Ax Vasilis Niarchos, Angelos Sirbu, Sokratis Trifinopoulos 5/11/2026

Neural Operators as Efficient Function Interpolators

Novel use of neural operators as function interpolators via auxiliary base-space composition.

Ax Xueying Ding, Leman Akoglu 5/11/2026

Toward Privileged Foundation Models:LUPI for Accelerated and Improved Learning

Framework integrating privileged information to accelerate training and improve generalization in tabular foundation models.

Ax Zhicheng Yang, Zhijiang Guo, Yifan Song, Minrui Xu, Yongxin Wang, Yiwei Wang, Xiaodan Liang, Jing Tang 5/11/2026

Prune-OPD: Efficient and Reliable On-Policy Distillation for Long-Horizon Reasoning

Efficient on-policy distillation method for long-horizon LLM reasoning that prunes diverged trajectories.

Ax Charlotte Peale, Siddartha Devic, Parikshit Gopalan, Udi Wieder, Aravind Gollakota 5/11/2026

Flexible Routing via Uncertainty Decomposition

Uncertainty-aware model routing system dynamically selecting between low-cost and expensive models based on query ambiguity.

Ax Yuxuan Lou, Yang You 5/11/2026

OrScale: Orthogonalised Optimization with Layer-Wise Trust-Ratio Scaling

Optimizer extension adding layer-wise trust-ratio scaling to orthogonalized matrix updates for neural network training.

Ax Oscar Davis, Anastasiia Filippova, Pierre Ablin, Victor Turrisi, Amitis Shidani, Marco Cuturi, Louis B\'ethune 5/11/2026

Scaling Categorical Flow Maps

Flow matching approach for discrete language modeling enabling faster sampling and flexible generation compared to autoregressive models.

Ax Subrat Prasad Panda, Blaise Genest, Arvind Easwaran 5/11/2026

Approximation-Free Differentiable Oblique Decision Trees

Differentiable training method for oblique decision trees with improved optimization and regularization for tabular data.

Ax Xingyue Huang, Louis Tichelman, Jinwoo Kim, Krzysztof Olejniczak, \.Ismail \.Ilkan Ceylan 5/11/2026

RelAgent: LLM Agents as Data Scientists for Relational Learning

LLM-based autonomous agent system for automated relational learning feature engineering and model selection.

Ax Hanzaleh Akbari Nodehi, Parsa Moradi, Soheil Mohajer, Mohammad Ali Maddah-Ali 5/11/2026

\mathsf{VISTA}: Decentralized Machine Learning in Adversary Dominated Environments

Decentralized ML framework for adversary-dominated environments using incentive-oriented robust aggregation.

Ax Aur\'elien Decelle, Alfonso de Jes\'us Navas G\'omez, Beatriz Seoane 5/11/2026

Distributional simplicity bias and effective convexity in Energy Based Models

Dynamical analysis of energy-based generative models through effective model theory and Fourier expansion.

Ax Usevalad Milasheuski, Piero Baraldi, Enrico Zio, Stefano Savazzi 5/11/2026

On the Tradeoffs of On-Device Generative Models in Federated Predictive Maintenance Systems

Analysis of generative models (VAE, GAN, diffusion) for anomaly detection in federated IoT predictive maintenance systems.

Ax Lucas Nerone Rillo, Zhanhong Jiang, Nastaran Saadati, Aditya Balu, Baskar Ganapathysubramanian, Chinmay Hegde, Soumik Sarkar 5/11/2026

ADKO: Agentic Decentralized Knowledge Optimization

Framework for decentralized multi-agent optimization using Gaussian Process surrogates and compact knowledge tokens.

Ax Minjae Oh, Sangjun Song, Gyubin Choi, Yunho Choi, Yohan Jo 5/11/2026

KL for a KL: On-Policy Distillation with Control Variate Baseline

Stabilized on-policy distillation method for LLM reasoning using control variate baseline to reduce gradient variance.

Ax Ozgu Goksu, Nicolas Pugeault 5/11/2026

Enhancing Federated Quadruplet Learning: Stochastic Client Selection and Embedding Stability Analysis

Federated learning approach using quadruplet learning with stochastic client selection for heterogeneous data.

Ax Aristotelis Ballas, Christos Diou 5/11/2026

Flatness and Gradient Alignment Are Both Necessary: Spectral-Aware Gradient-Aligned Exploration for Multi-Distribution Learning

Multi-distribution learning analysis showing both loss landscape flatness and gradient alignment are necessary for generalization improvement.

Ax Tue M. Cao, Hoang X. Nhat, Raed Alharbi, My T. Thai 5/11/2026

Tree SAE: Learning Hierarchical Feature Structures in Sparse Autoencoders

Sparse autoencoder method learning hierarchical feature structures without relying on activation coverage assumptions.

Ax Amin Karimi Monsefi, Dominic Culver, Nikhil Bhendawade, Manuel R. Ciosici, Yizhe Zhang, Irina Belousova 5/11/2026

Trajectory as the Teacher: Few-Step Discrete Flow Matching via Energy-Navigated Distillation

Discrete flow matching distillation for few-step text generation, using energy navigation to improve student model training trajectories.

Ax Xiao Tian, Jue Fan, Rachael Hwee Ling Sim, Bryan Kian Hsiang Low 5/11/2026

INO-SGD: Addressing Utility Imbalance under Individualized Differential Privacy

Optimization algorithm addressing utility imbalance under individualized differential privacy where data owners set heterogeneous privacy requirements.

Ax Atsushi Nitanda, Dake Bu, Yueming Lyu, Tanya Veeravalli 5/11/2026

Slowly Annealed Langevin Dynamics: Theory and Applications to Training-Free Guided Generation

Theoretical study of Slowly Annealed Langevin Dynamics sampler with applications to training-free guided generation using pretrained score models.

Ax Zhengkai Sun, Dibyakanti Kumar, Alejandro F Frangi, Anirbit Mukherjee, Mingfei Sun 5/11/2026

Convergent Stochastic Training of Attention and Understanding LoRA

Theoretical analysis of attention layer trainability with LoRA low-rank adaptation, establishing convergence guarantees under stochastic training.

Ax Hanlin Cai, Kai Li, Houtianfu Wang, Haofan Dong, Yichen Li, Falko Dressler, Ozgur B. Akan 5/11/2026

Graph Representation Learning Augmented Model Manipulation on Federated Fine-Tuning of LLMs

Privacy-preserving federated fine-tuning of LLMs using graph representation learning to detect and mitigate adversarial model manipulation attacks.

Ax Fabian Stricker, Jose A. Peregrina, David Bermbach, Christian Zirpins 5/11/2026

FLAM: Evaluating Model Performance with Aggregatable Measures in Federated Learning

Federated learning evaluation framework addressing metric aggregation challenges when assessing global model performance across distributed participants.

Ax Ahmad Aghapour, Erhan Bayraktar 5/11/2026

When Diffusion Model Can Ignore Dimension: An Entropy-Based Theory

Theoretical analysis of why diffusion models efficiently sample high-dimensional data using entropy-based convergence bounds beyond ambient dimension.

Ax Antoine Wehenkel, Michael Kagan, Lukas Heinrich, Chris Pollard 5/11/2026

It Just Takes Two: Scaling Amortized Inference to Large Sets

Neural posterior estimation for amortized inference on set-structured observations with shared factors, addressing high-dimensional conditioning problems.

Ax Seohyun Lee, Wenzhi Fang, Dong-Jun Han, Seyyedali Hosseinalipour, Christopher G. Brinton 5/11/2026

Self-Play Enhancement via Advantage-Weighted Refinement in Online Federated LLM Fine-Tuning with Real-Time Feedback

Feedback-based LLM fine-tuning system using advantage-weighted self-play in federated online settings without requiring ground-truth labels.

Ax Chris Elliott, Daniel Murfet 5/11/2026

Susceptibilities and Patterning: A Primer on Linear Response in Bayesian Learning

Theoretical framework interpreting neural networks through susceptibilities and Bayesian learning, connecting data perturbations to posterior covariances.

Ax Nicole Ma, Nick Rui 5/11/2026

Where's the Plan? Locating Latent Planning in Language Models with Lightweight Mechanistic Interventions

Mechanistic study identifying where planning representations form in LLMs using linear probing and activation patching across multiple model scales.

Ax Chris Elliott, Einar Urdshals, David Quarel, Daniel Murfet 5/11/2026

Interpreting Reinforcement Learning Agents with Susceptibilities

Susceptibilities technique for deep reinforcement learning interpretability studying how agent behaviors respond to loss perturbations.

Ax Zezheng Lin, Fengming Liu 5/11/2026

Position: Mechanistic Interpretability Must Disclose Identification Assumptions for Causal Claims

Position paper critiquing mechanistic interpretability papers for making causal claims without explicit identification assumptions.

Ax Ning Liu, Chuanneng Sun, Kristina Klinkner, Shervin Malmasi 5/11/2026

Beyond Pairs: Your Language Model is Secretly Optimizing a Preference Graph

Extension of DPO showing language models optimize preference graphs; proposes method to exploit rich preference structure beyond pairwise comparisons.

Ax Gugan Thoppe, L. A. Prashanth, Ankur Naskar, Sanjay Bhat 5/11/2026

Reinforcement Learning for Exponential Utility: Algorithms and Convergence in Discounted MDPs

Value-based RL algorithms for exponential-utility optimization in discounted MDPs with convergence analysis and Bellman-type equations.

Ax Zhexuan Wang, Xuebo Liu, Li Wang, Zifei Shan, Yutong Wang, Zhenxi Song, Min Zhang 5/11/2026

MASPO: Joint Prompt Optimization for LLM-based Multi-Agent Systems

MASPO: joint prompt optimization method for LLM-based multi-agent systems addressing misalignment between agent and system objectives.

Ax Alexandre Cristov\~ao Maiorano 5/11/2026

Evaluating Prompt Injection Defenses for Educational LLM Tutors: Security-Usability-Latency Trade-offs

Evaluation methodology for prompt-injection defenses in educational LLM tutors examining security-usability-latency trade-offs.

Ax Xiao Wang 5/11/2026

More Thinking, More Bias: Length-Driven Position Bias in Reasoning Models

Analysis of position bias in reasoning-tuned LLMs showing bias scales with reasoning trajectory length, not reduced by chain-of-thought.

Ax Jon-Paul Cacioli 5/11/2026

Domain-level metacognitive monitoring in frontier LLMs: A 33-model atlas

Study of 33 frontier LLMs showing domain-level variation in metacognitive monitoring across MMLU benchmarks using confidence calibration.

Ax Debashis Guha, Amritendu Mukherjee, Sanjay Kukreja, Tarun Kumar 5/11/2026

State Representation and Termination for Recursive Reasoning Systems

Framework for representing reasoning state as epistemic graph and determining termination conditions in recursive reasoning systems.

Ax Cameron Berg, Susan L. Schneider, Mark M. Bailey 5/11/2026

Hidden Coalitions in Multi-Agent AI: A Spectral Diagnostic from Internal Representations

Method for detecting hidden coalitions in multi-agent AI systems through spectral analysis of internal representations for safety alignment.

Ax Siyuan Guo, Yali Du, Hechang Chen, Yi Chang, Jun Wang 5/11/2026

CASCADE: Case-Based Continual Adaptation for Large Language Models During Deployment

Framework for enabling LLMs to continually adapt and learn during deployment through case-based examples and retrieval.

Ax Long Zhang, Wei-neng Chen, Feng-feng Wei, Zi-bo Qin 5/11/2026

When Does a Language Model Commit? A Finite-Answer Theory of Pre-Verbalization Commitment

Analysis of when language models stabilize answer preferences during reasoning generation using finite-answer projection.

Ax Xiaoyu Xu, Minxin Du, Qipeng Xie, Haobin Ke, Qingqing Ye, Haibo Hu 5/11/2026

When Routine Chats Turn Toxic: Unintended Long-Term State Poisoning in Personalized Agents

Security analysis of unintended long-term state poisoning in personalized LLM agents through routine interactions.

Ax Zhifeng Gu, Yuqi Wang, Bing Wang 5/11/2026

R$^3$L: Reasoning 3D Layouts from Relative Spatial Relations

Framework improving reliability of multimodal LLMs in inferring spatial relations for 3D layout generation.

Ax O\u{g}uzhan Fatih Kar, Roman Bachmann, Yuanzheng Gong, Anders Boesen Lindbo Larsen, Afshin Dehghan 5/11/2026

Weblica: Scalable and Reproducible Training Environments for Visual Web Agents

Framework for constructing scalable, reproducible web environments for training visual web agents with realistic diversity.

Ax Yuwei Yin, Chuyuan Li, Giuseppe Carenini 5/11/2026

IntentGrasp: A Comprehensive Benchmark for Intent Understanding

Comprehensive benchmark for evaluating LLM intent understanding capabilities across 12 domains with 49 corpora.

Ax Qinshi Zhang (University of California, San Diego), Weipeng Deng (University of Hong Kong), Zhihan Jiang (Columbia University), Jiaming Qu (Amazon), Qianren Li (City University of Hong Kong), Weitao Xu (City University of Hong Kong), Ray LC (City University of Hong Kong) 5/11/2026

AGWM: Affordance-Grounded World Models for Environments with Compositional Prerequisites

Model-based learning framework for agents in environments with compositional action prerequisites, using affordance-grounded world models.

Ax Keegan Harris, Siddharth Prasad, Asher Trockman 5/11/2026

In-Context Credit Assignment via the Core

Cooperative game theory approach for assigning credit to creators whose IP appears in LLM context windows, using least core solution concept.

Ax Joseph Quaratiello, Anthony Rizzo 5/11/2026

Physics-Based Flow Matching for Full-Field Prediction of Silicon Photonic Devices

PIC-Flow generative neural surrogate combining physics-based learning to predict electromagnetic fields for photonic devices.

Ax Syed Ihtesham Hussain Shah, Floris den Hengst, Aneta Lisowska, Annette ten Teije 5/11/2026

Multi-Objective Constraint Inference using Inverse reinforcement learning

Inverse reinforcement learning approach for multi-objective constraint inference from heterogeneous expert demonstrations.

Ax Haitham Kanj, Kiryung Lee 5/11/2026

Locally Near Optimal Piecewise Linear Regression in High Dimensions via Difference of Max-Affine Functions

Piecewise linear regression via Adaptive Block Gradient Descent using difference of max-affine functions parametrization.

Ax Dongqing Li, Geoff K. Nicholls, Shiyi Sun, You Luo 5/11/2026

A Differentiable Bayesian Relaxation for Latent Partial-Order Inference

Differentiable Bayesian method for inferring latent partial orders from linearized agent and workflow traces.

Ax Adam R. Klivans, Shyamal Patel, Konstantinos Stavropoulos, Arsen Vasilyan 5/11/2026

Equivalence of Coarse and Fine-Grained Models for Learning with Distribution Shift

Theoretical equivalence between PQ and TDS learning models for algorithms handling distribution shift in training and test data.