision-language-action policies using learned critic to guide flow-matching inference without retraining base models.

Ax Haoran Wang, Jinchuan Tian, Siddhant Arora, Shinji Watanabe 29d ago

An Efficient vLLM-Based Inference Pipeline for Unified Audio Understanding and Generation

Presents vLLM-based inference pipeline for unified audio understanding and generation in speech language models with multi-token prediction support.

Ax William Hackett, Peter Garraghan 29d ago

Behind the Refusal: Determining Guardrail Activation via Behavioral Monitoring

Develops behavioral monitoring techniques to detect and analyze guardrail activations in LLMs, enabling black-box security testing of production AI systems.

Ax Javier Irigoyen, Roberto Daza, Aythami Morales, Julian Fierrez, Ruben Tolosana, Ruben Vera-Rodriguez, Francisco Jurado, Alvaro Ortigosa 29d ago

Overview of Risk Assessment and Management for Intelligent Systems under the AI Act and Beyond

Reviews AI risk assessment and management methodologies under EU AI Act and other regulatory frameworks, covering identification, analysis, and mitigation approaches.

Ax Nathan Hughes, Ibrahim Habli 29d ago

What Types of Human-AI Teams Exist?

Systematically categorizes 53 human-AI team studies into five clusters using psychological teaming taxonomies to understand collaboration patterns and diversity.

Ax Gemma Galdon Clavell, Pablo Accuosto, Usman Gohar 29d ago

The Eticas AI Risk Taxonomy: Open Infrastructure for Operationalizing AI Audits

Presents open infrastructure for operationalizing AI audits by moving beyond taxonomies to actionable tests. Reviews 74 existing risk taxonomies and proposes executable evaluation framework.

Ax Haokun Liu, Zhaoqi Ma, Yicheng Chen, Wentao Zhang, Masaki Kitagawa, Zicen Xiong, Jinjie Li, Moju Zhao 29d ago

CoFL-S: Spatially Queryable Sector Flow Fields for Local Language-Conditioned Navigation

Proposes CoFL-S framework for vision-language navigation using language-conditioned flow fields for low-level robot action generation and trajectory planning.

Ax A. Seza Do\u{g}ru\"oz, Xixian Liao, Verena Blaschke, Jakob Prange, Senyu Li, David Ifeoluwa Adelani 29d ago

Challenges and Recommendations for LLMs-as-a-Judge in Multilingual Settings and Low-Resource Languages

Analyzes limitations of LLM-as-a-Judge evaluation paradigm for multilingual and low-resource language tasks, showing proficiency gaps and validation challenges.

Ax Ziyun Qiao, Yue Min, Ruining Chen, Yujun Li 29d ago

HERMES: A Multi-Granularity Labeling Substrate for Pre-training Data Mixtures

Proposes HERMES, a hierarchical multi-granularity labeling system for pre-training data mixtures to improve flexibility in corpus partitioning and semantic organization.

Ax Rintaro Otsubo, Ryo Fujii, Reina Ishikawa, Taiki Kanaya, Kanta Sawafuji, Hiroki Kajita, Shigeki Sakai, Hideo Saito, Ryo Hachiuma 29d ago

AnyGroundBench: A Specialized-Domain Benchmark for Video Grounding in Vision-Language Models

Introduces specialized benchmark for evaluating vision-language models on rare concepts and complex spatio-temporal video grounding beyond general datasets.

Ax Jinwei Hu, Yi Dong, Youcheng Sun, Xiaowei Huang 29d ago

SkillFuzz: Fuzzing Skill Composition for Implicit Intents Discovery in Open Skill Marketplaces

Research on security vulnerabilities in LLM-based agent skill marketplaces where benign skills can interact unexpectedly. Addresses fuzzing skill composition to discover implicit intent attacks.

Ax Batu Guan, Zirui Wang, Shaohua Li 29d ago

Understanding Agent-Based Patching of Compiler Missed Optimizations

Systematic study of AI agents patching compiler missed optimizations showing generalization beyond single cases is the key challenge.

Ax Cristian-Gabriel Florea, Stelian Sp\^inu 29d ago

VisionAId: An Offline-First Multimodal Android Assistant for People with Visual Impairment, Featuring Personalized Object Retrieval

Offline-first Android assistant application for people with visual impairment using multimodal models with personalized object retrieval.

Ax Gawon Seo, Dongwon Kim, Suha Kwak 29d ago

ACID: Action Consistency via Inverse Dynamics for Planning with World Models

ACID method uses inverse dynamics to verify trajectory realizability during decision-time planning with world models for embodied control.

Ax Zhuowei Chen, Liwei Chen, Christian Schunn, Raquel Coelho, Xiang Lorraine Li 29d ago

Neuron-Aware Active Few-Shot Learning for LLMs

Neuron-aware active learning method for LLMs identifies valuable unlabeled samples for annotation reducing human labeling costs in few-shot adaptation.