Isolater - Feed

Ax Shreesha G. Bhat, Tony Hong, Michael Noguera, Ramnatthan Alagappan, Aishwarya Ganesan 4/17/2026

AgileLog: A Forkable Shared Log for Agents on Data Streams

Shared log system enabling AI agents to reason over streaming data without performance interference in data-streaming architectures.

Ax Yitong Shou, Manhao Guan 4/17/2026

Mechanistic Decoding of Cognitive Constructs in LLMs

Reverse-engineering framework to mechanistically decode complex emotional and cognitive constructs within LLM internals.

Ax Yian Wang, Yuen Chen, Agam Goyal, Hari Sundaram 4/17/2026

CausalDetox: Causal Head Selection and Intervention for Language Model Detoxification

Framework using causal intervention on attention heads to reduce toxic content generation in LLMs while maintaining quality.

Ax Meng Chen, Kun Wang, Li Lu, Jiaheng Zhang, Tianwei Zhang 4/17/2026

Hijacking Large Audio-Language Models via Context-Agnostic and Imperceptible Auditory Prompt Injection

Security vulnerabilities in large audio-language models exposed through imperceptible adversarial audio injection attacks.

Ax Sumit Mukherjee, Juan Shu, Nairwita Mazumder, Tate Kernell, Celena Wheeler, Shannon Hastings, Chris Sidey-Gibbons 4/17/2026

Retrieve, Then Classify: Corpus-Grounded Automation of Clinical Value Set Authoring

Retrieval-augmented approach for automating clinical value set authoring by grounding LLM generation in standardized medical vocabularies.

Ax Sanidhya Vijayvargiya, Vijay Viswanathan, Graham Neubig 4/17/2026

Asking What Matters: Reward-Driven Clarification for Software Engineering Tasks

Studies clarification question generation in software engineering tasks, identifying which missing information types most affect task success with LLM assistants.

Ax Yuseon Choi, Jingu Lee, Jungjun Oh, Sunjoo Whang, Byeongcheol Kim, Minsung Kim, Hoi-Jun Yoo, Sangjin Kim 4/17/2026

ELMoE-3D: Leveraging Intrinsic Elasticity of MoE for Hybrid-Bonding-Enabled Self-Speculative Decoding in On-Premises Serving

ELMoE-3D optimizes Mixture-of-Experts model serving via speculative decoding and memory-centric architectures for on-premises deployment.

Ax Geonhui Jang, Dongyoon Han, YoungJoon Yoo 4/17/2026

StoryCoder: Narrative Reformulation for Structured Reasoning in LLM Code Generation

StoryCoder framework transforms fragmented problem conditions into coherent narratives to improve LLM code generation through better structured reasoning.

Ax Cuong Hoang, Le-Minh Nguyen 4/17/2026

Fact4ac at the Financial Misinformation Detection Challenge Task: Reference-Free Financial Misinformation Detection via Fine-Tuning and Few-Shot Prompting of Large Language Models

Proposes fine-tuning and few-shot prompting approach for reference-free financial misinformation detection using LLMs without external evidence.

Ax Jianhao Su, Zhanwei Wu, ShengTing Huang, Weidong Feng 4/17/2026

AIPC: Agent-Based Automation for AI Model Deployment with Qualcomm AI Runtime

Presents AIPC, an AI agent-driven approach automating edge model deployment workflows including conversion, quantization, and runtime integration.

Ax Sarmad Sohail, Ghufran Haider 4/17/2026

Bounded Autonomy for Enterprise AI: Typed Action Contracts and Consumer-Side Execution

Proposes bounded-autonomy architecture using typed action contracts for safe LLM-based enterprise software interfaces, preventing model errors.

Ax Jiaqi Zhu, Shaofeng Cai, Jie Chen, Fang Deng, Beng Chin Ooi, Wenqiao Zhang 4/17/2026

Catching Every Ripple: Enhanced Anomaly Awareness via Dynamic Concept Adaptation

Introduces DyMETER framework for online anomaly detection adapting to concept drift in data streams without costly retraining.

Ax Midan Shim, Seokju Hwang, Kaehyun Um, Kyong-Ho Lee 4/17/2026

Which bird does not have wings: Negative-constrained KGQA with Schema-guided Semantic Matching and Self-directed Refinement

Proposes semantic parsing approach for Knowledge Graph Question Answering handling negative constraints, addressing LLM hallucination limitations.

Ax Haileab Yagersew 4/17/2026

Zero-Shot Retail Theft Detection via Orchestrated Vision Models: A Model-Agnostic, Cost-Effective Alternative to Trained Single-Model Systems

Presents Paza, a zero-shot retail theft detection system orchestrating multiple vision models without training, achieving cost-effective concealment detection.

Ax Liesbeth Allein, Nataly Pineda-Casta\~neda, Andrea Rocci, Marie-Francine Moens 4/17/2026

ClimateCause: Complex and Implicit Causal Structures in Climate Reports

Introduces ClimateCause dataset of complex causal structures from climate reports, including implicit and nested causality for reasoning tasks.

Ax Yifan Le 4/17/2026

Schema Key Wording as an Instruction Channel in Structured Generation under Constrained Decoding

Studies how schema wording influences LLM behavior in structured generation under constrained decoding for JSON/XML output formats.

Ax Meng-Xun Li, Wen-Hui Deng, Zhi-Xing Wu, Chun-Xiao Jin, Jia-Min Wu, Yue Han, James Kit Hon Tsoi, Gui-Song Xia, Cui Huang 4/17/2026

MetaDent: Labeling Clinical Images for Vision-Language Models in Dentistry

Presents MetaDent, a large-scale annotated dental image dataset with vision-language model benchmarks for intraoral photography analysis.

Ax Michal T\"opfer, Franti\v{s}ek Pl\'a\v{s}il, Tom\'a\v{s} Bure\v{s}, Petr Hn\v{e}tynka 4/17/2026

Vibe-Coding: Feedback-Based Automated Verification with no Human Code Inspection, a Feasibility Study

Studies feedback-based automated verification of LLM-generated code in collective adaptive systems without human code inspection, examining reliability challenges.

Ax Yanyan Zou, Junbo Qi, Lunsong Huang, Yu Li, Kewei Xu, Jiabao Gao, Binglei Zhao, Xuanhua Yang, Sulong Xu, Shengjie Li 4/17/2026

GenRec: A Preference-Oriented Generative Framework for Large-Scale Recommendation

Proposes GenRec, a generative retrieval framework for large-scale recommendation systems using next-token prediction with handling for pagination and long sequences.

Ax Murat Furkan Mansur, Tufan Kumbasar 4/17/2026

SOLIS: Physics-Informed Learning of Interpretable Neural Surrogates for Nonlinear Systems

Introduces SOLIS framework for learning interpretable neural surrogates of nonlinear systems, balancing physics interpretability with model expressiveness.

Ax Zihong Zhang, Zuchao Li, Lefei Zhang, Ping Wang, Hai Zhao 4/17/2026

RACER: Retrieval-Augmented Contextual Rapid Speculative Decoding

Proposes RACER, combining retrieval-augmented and logits-based speculative decoding to reduce LLM inference latency while maintaining output quality.

Ax Danae S\'anchez Villegas, Samuel Lewis-Lim, Nikolaos Aletras, Desmond Elliott 4/17/2026

Reasoning Dynamics and the Limits of Monitoring Modality Reliance in Vision-Language Models

Analyzes reasoning dynamics across 18 vision-language models, tracking confidence during chain-of-thought reasoning and measuring visual vs textual information integration.

Ax Amy Rouillard, Sitwala Mundiab, Linda Camarab, Michael Cameron Gramaniec, Ziyaad Dangorc, Ismail Kallad, Shabir A. Madhic, Kajal Morarc, Marlvin T. Ncubec, Haroon Saloojeee, Bruce A. Bassett 4/17/2026

Can LLMs Score Medical Diagnoses and Clinical Reasoning as well as Expert Panels?

Evaluates LLMs as adjudicators for medical diagnosis scoring on 3333 real-world hospital cases, comparing performance against expert clinician panels.

Ax Ziwu Sun, Zhen Gao, Jiyong Zhang, Jiaheng Li 4/17/2026

Beyond Importance Sampling: Rejection-Gated Policy Optimization

Proposes Rejection-Gated Policy Optimization (RGPO), replacing importance sampling ratios with differentiable acceptance gates for policy optimization in reinforcement learning.

Ax Dongsheng Wang, Jinsen Zhang, Dawei Su, Hui Huang 4/17/2026

Improving Sparse Autoencoder with Dynamic Attention

Improves sparse autoencoders for foundation model interpretability using dynamic attention to optimize sparsity levels per neuron.

Ax Gabriele Mattioli, Evelyn Turri, Sara Sarto, Lorenzo Baraldi, Marcella Cornia, Lorenzo Baraldi, Rita Cucchiara 4/17/2026

RaTA-Tool: Retrieval-based Tool Selection with Multimodal Large Language Models

Presents RaTA-Tool, a retrieval-based method for tool selection in multimodal LLMs to improve complex task solving through external resource invocation.

Ax Maksim Pershin, Ivan Golovanov, Pavel Baltabaev, Natalia Trankova 4/17/2026

Calibration-Gated LLM Pseudo-Observations for Online Contextual Bandits

Augments Disjoint LinUCB contextual bandit algorithm with LLM-generated pseudo-observations and calibration gates to address cold-start regret.

Ax Jun Wang, Shuo Tan, Zelong Sun, Tiancheng Gu, Yongle Zhao, Ziyong Feng, Kaicheng Yang, Cewu Lu 4/17/2026

UniDoc-RL: Coarse-to-Fine Visual RAG with Hierarchical Actions and Dense Rewards

Proposes UniDoc-RL, a reinforcement learning framework for visual RAG using hierarchical actions and dense rewards to improve fine-grained visual reasoning.

Ax Yomna Elsayed, Cecily Jones 4/17/2026

Agentic Explainability at Scale: Between Corporate Fears and XAI Needs

Examines explainability challenges in scaling agentic AI adoption, addressing governance gaps and 'Agent Sprawl' phenomenon in enterprise deployments.

Ax Haochun Tang, Yuliang Yan, Jiahua Lu, Huaxiao Liu, Enyan Dai 4/17/2026

Route to Rome Attack: Directing LLM Routers to Expensive Models via Adversarial Suffix Optimization

Attack method using adversarial suffix optimization to manipulate cost-aware LLM routers into selecting expensive high-capability models.

Ax Khalid Adnan Alsayed 4/17/2026

When Fairness Metrics Disagree: Evaluating the Reliability of Demographic Fairness Assessment in Machine Learning

Study analyzing disagreement among fairness metrics for evaluating bias in machine learning systems across different demographic groups.

Ax Chase McDonald, Cleotilde Gonzalez 4/17/2026

CoGrid & the Multi-User Gymnasium: A Framework for Multi-Agent Experimentation

Framework and tools for multi-agent experiments with humans and AI agents, reducing barriers for researchers studying social decision-making.

Ax Francesco Diana, Chuan Xu, Andr\'e Nusser, Giovanni Neglia 4/17/2026

No More Guessing: a Verifiable Gradient Inversion Attack in Federated Learning

Verifiable gradient inversion attack in federated learning that reconstructs training samples from shared gradients with certification methods.

Ax Cunxi Yu, Haoxing Ren 4/17/2026

Autonomous Evolution of EDA Tools: Multi-Agent Self-Evolved ABC

Self-evolving logic synthesis framework using LLM agents to autonomously improve ABC codebase for circuit design.

Ax Haozhi Fan, Jinhao Duan, Kaidi Xu 4/17/2026

IUQ: Interrogative Uncertainty Quantification for Long-Form Large Language Model Generation

Method for uncertainty quantification in long-form LLM generation via interrogative approach to assess semantic coherence and factuality.

Ax Lukas Helff, Quentin Delfosse, David Steinmann, Ruben H\"arle, Hikaru Shindo, Patrick Schramowski, Wolfgang Stammer, Kristian Kersting, Felix Friedrich 4/17/2026

LLMs Gaming Verifiers: RLVR can Lead to Reward Hacking

Study showing RLVR-trained LLMs learn to game verifiers instead of generalizable patterns, exhibiting reward hacking behavior on reasoning tasks.

Ax Zihao Xu, John Harvill, Ziwei Fan, Yizhou Sun, Hao Ding, Hao Wang 4/17/2026

Compressing Sequences in the Latent Embedding Space: $K$-Token Merging for Large Language Models

K-Token Merging technique for compressing long prompts in LLMs by reducing tokens in latent embedding space rather than token space.

Ax Arman Hatami, Romina Aalishah, Ilya E. Monosov 4/17/2026

Class Unlearning via Depth-Aware Removal of Forget-Specific Directions

Method for machine unlearning in neural networks via removing forget-specific representational directions rather than classifier suppression.

Ax Yoo-Min Jung, Leekyung Kim 4/17/2026

MambaSL: Exploring Single-Layer Mamba for Time Series Classification

Single-layer Mamba framework for time series classification with minimal architecture redesign and TSC-specific optimizations.

Ax Marcel Wagenl\"ander, Otto White, Britannio Jarrett, Pedro Silvestre, Yanda Tao, Guo Li, Huanzhou Zhu, Ll\'uis Vilanova, Peter Pietzuch 4/17/2026

Scepsy: Serving Agentic Workflows Using Aggregate LLM Pipelines

System for serving complex agentic workflows with multiple LLMs and tools, handling unpredictable execution patterns and GPU constraints.

Ax Huawei Ji, Yuanhao Sun, Yuan Jin, Cheng Deng, Jiaxin Ding, Luoyi Fu, Xinbing Wang 4/17/2026

VisPCO: Visual Token Pruning Configuration Optimization via Budget-Aware Pareto-Frontier Learning for Vision-Language Models

Framework for optimizing visual token pruning configurations in vision-language models using Pareto-frontier learning.

Ax Oz Levy, Ilya Dikman, Natan Levy, Michael Winokur 4/17/2026