Isolater - Feed

Ax Disheng Liu, Tuo Liang, Chaoda Song, Yu Yin 7/7/2026

Post-Generation Curation of Synthetic Images via Homogeneous-Heterogeneous Splitting

Post-hoc curation method for synthetic images via homogeneous-heterogeneous splitting to improve training data quality without retraining generators.

Ax Kristina Zhang, Junior Chinomso Okoroafor, Benjamin Maltbie, Andrew Lin, Abhitej Bokka, Elliott Thornley 7/7/2026

Out-of-Distribution Generalization of Risk Aversion in Language Models

Research on training risk-averse language models to generalize risk aversion from low-stakes to high-stakes scenarios as AI safety approach.

Ax Yaswanth Chittepu, Ativ Joshi, Sohini Chintala, Scott Niekum 7/7/2026

Safe Inference-Time Alignment via Lagrangian Reward Augmentation

Lagrangian Reward Augmentation (LARA) framework for inference-time alignment of frozen LLMs using auxiliary reward signals with explicit safety constraints.

Ax Francesco D'Angelo, Oguz Kaan Yuksel, Swathi Shree Narashiman, Nicolas Flammarion 7/7/2026

Induction Heads Interpolate N-Grams

Study of induction heads in transformers, identifying attention circuits that implement soft context-matching estimators for in-context learning via smoothing mechanisms.

Ax Pranshu Chaturvedi, Parth Shroff, Tarun Suresh, Hangoo Kang, Kaiyue Wen 7/7/2026

Training Hybrid Block Diffusion Language Models with Partial Bidirectionality

Hybrid block diffusion language model training using partial bidirectionality to improve long-context generation efficiency and memory bandwidth.

Ax Trevor Chen, Ariel Dai, Jason Yang, Riccardo De Santi, Daniel Khalil, Wenda Chu, Nate Gruver, Pranav Murugan, Alexander F. G. Goldberg, Maruan Al-Shedivat, Yisong Yue 7/7/2026

On the Design Space of Discrete Diffusion Online Adaptation for Molecular Optimization

Study of discrete diffusion model adaptation for molecular optimization with limited oracle budget and test-time task-specific optimization.

Ax Lei Sun, Yusuke Tanaka, Tomoharu Iwata 7/7/2026

Labeled-Data-Free Meta-Learning: Efficient Task Generation Using Pre-trained Models and Unlabeled Data

Data-free meta-learning method using pre-trained models to generate training tasks without access to labeled data.

Ax Lin Li, Li Rong Wang, Hsuan Fu, Xiuyi Fan 7/7/2026

Trading Confidence: Comprehensive Uncertainty Estimation in Algorithmic Trading

Uncertainty estimation approach for RL-based algorithmic trading agents to handle market volatility and regime shifts.

Ax Anagha Radhakrishna Palandye, Rebecca Glick, Osheen Kaul 7/7/2026

Reward Granularity in RLVR: Comparing Process and Outcome Reward Structures for Mathematical Reasoning in Small Language Models

Study comparing process-level vs outcome-only reward structures in reinforcement learning for mathematical reasoning in small language models.

Ax Nan Fang, Yijun Wang, Hao Liao, Sikun Yang 7/7/2026

Poisson-Gamma Modeling of Inter-Relational Dependencies in Dynamic Knowledge Graphs

PGRE probabilistic model for temporal and relational dependencies in dynamic knowledge graphs.

Ax Zihao Hu, Yuan Yao, Jiheng Zhang, Zhengyuan Zhou 7/7/2026

Dynamic Regret for Non-Stationary Linear Bandits via Misspecification Reductions

Dynamic regret analysis for non-stationary linear bandits with time-varying action sets and drifting reward models.

Ax Hamish Ogilvy 7/7/2026

Variable Bit-width Quantization: Learning Per-Group Precision for "Bigger-but-Smaller" Language Models

Variable Bit-width Quantization technique where weight groups learn per-group precision for efficient language model compression.

Ax Binglin Ji, Anindya Sarkar, Hengchang Lu, Jens Sj\"olund, Yevgeniy Vorobeychik 7/7/2026

Bootstrap Flow-Map Tree Sampling Enables Online Feedback Driven Search

Bootstrap sampling method for observation-guided exploration in scientific discovery with limited sampling budget and sequential feedback.

Ax Di Wu, Hongyi Sun, Haichao Xu, Jia Chen, Zhong Chen, Jie Yang 7/7/2026

CoFEND: A Cross-Modal Fusion End-to-End Network for Cold-Start Drug-Drug Interaction Prediction

CoFEND network for cold-start drug-drug interaction prediction using cross-modal fusion of biomedical entity relationships.

Ax Amirpasha Hedayat, Laura Balzano, Karthik Duraisamy 7/7/2026

In-span learning: adapting reduced-order models using their own predictions

In-span learning method for adapting reduced-order models using their own predictions without external data.

Ax Soyeon Park, Charmgil Hong 7/7/2026

Missingness as Signal: Channel-Independent Spectrogram Learning for Clinical Time Series Prediction

CISM framework for clinical time series prediction that models missingness as informative signal in ICU data.

Ax Yujin Kim, Charmgil Hong 7/7/2026

A Precedent-Guided Co-Scientist for Side-Effect-Aware Drug Redesign

PRECEDE system uses LLM orchestration with knowledge graphs for side-effect-aware drug redesign, framing drug development as evidence-grounded reasoning.

Ax Joy Bose 7/7/2026

Rank-Order N-of-M Codes for Sparse Distributed Memory: Disentangling Representation and Learning Effects in Noise Robustness Against Contemporary Neuromorphic Architectures

Evaluation of rank-order N-of-M encoding for sparse distributed memory as alternative to threshold-binary encoding for continual learning systems.

Ax Yaniv Shulman, Shaghayegh Akbarpour, Jack B. Muir 7/7/2026

MABLE: Masked Autoencoding with Bi-Lipschitz Decoding for Embeddings and Graph Metric Learning

MABLE self-supervised framework combining masked reconstruction with bi-Lipschitz decoding for learning node and graph embeddings from heterogeneous graphs.

Ax Joonho Kim, Seyoung Park 7/7/2026

Transfer Learning in High-dimensional Ising Models

Trans-Ising transfer learning method for high-dimensional Ising model estimation using source screening and two-stage estimation with auxiliary data.

Ax Wenda Wang, Jinjia Feng, Zhewei Wei 7/7/2026

Back to Basics: Improving Molecular Understanding in LLMs via SMILES-Graph Translation

Molecular LLM improvement via SMILES-graph translation for better structural grounding in chemistry understanding, aligning with structure-determines-function principle.

Ax Beatrice Zanchi, Giuliana Monachino, Alvise Dei Rossi, Luigi Fiorillo, Georgia Sarquella-Brugada, Giulio Conte, Francesca Dalia Faraci 7/7/2026

Do ECG Foundation Models Transfer to Rare Cardiac Diseases? Evidence from Brugada Syndrome Detection

Evaluation of ECG foundation models on rare cardiac disease detection (Brugada syndrome), assessing transferability to clinically rare phenotypes.

Ax Stefan Horoi, Benjamin Th\'erien, Guy Wolf, Eugene Belilovsky 7/7/2026

Can Model Merging Improve Aggregation in DiLoCo?

Investigation of model merging techniques for improving aggregation in DiLoCo distributed learning, combining independent finetuned models.

Ax Yuan-Bin Zhu, Shuang Qiao, Shi-Ju Ran 7/7/2026

Out-of-distribution Neural Inference in Dynamical Ising Models

Study of neural networks for inferring interaction graphs in Ising models, evaluating out-of-distribution generalization across CNN, GNN, and Transformer architectures.

Ax Shijie Cao, Qingyu Zhang, Boxi Yu, Yuzhong Zhang, Boxi Cao, Yaojie Lu, Hongyu Lin, Xianpei Han, Le Sun 7/7/2026

OmniFocus: Query-Guided Modality-Balanced Token Compression for Omni-Modal Large Language Models

OmniFocus token compression for multimodal LLMs processing audio-video inputs using query-guided modality-balanced compression to reduce inference cost.

Ax Tomoya Mizuguchi, Bum Jun Kim 7/7/2026

SHiPPO: Recurrent Memory with Transported Polynomial Projections

SHiPPO extends HiPPO with transported polynomial projections for selective SSMs, enabling token-dependent control and channel interaction in recurrent memory.

Ax Zhuowen Liu, Longkun Hao, Shiyu Feng, Xiaowen Chang, Ruiqun Li, Changqun Li 7/7/2026

LACE-SVD: Loss-Aware SVD with Cumulative Error Correction for LLM Compression

LACE-SVD compression method for LLMs using loss-aware SVD with cumulative error correction for efficient low-rank compression.

Ax Zhilong Zhang, Hongli Yu, Huan-ang Gao, Hanlin Wu, Yuxuan Song, Wei-Ying Ma, Ya-Qin Zhang, Hao Zhou 7/7/2026

Spectral Rewiring for Exploration, Purification, and Model Merging

Spectral rewiring technique for RL post-training of LLMs, addressing reasoning saturation and model merging interference through targeted parameter updates.

Ax Nicolas Sournac, Ahmed Baha Ben Jmaa, Bertrand Braeckeveldt 7/7/2026

Robustness Meets Uncertainty: Evidential Adversarial Training for Robust Selective Classification

Framework combining adversarial training with evidential uncertainty for robust selective classification in safety-critical applications.

Ax Nirhoshan Sivaroopan, Albert Zomaya, Kanchana Thilakarathna 7/7/2026

STELLA: Efficient Sensor-to-LLM Translation for On-Device Human Activity Recognition

STELLA framework for on-device human activity recognition using efficient sensor-to-LLM translation with lightweight tokenization for edge deployment.

Ax Aymen Sarhane, Fouad Lbakali, Mouad Souissi, Jonathan Lys, Giulia Lioi 7/7/2026

Stacked LoRA for Subject-Adaptive EEG Foundation Models in Motor Imagery Decoding

Stacked LoRA adapters for subject-adaptive EEG foundation models in motor imagery decoding, addressing cross-subject generalization challenges.

Ax Naoya Chiba, Satoshi Sugiyama, Yuki Uranishi 7/7/2026

Observable- and Positional-Encoding-Dependent Symmetry Readout from Neural Network Weights

Analysis of symmetry structures recovered from neural network weights with positional encodings and observability hierarchies.

Ax Georg Sch\"afer, Jakob Rehrl, Stefan Huber 7/7/2026

Integrating Physics-Informed Neural Networks for Safe Reinforcement Learning in a 1-DoF Helicopter System

Physics-informed neural networks integrated into PPO actor loss for safe DRL in cyber-physical systems with hardware constraints.

Ax Zijun Xie, Yuyang You, Yongzhi Li, Enlei Gong, Zeyu Chen, Quan Chen, Yanhua Cheng, Peng Jiang, Yadong Mu 7/7/2026

ACPO: Adaptive Credit Policy Optimization via Fine-Grained Surrogate Entropy

ACPO method for token-level credit assignment in RL-finetuned LLMs, using fine-grained surrogate entropy for improved reasoning ability.

Ax Georg Sch\"afer, Jakob Rehrl, Stefan Huber, Simon Hirlaender 7/7/2026

Anticipatory Reinforcement Learning for Trajectory Tracking

Deep RL with predictive formulation for trajectory tracking using PPO, augmenting state space with target velocities to reduce lag and overshoot.

Ax Georg Sch\"afer, Jakob Rehrl, Stefan Huber, Simon Hirlaender 7/7/2026

Sample-Efficient Pareto Front Modeling for Energy-Aware Reinforcement Learning Using Bayesian Optimization

Multi-objective RL for industrial automation using Bayesian optimization to model Pareto fronts for energy-efficient control strategies.

Ax Alexandre L. M. Levada 7/7/2026

CuBAS: Information Geometric Curvature-Based Adaptive Sampling for Supervised Classification

CuBAS: information-geometric framework for adaptive data sampling in supervised classification using curvature-based selection.

Ax Muhammad Sabih, Frank Hannig, J\"urgen Teich 7/7/2026

Rethinking Neural Nonlinearity as Gating

Shows neural nonlinearity can be achieved through input-conditioned threshold gating as alternative to activation functions, unifying standard activations.

Ax Rajat Ghosh 7/7/2026

Reduced-Order Models: The Mother of World Models

Connects world models in modern deep learning to classical model-order-reduction and control theory, showing shared functional anatomy.

Ax Jialiang Wang, Xianming Liu, Xiong Zhou, Hui Liu, Haoliang Li 7/7/2026

Unbiased Alignment for Large Language Models with Noisy Preferences

Proposes Unbiased Reward Model loss and Unbiased DPO for robust LLM alignment from noisy preference datasets.

Ax Bing Cheng, Yi-Shuai Niu, Howell Tong, Shing-Tung Yau 7/7/2026

Statistically Meaningful Geometry (SMG) Beyond the Euclidean Paradigm, with Application to Generative AI

Proposes statistically meaningful geometry framework for analyzing generalization in over-parameterized models like transformers, addressing hallucination issues.

Ax Jakob Hartmann, James Harvey, Jhonathan Navott, Erik Y. Wang, Luckeciano C. Melo, Flaviu Cipcigan, Cheng Zhang, Alessandro Abate 7/7/2026

Amortising Bayesian Experimental Design for Sequential Information Gathering in LLMs

Proposes ASIG, a fine-tuning approach using Bayesian Experimental Design to improve information gathering in multi-turn LLM decision-making settings.

Ax Teng-Ruei Chen 7/7/2026

How Much of the Routing Gap Is Real? Decomposing the Router-to-Oracle Gap into Reproducible Specialist Advantage and Single-Draw Label Noise

Analyzes routing gaps between learned routers and oracles for LLM selection, decomposing the gap into reproducible specialist advantage and label noise components.

Ax Yanbo Wang, Jinhua Hao, Yuze Shi, Kun Yuan, Ming Sun 7/7/2026