Ax Konstantinos Emmanouilidis, Lachlan MacDonald, Salma Tarmoun, Rene Vidal 7/1/2026

SGD at the Edge of Stability: Stochastic Stabilization with Large Learning Rates

Sharp convergence guarantees for SGD at the edge of stability with large learning rates in multiclass cross-entropy settings.

Ax Anish Acharya, Kris W Pan, Brian Verkhovsky 7/1/2026

RoPoLL: Robust Panel of LLM Judges

RoPoLL formalizes LLM jury evaluation under contamination models, analyzing robustness of panel-based LLM judges for consensus scoring.

Ax Long Kiu Chung, Shreyas Kousik 7/1/2026

ShardNet: Training Neural Controllers with Hard, Non-Convex Constraints

ShardNet: neural network architecture that strictly enforces hard non-convex safety constraints by construction using differentiable methods.

Ax Nicolaie Popescu-Bodorin, Madeleine Togher 7/1/2026

Neuro-Bayesian-Symbolic Residual Attention Shallow Network: Explainable Deep Learning for Cybersecurity Risk Assessment

NBS-RASN: hybrid neural architecture combining Bayesian and symbolic reasoning for explainable cybersecurity risk assessment in open-source systems.

Ax Jacob W. Toney, Samir Darouich, Yiran Wang, Aaron G. Garrison, Johannes K\"astner, Heather J. Kulik 7/1/2026

ElemeNet: Multiscale Molecular Machine Learning with Uncertainty Quantification Across the Periodic Table

ElemeNet: open-source ML software package for molecular property prediction with uncertainty quantification across chemical species.

Ax Chenyu Zhou, Qiliang Jiang, Shuning Wu, Xu Zhou 7/1/2026

Certified Speculative Execution for Untrusted AI Agents

Certificate-Gated Prefix Acceptance enables certified speculative execution for untrusted AI agents, combining speed with safety guarantees.

Ax Hyeonchang Jeon, Kyungbeom Kim, Eugene Vinitsky, Kyung-Joong Kim 7/1/2026

What Probing Reveals about Autonomous Driving: Linking Internal Prediction Errors to Ego Planning

Probing internal prediction errors in autonomous driving policies to link reasoning to ego vehicle planning and motion prediction.

Ax Mohammed Latif Siddiq, Md. Nafiu Rahman, Joanna C. S. Santos 7/1/2026

An Empirical Study of Security Calibration in Large Language Models for Code

First large-scale empirical study measuring security calibration in LLM-generated code across GPT-4o-mini, Gemini-2, and other models.

Ax Lingjie Chen, Yuanchen Bei, Haobo Xu, Yanjun Zhao, Yuzhong Chen, Hanghang Tong 7/1/2026

TAG-DLM: Diffusion Language Models for Text-Attributed Graph Learning

TAG-DLM unifies text-attributed graphs via diffusion language models for joint reasoning over text and topology.

Ax Zhichao Fan, Zexin Zhuang, Yanhang Li 7/1/2026

Probe Choice Changes Canary-Memorization Verdicts: Three Post-Hoc Disagreement Case Studies in a Text-Dominant LoRA-Tuned Autoregressive Testbed

Audits memorization probes on LoRA-tuned LLM testbed, finding probe choice significantly affects canary-memorization detection verdicts.

Ax Amos Alwala, Gabriel da Silva Lima, Wallace Moreira Bessa 7/1/2026

Machine Learning-based Feedback Linearization Control of Quadrotor Subject to Unmodeled Dynamics

Gaussian RBF neural network for feedback-linearization control of quadrotors with unmodeled nonlinear dynamics.

Ax Jongchan Choi, Nari Yang, Sung Soo Park, Jaemin Cho, Han Seoyoung, Haerin Shin, Jun-Hyung Park 7/1/2026

Can LLMs Imagine Moral Alternatives Beyond Binary Dilemmas?

MoralAltDataset evaluates whether LLMs can imagine moral alternatives beyond binary dilemmas as moral advisors and agents.

Ax Xueqiao Sun, Xiaohan Wang, Ludwig Schmidt, Serena Yeung-Levy, Yuhui Zhang 7/1/2026

Learning from Failure: Inference-Time Self-Improvement for Computer-Use Agents

Computer-use agents with MLLMs improve at inference time by learning from task failures without requiring large-scale trajectory collection.

Ax Hongliang Liu, Yuhao Wu, Tung-Ling Li 7/1/2026

The Decomposition Is the Fingerprint: Per-Component Identity for Agent Skills

Proposes locality-sensitive fingerprinting for identifying AI agent skills fetched from marketplaces to enable skill governance at runtime.

Ax Zhengxing Li, David J. Miller, Guangmingmei Yang, George Kesidis 7/1/2026

CSO-LLM: Class Subspace Orthogonalization for Post-Training Backdoor Detection and Trigger Inversion in LLMs

CSO-LLM method for post-training backdoor detection and trigger inversion in LLMs via class subspace orthogonalization.

Ax Gennady Andrienko, Natalia Andrienko 7/1/2026

From Idea to Prototype in an Afternoon: Scaffolded, AI-Assisted Rapid VA Prototyping

Describes using workflow language and AI assistant to reduce visual analytics prototyping from months to one afternoon.

Ax Seyed Alireza Molavi, Zhan Su, Yan Hu, Peyman Sheikholharam Mashhadi, Stefan Byttner, Prayag Tiwari 7/1/2026

Learning to Select, Not Relearn: Hard-Routed Mixtures of Reasoning LoRAs

Proposes hard-routed mixture-of-experts approach for composing LoRA adapters in LLMs for multi-domain adaptation without soft weighted combinations.

Ax Louis Mouchon 7/1/2026

Surprise as a Signal for Plasticity and Metacognition

Studies prediction-error signals in frozen encoders for plasticity gating and metacognition via episodic memory and offline replay.

Ax Mehmet Iscan 7/1/2026

Falsification, Not Exposure: An Internally Preregistered Placebo-Controlled Decomposition of Self-Repair Feedback in Frozen Small Code Models

Studies whether frozen small code models improve programs via self-repair feedback through falsification rather than re-exposure.

Ax Johan Land 7/1/2026

Modality-Driven Search with Holistic Trace Judging for ARC-AGI-2

Solver for ARC-AGI-2 visual reasoning using modality-driven search with diverse candidates across text and image modalities.

Ax Cl\'ement Fuchs, Tim Bary, Beno\^it Macq 7/1/2026

Localized Conformal Prediction for Image Classification with Vision-Language Models

Adapts conformal prediction sets for image classification with vision-language models using localized similarity to calibration examples.

Ax Amel Bennaceur, Gopi Krishnan Rajbahadur, Prince Mercy, Bashar Nuseibeh, Faeq Alrimawi 7/1/2026

From Failure to Alignment: A Requirements Engineering Framework for Machine Learning Systems

Proposes requirements engineering framework for ML systems focusing on stakeholder alignment and trustworthiness.

Ax Sayan Das, Bahram Yaghooti, Todd A. Kuffner, Soumendra N. Lahiri 7/1/2026

On Optimal Data Splitting for Split Conformal Prediction

Characterizes optimal data splitting for split conformal prediction to maximize statistical efficiency.

Ax Nikolai R\"ohrich, Julian Glei{\ss}ner, Ahmed H. A. Ibrahim, Silvan Mertes, Tobias Huber 7/1/2026

Preserve the Hard, Regenerate the Rest: Uncertainty-Guided Synthetic Training Data Augmentation with Diffusion Models

Uses uncertainty guidance and diffusion models to augment synthetic training data while preserving hard samples in semantic segmentation.

Ax Utsav Garg, Sungjin Hong, Jason Jung, Justin Lee, Shaan Desai, Joon Hee Kim, Anirudh Shrinivason, Edmond Wen, Susie Park 7/1/2026

Think in English, Answer in Korean: Efficient Adaptation of Multilingual Tool-Using Agents

Presents LuckyStar 111B, hybrid reasoning model for multilingual Korean-English enterprise agents with tool use via preamble conditioning.

Ax Haoming Liu, Yuanhe Guo, Yijia Cao, Shenji Wan, Hongyi Wen 7/1/2026

Histogram-constrained Image Generation

Proposes histogram-constrained diffusion models for fine-grained control over generative outputs.

Ax Rafael Rivera Soto, Anna Wegmann, Cristina Aggazzotti 7/1/2026

STEB: Style Text Embedding Benchmark

Introduces STEB, open-source benchmark standardizing evaluation of style embeddings across 96 datasets in 7 languages.

Ax Moreno D'Inc\`a, Massimiliano Mancini, Nicu Sebe 7/1/2026

Harnessing Textual Refusal Directions for Multimodal Safety

Improves multimodal LLM safety by extracting textual refusal directions from LLM backbone and transferring to MLLMs.

Ax Jiachen Cong, Jingbo Liu 7/1/2026

Accelerating Conformal Prediction via Approximate Leave-One-Out

Accelerates conformal prediction via approximate leave-one-out methods for faster uncertainty quantification.

Ax Ekaterina Alimaskina, Denis Shveykin, Gleb Molodtsov, Igor Shalygin, Alexey Kadeishvili, Aleksandr Beznosikov 7/1/2026

Self-Study Reconsidered: The Hidden Fragility of Learning from Self-Generated QA

Studies fragility of using synthetic QA pairs to fine-tune LLMs, showing generation is an implicit policy affecting training signal selection and quality.

Ax Sameer Malik, Ayush Singh, Amar Prakash Azad 7/1/2026

PolicyGuard: From Organizational Policies to Neuro-SymbolicCompliance Review Engines

PolicyGuard: neuro-symbolic framework converting organizational policies into interpretable compliance review engines using language models.

Ax Cesar Roder, Kajetan Schweighofer 7/1/2026

Automated Background Swapping for Robustness against Spurious Backgrounds

Automated background swapping technique to prevent deep learning classifiers from relying on spurious background correlations.

Ax Marcel Torne, Anubha Mahajan, Abhijnya Bhat, Chelsea Finn 7/1/2026

Freeform Preference Learning for Robotic Manipulation

FPL method for learning robot manipulation policies from freeform human preferences rather than binary comparisons.

Ax Zifan Carl Guo, Laura Ruis, Jacob Andreas, Belinda Z. Li 7/1/2026

Introspective Coupling: Self-Explanation Training Tracks Behavioral Change Despite Fixed Supervision

Study on language models generating faithful explanations via counterfactual supervision, tracking behavioral changes despite fixed training.

Ax Debmalya Mandal, Andi Nika, Parameswaran Kamalaruban, Adish Singla, Goran Radanovi\'c 7/1/2026

Corruption Robust Offline Reinforcement Learning with Human Feedback

RLHF algorithm robust to corrupted feedback/trajectories in offline setting for learning near-optimal policies with adversarial noise.

Ax Xuran Li, Hao Xue, Peng Wu, Xingjun Ma, Zhen Zhang, Huaming Chen, Flora D. Salim 7/1/2026

Perturbation Effects on Robustness and Individual Fairness

Study on adversarial perturbation effects on neural network robustness and individual fairness under semantic-preserving transformations.

Ax Anestis Kaimakamidis, Ioannis Mademlis, Ioannis Pitas 7/1/2026

Collaborative Knowledge Distillation via a Learning-by-Education Node Community

LENC framework for collaborative knowledge distillation enabling peer DNNs to dynamically adopt student/teacher roles for continual learning.

Ax Zeyu Jiang, Hai Huang, Xingquan Zuo 7/1/2026

Compositional Concept-Based Neuron-Level Interpretability for Deep Reinforcement Learning

Concept-based neuron-level interpretability method for understanding deep reinforcement learning policies in continuous state spaces.

Ax Vasilii Feofanov, Songkang Wen, Shifeng Xie, Simon Roschmann, Marius Alonso, Hongbo Guo, Romain Ilbert, Malik Tiomoko, Quentin Bouniot, Zeynep Akata, Lujia Pan, Jianfeng Zhang, Ievgen Redko 7/1/2026

Mantis: Lightweight Foundation Model for Time Series Classification

Mantis: transformer-based foundation model for time series classification pre-trained on synthetic data via self-supervised contrastive learning.

Ax Siow Meng Low, Ze Gong, Akshat Kumar 7/1/2026

TraCeS: Learning Per-Timestep Constraint-Violation Credit from Sparse Trajectory-Level Labels

TraCeS method for learning per-timestep constraint violation credit in safe reinforcement learning from sparse trajectory labels.

Ax Davide Prosperino, Haochun Ma, Christoph R\"ath 7/1/2026

Tailored minimal reservoir computing: on the bidirectional connection between nonlinearities in the reservoir and in data

Research on reservoir computing design using tunable nonlinearity parameters to match data nonlinearity for time series prediction.

Ax Mohammad Yaghini, Tudor Cebere, Michael Menart, Aur\'elien Bellet, Nicolas Papernot 7/1/2026

Private Rate-Constrained Optimization with Applications to Fair Learning

Differential privacy optimization for fair machine learning with rate constraints across subpopulations.

Ax Tommaso Marzi, Cesare Alippi, Andrea Cini 7/1/2026

Hierarchical Message-Passing Policies for Multi-Agent Reinforcement Learning

Research on hierarchical message-passing coordination for decentralized multi-agent reinforcement learning with temporal abstraction.

Ax Jan Tauberschmidt, Sophie Fellenz, Sebastian J. Vollmer, Andrew B. Duncan 7/1/2026

Physics-Constrained Fine-Tuning of Flow-Matching Models for Generation and Inverse Problems

Framework for fine-tuning flow-matching generative models with PDE constraints for physics-informed generation and inverse problems.

Ax Gal Lifshitz, Shahar Zuler, Ori Fouks, Dan Raviv 7/1/2026

L-SR1: Learned Symmetric-Rank-One Preconditioning

Learned symmetric-rank-one optimizer combining deep learning with classical optimization for data-efficient training.

Ax Miko{\l}aj Janusz, Tomasz Wojnar, Yawei Li, Luca Benini, Kamil Adamczewski 7/1/2026

One Shot vs. Iterative: Rethinking Pruning Strategies for Model Compression

Empirical comparison of one-shot versus iterative neural network pruning strategies for model compression effectiveness.

Ax Rupert Mitchell, Kristian Kersting 7/1/2026

Multipole Semantic Attention: A Fast Approximation of Softmax Attention for Pretraining

MuSe: Fast softmax attention approximation using query-key clustering for efficient long-sequence transformer pretraining on code and documents.

Ax Santosh Patapati, Ian Noronha 7/1/2026

Graph Coloring for Multi-Task Learning

SON-GOKU scheduler using graph coloring to resolve gradient conflicts in multi-task learning by partitioning tasks.

Ax Brian Godwin Lim, Galvin Brice Lim, Renzo Roel Tan, Irwin King, Kazushi Ikeda 7/1/2026

Enhancing Graph Representations with Neighborhood-Contextualized Message-Passing

Graph neural network improvement using neighborhood-contextualized message passing to capture relational data features.

Ax Austin Feng, Marius Alonso, Ambroise Odonnat, Vasilii Feofanov, Ievgen Redko 7/1/2026

Optimal Self-Consistency for Efficient Reasoning with Large Language Models

Optimized self-consistency approach for efficient test-time inference in LLM chain-of-thought reasoning reducing computational cost.