Ax Jose Marie Antonio Mi\~noza 2/13/2026

UltraLIF: Fully Differentiable Spiking Neural Networks via Ultradiscretization and Max-Plus Algebra

UltraLIF: fully differentiable spiking neural networks using ultradiscretization and tropical geometry instead of surrogate gradients.

Ax Danlong Yuan, Wei Wu, Zhengren Wang, Xueliang Zhao, Huishuai Zhang, Dongyan Zhao 2/13/2026

SWE-MiniSandbox: Container-Free Reinforcement Learning for Building Software Engineering Agents

SWE-MiniSandbox: lightweight container-free environment for scalable reinforcement learning training of software engineering agents.

Ax Tanmoy Mukherjee, Marius Kloft, Pierre Marquis, Zied Bouraoui 2/13/2026

Credal Concept Bottleneck Models: Structural Separation of Epistemic and Aleatoric Uncertainty

Credal Concept Bottleneck Models: framework separating epistemic and aleatoric uncertainty in predictive models for reliable decisions.

Ax Nikhil Garg, Jon Kleinberg, Kenny Peng 2/13/2026

How Many Features Can a Language Model Store Under the Linear Representation Hypothesis?

Mathematical framework for linear representation hypothesis: analysis of how many neurons encode features in language model layers.

Ax Yuanyong Luo, Jing Huang, Yu Cheng, Ziwei Yu, Kaihua Zhang, Kehong Hong, Xinda Ma, Xin Wang, Anping Tong, Guipeng Hu, Yun Xu, Mehran Taghian, Peng Wu, Guanglin Li, Yunke Peng, Tianchi Hu, Minqi Chen, Michael Bi Mi, Hu Liu, Xiping Zhou, Junsong Wang, Qiang Lin, Heng Liao 2/13/2026

HiFloat4 Format for Language Model Inference

HiFloat4: block floating-point data format for efficient language model inference with 4.5 bits per value on average.

Ax Anushri Eswaran, Oleg Golev, Darshan Tank, Sidhant Rahi, Himanshu Tyagi 2/13/2026

CryptoAnalystBench: Failures in Multi-Tool Long-Form LLM Analysis

CryptoAnalystBench: benchmark for LLM agent failures when reasoning over complex multi-tool outputs and large document sets.

Ax Jason Dury 2/13/2026

Predictive Associative Memory: Retrieval Beyond Similarity Through Temporal Co-occurrence

Predictive Associative Memory: neural architecture using temporal co-occurrence for memory retrieval instead of similarity-based approaches.

Ax Zeynab Anbiaee, Mahdi Rabbani, Mansur Mirani, Gunjan Piya, Igor Opushnyev, Ali Ghorbani, Sajjad Dadkhah 2/13/2026

Security Threat Modeling for Emerging AI-Agent Protocols: A Comparative Analysis of MCP, A2A, Agora, and ANP

Security threat modeling analysis of AI agent communication protocols (MCP, A2A, Agora, ANP) identifying vulnerabilities in multi-agent interaction systems.

Ax Qiaosi Wang (Hyun Jin), Jini Kim (Hyun Jin), Avanita Sharma (Hyun Jin), Alicia (Hyun Jin), Lee, Jodi Forlizzi, Hong Shen 2/13/2026

Situated, Dynamic, and Subjective: Envisioning the Design of Theory-of-Mind-Enabled Everyday AI with Industry Practitioners

Co-design study exploring how theory-of-mind capabilities could be integrated into everyday user-facing AI products and services based on practitioner input.

Ax Esha Singh, Dongxia Wu, Chien-Yi Yang, Tajana Rosing, Rose Yu, Yi-An Ma 2/13/2026

Divide and Learn: Multi-Objective Combinatorial Optimization at Scale

Reformulates multi-objective combinatorial optimization as online learning over decomposed spaces using adaptive expert-guided sequential construction.

Ax Zachary Pedram Dadfar 2/13/2026

When Models Examine Themselves: Vocabulary-Activation Correspondence in Self-Referential Processing

Analyzes whether self-referential vocabulary in LLM introspection reflects internal computation or confabulation using activation analysis and pull methodology.

Ax Sara Matijevic, Christopher Yau 2/13/2026

Bootstrapping-based Regularisation for Reducing Individual Prediction Instability in Clinical Risk Prediction Models

Proposes bootstrapping-based regularization to reduce prediction instability in clinical deep learning models for patient care.

Ax Weili Shi, Dongliang Guo, Lehan Yang, Tianlong Wang, Hanzhang Yuan, Sheng Li 2/13/2026

Finding the Cracks: Improving LLMs Reasoning with Paraphrastic Probing and Consistency Verification

Improves LLM reasoning by identifying critical tokens and using paraphrastic probing with consistency verification to reduce hallucinations and error accumulation.

Ax Arpit Singh Gautam, Kailash Talreja, Saurabh Jha 2/13/2026

The Energy of Falsehood: Detecting Hallucinations via Diffusion Model Likelihoods

Proposes DiffuTruth framework using diffusion model likelihoods and non-equilibrium thermodynamics to detect LLM hallucinations without supervision.

Ax Aviv Bick, Eric P. Xing, Albert Gu 2/13/2026

Retrieval-Aware Distillation for Transformer-SSM Hybrids

Uses distillation to convert Transformers into SSM hybrids by preserving attention heads that perform in-context retrieval and distilling remaining layers.

Ax Qingsong Wang, Mikhail Belkin, Yusu Wang 2/13/2026

General and Efficient Steering of Unconditional Diffusion

Proposes efficient steering methods for unconditional diffusion models without gradient-based guidance during inference, enabling faster controllable generation.

Ax Chongyi Zheng, Royina Karegoudra Jayanth, Benjamin Eysenbach 2/13/2026

Can We Really Learn One Representation to Optimize All Rewards?

Investigates whether a single learned representation can optimize performance across multiple reward functions in reinforcement learning using large model priors.

Ax Hanjing Shi, Dominic DiFranzo 2/13/2026

When Visibility Outpaces Verification: Delayed Verification and Narrative Lock-in in Agentic AI Discourse

Examines how social media engagement signals create credibility proxies that may delay verification and lock in narratives about agentic AI systems before direct evaluation.

Ax Zecheng Wang, Deyuan Liu, Chunshan Li, Yupeng Zhang, Zhengyun Zhao, Dianhui Chu, Bingning Wang, Dianbo Sui 2/13/2026

Gradients Must Earn Their Influence: Unifying SFT with Generalized Entropic Objectives

Training method unifying supervised fine-tuning with entropic objectives using gradient-based token weighting to resolve plasticity-stability tradeoffs in SFT.

Ax Muskaan Chopra, Lorenz Sparrenberg, Rafet Sifa 2/13/2026

Towards Reliable Machine Translation: Scaling LLMs for Critical Error Detection and Safety

LLM-based approach for detecting critical errors in machine translation including factual distortions and intent reversals for safety and reliability.

Ax Sajad Ebrahimi, Bhaskar Mitra, Negar Arabzadeh, Ye Yuan, Haolun Wu, Fattane Zarrinkalam, Ebrahim Bagheri 2/13/2026

From Noise to Order: Learning to Rank via Denoising Diffusion

Denoising diffusion-based generative approach to learning-to-rank that models joint distribution over feature vectors for information retrieval.

Ax Minda Li, Bhaskar Krishnamachari 2/13/2026

Compiler-Guided Inference-Time Adaptation: Improving GPT-5 Programming Performance in Idris

Compiler-guided inference-time adaptation enabling GPT-5 to improve programming performance in low-resource languages like Idris through iterative feedback.

Ax Stephan Vonschallen, Friederike Eyssel, Theresa Schmiedel 2/13/2026

Understanding Persuasive Interactions between Generative Social Agents and Humans: The Knowledge-based Persuasion Model (KPM)

Theoretical framework (KPM) for studying persuasive interactions between generative social agents and humans based on knowledge and adaptive communication.

Ax Zhen Bi, Xueshu Chen, Luoyang Sun, Yuhang Yao, Qing Shen, Jungang Lou, Cheng Deng 2/13/2026

RooflineBench: A Benchmarking Framework for On-Device LLMs via Roofline Analysis

Roofline-based benchmarking framework for characterizing small language model performance on resource-constrained edge hardware and heterogeneous platforms.

Ax David Wan, Han Wang, Ziyang Wang, Elias Stengel-Eskin, Hyunji Lee, Mohit Bansal 2/13/2026

Multimodal Fact-Level Attribution for Verifiable Reasoning

Framework for fact-level attribution in multimodal LLMs enabling verification of multi-step reasoning and grounding outputs across heterogeneous input sources.

Ax Yujie Gu, Richeng Jin, Xiaoyu Ji, Yier Jin, Wenyuan Xu 2/13/2026

Differentially Private and Communication Efficient Large Language Model Split Inference via Stochastic Quantization and Soft Prompt

Differential privacy and communication efficiency approach for LLM split inference using stochastic quantization and soft prompts on resource-constrained devices.

Ax Sidong Feng, Chunyang Chen 2/13/2026

How Smart Is Your GUI Agent? A Framework for the Future of Software Interaction

Framework proposing six autonomy levels for GUI agents to standardize capability assessment and benchmark progress toward trustworthy software interaction.

Ax Congmin Zheng, Xiaoyun Mo, Xinbei Ma, Qiqiang Lin, Yin Zhao, Jiachen Zhu, Xingyu Lou, Jun Wang, Zhaoxiang Wang, Weiwen Liu, Zhuosheng Zhang, Yong Yu, Weinan Zhang 2/13/2026

Adaptive Milestone Reward for GUI Agents

Adaptive milestone reward mechanism for reinforcement learning-based GUI agents that resolves temporal credit assignment in long-horizon tasks.

Ax Dong Yan, Jian Liang, Ran He, Tieniu Tan 2/13/2026

Stop Tracking Me! Proactive Defense Against Attribute Inference Attack in LLMs

Proactive defense method against attribute inference attacks in LLMs using word-level precision anonymization to prevent privacy breaches.

Ax Jingkun Liu, Yisong Yue, Max Welling, Yue Song 2/13/2026

Krause Synchronization Transformers

Krause Attention mechanism for transformers that prevents representation collapse and attention sink phenomena through bounded-confidence dynamics.

Ax Yuanfu Wang, Zhixuan Liu, Xiangtian Li, Chaochao Lu, Chao Yang 2/13/2026

Native Reasoning Models: Training Language Models to Reason on Unverifiable Data

Training approach for reasoning models that learn from unverifiable data without relying on external verifiers or human-annotated reasoning examples.

Ax Yiteng Tu, Shuo Miao, Weihang Su, Yiqun Liu, Qingyao Ai 2/13/2026

Analytical Search

Analytical Search framework for trend analysis and causal assessment using LLMs at corpus scale beyond standard RAG.

Ax Yujie Gu, Richeng Jin, Zhaoyang Zhang, Huaiyu Dai 2/13/2026

Gradient Compression May Hurt Generalization: A Remedy by Synthetic Data Guided Sharpness Aware Minimization

Analysis showing gradient compression hurts generalization in federated learning; proposes sharpness-aware minimization remedy.

Ax Zedong Chu, Shichao Xie, Xiaolong Wu, Yanfen Shen, Minghua Luo, Zhengbo Wang, Fei Liu, Xiaoxu Leng, Junjun Hu, Mingyang Yin, Jia Lu, Yingnan Guo, Kai Yang, Jiawei Han, Xu Chen, Yanqing Zhu, Yuxiang Zhao, Xin Liu, Yirong Yang, Ye He, Jiahang Wang, Yang Cai, Tianlin Zhang, Li Gao, Liu Liu, Mingchao Sun, Fan Jiang, Chiyu Wang, Zhicheng Liu, Hongyu Pan, Honglin Han, Zhining Gu, Kuan Yang, Jianfang Zhang, Di Jing, Zihao Guan, Wei Guo, Guoqing Liu, Di Yang, Xiangpo Yang, Menglin Yang, Hongguang Xing, Weiguo Li, Mu Xu 2/13/2026

ABot-N0: Technical Report on the VLA Foundation Model for Versatile Embodied Navigation

ABot-N0 Vision-Language-Action foundation model unifying embodied navigation tasks via LLM reasoning and flow matching.

Ax Wenqian Chen, Yucheng Fu, Michael Penwarden, Pratanu Roy, Panos Stinis 2/13/2026

ArGEnT: Arbitrary Geometry-encoded Transformer for Operator Learning

ArGEnT transformer architecture for learning solution operators across varying geometries in scientific machine learning.

Ax Changti Wu, Jiahuai Mao, Yuzhuo Miao, Shijie Lian, Bin Yu, Xiaopeng Lin, Cong Huang, Lei Zhang, Kai Chen 2/13/2026

ScalSelect: Scalable Training-Free Multimodal Data Selection for Efficient Visual Instruction Tuning

ScalSelect training-free data selection method for efficient visual instruction tuning of vision-language models.

Ax Enhao Huang, Frank Li, Tony Lin, Lowes Yang 2/13/2026

DMind-3: A Sovereign Edge--Local--Cloud AI System with Controlled Deliberation and Correction-Based Tuning for Safe, Low-Latency Transaction Execution

DMind-3 system for secure financial transaction execution in Web3 using edge-local-cloud AI stack with latency constraints.

Ax Christian Rondanini, Barbara Carminati, Elena Ferrari, Niccol\`o Lardo, Ashish Kundu 2/13/2026

LoRA-based Parameter-Efficient LLMs for Continuous Learning in Edge-based Malware Detection

LoRA-based parameter-efficient tuning for deploying LLMs on edge devices for malware detection with memory constraints.

Ax Seo Hyun Kim, Jin Bok Park, Do Yeon Koo, Ho Gun Park, Il Yong Chun 2/13/2026

SToRM: Supervised Token Reduction for Multi-modal LLMs toward efficient end-to-end autonomous driving

SToRM: supervised token reduction method for multimodal LLMs in autonomous driving with human-vehicle interaction.

Ax Kyungbok Lee, Angelica Cristello Sarteau, Michael R. Kosorok 2/13/2026

Provable Offline Reinforcement Learning for Structured Cyclic MDPs

Theoretical framework for offline reinforcement learning in cyclic MDPs with heterogeneous stage-specific dynamics.

Ax Sahand Sabour, TszYam NG, Minlie Huang 2/13/2026

PatientHub: A Unified Framework for Patient Simulation

PatientHub: unified modular framework for simulating patients with LLMs for training counselors and therapeutic assessment.

Ax Joey Zhong, Hao Zhang, Clare Southern, Jeremy Yang, Thomas Wang, Kate Jung, Shu Zhang, Denis Yarats, Johnny Ho, Jerry Ma 2/13/2026

DRACO: a Cross-Domain Benchmark for Deep Research Accuracy, Completeness, and Objectivity

DRACO: benchmark of 10 domain deep research tasks from real-world usage patterns for evaluating complex research capabilities.

Ax Oliver Zahn, Matt Beton, Simran Chana 2/13/2026

ANML: Attribution-Native Machine Learning with Guaranteed Robustness

ANML: machine learning framework that weights training samples by quality factors for improved robustness with specialized expert data.

Ax Yongyao Wang, Ziqi Miao, Lu Yang, Haonan Jia, Wenting Yan, Chen Qian, Lijun Li 2/13/2026

TabSieve: Explicit In-Table Evidence Selection for Tabular Prediction

TabSieve: select-then-predict framework that makes in-table evidence selection explicit for tabular prediction tasks.

Ax Arafa Yoncalik, Wouter Jansen, Nico Huebel, Mohammad Hasan Rahmani, Jan Steckel 2/13/2026

LLM-Driven 3D Scene Generation of Agricultural Simulation Environments

LLM-driven 3D scene generation for agricultural simulation environments with domain-specific reasoning and verification mechanisms.

Ax Matteo Nulli, Vladimir Orshulevich, Tala Bazazo, Christian Herold, Michael Kozielski, Marcin Mazur, Szymon Tuzel, Cees G. M. Snoek, Seyyed Hadi Hashemi, Omar Javed, Yannick Versley, Shahram Khadivi 2/13/2026

Adapting Vision-Language Models for E-commerce Understanding at Scale

Strategy for adapting vision-language models to e-commerce data with multiple images, structured attributes, and noisy information.

Ax Jiazheng Sun, Mingxuan Li, Yingying Zhang, Jiayang Niu, Yachen Wu, Ruihan Jin, Shuyu Lei, Pengrongrui Tan, Zongyu Zhang, Ruoyi Wang, Jiachen Yang, Boyu Yang, Jiacheng Liu, Xin Peng 2/13/2026

AmbiBench: Benchmarking Mobile GUI Agents Beyond One-Shot Instructions in the Wild

AmbiBench: benchmark for mobile GUI agents that require clarification and interaction to understand ambiguous user instructions.

Ax Keita Nishimoto, Kimitaka Asatani, Ichiro Sakata 2/13/2026

Cooperation Breakdown in LLM Agents Under Communication Delays

FLCOA framework analyzing cooperation breakdown in LLM-based multi-agent systems under communication delays and computational constraints.

Ax MiniCPM Team, Wenhao An, Yingfa Chen, Yewei Fang, Jiayi Li, Xin Li, Yaohui Li, Yishan Li, Yuxuan Li, Biyuan Lin, Chuan Liu, Hezi Liu, Siyuan Liu, Hongya Lyu, Yinxu Pan, Shixin Ren, Xingyu Shen, Zhou Su, Haojun Sun, Yangang Sun, Zhen Leng Thai, Xin Tian, Rui Wang, Xiaorong Wang, Yudong Wang, Bo Wu, Xiaoyue Xu, Dong Xu, Shuaikang Xue, Jiawei Yang, Bowen Zhang, Jinqian Zhang, Letian Zhang, Shengnan Zhang, Xinyu Zhang, Xinyuan Zhang, Zhu Zhang, Hengyu Zhao, Jiacheng Zhao, Jie Zhou, Zihan Zhou, Shuo Wang, Chaojun Xiao, Xu Han, Zhiyuan Liu, Maosong Sun 2/13/2026

MiniCPM-SALA: Hybridizing Sparse and Linear Attention for Efficient Long-Context Modeling

MiniCPM-SALA: 9B hybrid LLM architecture combining sparse and linear attention for efficient long-context modeling.

Ax Keita Broadwater 2/13/2026

Evaluating LLM Safety Under Repeated Inference via Accelerated Prompt Stress Testing

Accelerated Prompt Stress Testing framework for evaluating LLM safety under repeated inference with identical or similar prompts.