3/24/2026
Detection of adversarial intent in Human-AI teams using LLMs
Detection framework using LLMs to identify adversarial attacks against human-AI teams, covering data poisoning, prompt injection, and prompt engineering threats.
3/24/2026
From Causal Discovery to Dynamic Causal Inference in Neural Time Series
Method for discovering time-varying causal networks in neural time series without assuming known causal structure a priori.
3/24/2026
Joint Surrogate Learning of Objectives, Constraints, and Sensitivities for Efficient Multi-objective Optimization of Neural Dynamical Systems
Optimization framework using jointly learned surrogate models for multi-objective optimization of neural dynamical systems in biophysical simulations.
3/24/2026
Interpreting the Synchronization Gap: The Hidden Mechanism Inside Diffusion Transformers
Theoretical analysis of synchronization gaps in diffusion transformers using coupled Ornstein-Uhlenbeck systems to explain mode interaction hierarchies in the reverse process.
3/24/2026
Structural Sensitivity in Compressed Transformers: Error Propagation, Lyapunov Stability, and Formally Verified Bounds
Study of sensitivity in compressed transformers across architectures, identifying which components degrade catastrophically vs. compress well, with formal bounds on error propagation.
3/24/2026
Long-Term Outlier Prediction Through Outlier Score Modeling
Novel unsupervised method for long-term outlier prediction in time series data using outlier score modeling.
3/24/2026
When Does Content-Based Routing Work? Representation Requirements for Selective Attention in Hybrid Sequence Models
Analysis of content-based routing in hybrid recurrent-attention architectures through controlled experiments on multiple benchmarks.
3/24/2026
CLT-Forge: A Scalable Library for Cross-Layer Transcoders and Attribution Graphs
CLT-Forge library for mechanistic interpretability of LLMs using cross-layer transcoders and sparse feature attribution graphs.
3/24/2026
Deep Attention-based Sequential Ensemble Learning for BLE-Based Indoor Localization in Care Facilities
Deep attention-based sequential ensemble learning framework for BLE-based indoor localization in care facilities.
3/24/2026
Fuel Consumption Prediction: A Comparative Analysis of Machine Learning Paradigms
Comparative analysis of machine learning approaches for vehicle fuel consumption prediction using Motor Trend dataset.
3/24/2026
Benchmarking Scientific Machine Learning Models for Air Quality Data
Benchmarking physics-guided and deep learning models for air quality index forecasting on standardized datasets.
3/24/2026
Confidence Freeze: Early Success Induces a Metastable Decoupling of Metacognition and Behaviour
Cognitive science study on human decision-making persistence in bandit tasks using confidence-freeze theory.
3/24/2026
Harmful Visual Content Manipulation Matters in Misinformation Detection Under Multimedia Scenarios
Multimodal misinformation detection system addressing visual content manipulation in social media.
3/24/2026
Semi-Supervised Learning with Balanced Deep Representation Distributions
Semi-supervised text classification using self-training with pseudo-labels to improve deep classifier performance on unlabeled data.
3/24/2026
Mixture of Chapters: Scaling Learnt Memory in Transformers
Learnable sparse memory banks with chapter-based routing for scaling knowledge storage in Transformers without prohibitive attention costs.
3/24/2026
ResPrune: Text-Conditioned Subspace Reconstruction for Visual Token Pruning in Large Vision-Language Models
Training-free visual token pruning framework for efficient Vision-Language Model inference through text-conditioned subspace reconstruction.
3/24/2026
DMMRL: Disentangled Multi-Modal Representation Learning via Variational Autoencoders for Molecular Property Prediction
Disentangled multi-modal representation learning via VAEs for molecular property prediction in drug discovery and materials science.
3/24/2026
Learning from Label Proportions with Dual-proportion Constraints
Weakly supervised learning method for classification from bag-level label proportions using dual-proportion constraints.
3/24/2026
Model Evolution Under Zeroth-Order Optimization: A Neural Tangent Kernel Perspective
Theoretical analysis of zeroth-order optimization training dynamics through Neural Tangent Kernel perspective for gradient-free neural network training.
3/24/2026
Pruned Adaptation Modules: A Simple yet Strong Baseline for Continual Foundation Models
Lightweight continual learning method using pruned adaptation modules for foundation models compared against recent FM-based approaches.
3/24/2026
Rethinking Plasticity in Deep Reinforcement Learning
Investigation of plasticity loss in deep reinforcement learning with proposed Optimization-Centric Plasticity hypothesis explaining adaptation dynamics.
3/24/2026
Reward Sharpness-Aware Fine-Tuning for Diffusion Models
Sharpness-aware fine-tuning approach for diffusion models to reduce reward hacking in reinforcement learning from human feedback.
3/24/2026
Prompt replay: speeding up grpo with on-policy reuse of high-signal prompts
Online data selection method for GRPO reinforcement learning that reuses high-signal prompts to improve LLM reasoning training efficiency.
3/24/2026
ALMAB-DC: Active Learning, Multi-Armed Bandits, and Distributed Computing for Sequential Experimental Design and Black-Box Optimization
Bayesian sequential design framework combining active learning, multi-armed bandits, and distributed computing for black-box optimization.
3/24/2026
On the Role of Batch Size in Stochastic Conditional Gradient Methods
Theoretical analysis of batch size effects in stochastic conditional gradient optimization methods.
3/24/2026
Pretrained Video Models as Differentiable Physics Simulators for Urban Wind Flows
Pretrained video diffusion model repurposed as differentiable physics simulator for urban wind flow prediction.
3/24/2026
Does Mechanistic Interpretability Transfer Across Data Modalities? A Cross-Domain Causal Circuit Analysis of Variational Autoencoders
Analysis of mechanistic interpretability in VAEs across image and tabular data modalities using causal circuit analysis.
3/24/2026
Amortized Variational Inference for Logistic Regression with Missing Covariates
Amortized variational inference method for logistic regression with missing covariate data using VAE-based approach.
3/24/2026
Aggregation Alignment for Federated Learning with Mixture-of-Experts under Data Heterogeneity
Federated learning framework for fine-tuning Mixture-of-Experts LLMs on distributed data with privacy preservation.
3/24/2026
Fusing Memory and Attention: A study on LSTM, Transformer and Hybrid Architectures for Symbolic Music Generation
Comparative study of LSTM, Transformer, and hybrid architectures for symbolic music generation tasks.
3/24/2026
Sonny: Breaking the Compute Wall in Medium-Range Weather Forecasting
Data-driven weather forecasting using deep learning with reduced computational requirements compared to existing models.
3/24/2026
Direct Interval Propagation Methods using Neural-Network Surrogates for Uncertainty Quantification in Physical Systems Surrogate Model
Neural network surrogates for uncertainty quantification in physical systems through interval propagation methods.
3/24/2026
FluidWorld: Reaction-Diffusion Dynamics as a Predictive Substrate for World Models
Research on world models using reaction-diffusion dynamics as alternative to Transformers for predicting future environment states with better spatial inductive bias.
3/24/2026
Stream separation improves Bregman conditioning in transformers
Analysis of Bregman geometry in transformer representations, showing how stream separation improves steering methods.
3/24/2026
Active Inference Agency Formalization, Metrics, and Convergence Assessments
Formal framework for defining and analyzing agency in AI systems through continuous representation and mesa-optimization dynamics.
3/24/2026
AutoKernel: Autonomous GPU Kernel Optimization via Iterative Agent-Driven Search
AutoKernel: open-source autonomous agent framework for GPU kernel optimization using iterative search on PyTorch models.
3/24/2026
The Workload-Router-Pool Architecture for LLM Inference Optimization: A Vision Paper from the vLLM Semantic Router Project
vLLM Semantic Router architecture for LLM inference optimization covering routing, caching, safety, and adaptive mechanisms.
3/24/2026
TIDE: Token-Informed Depth Execution for Per-Token Early Exit in LLM Inference
TIDE: post-training system with learned routers for per-token early exit in LLM inference, no retraining required.
3/24/2026
PLR: Plackett-Luce for Reordering In-Context Learning Examples
PLR: method using Plackett-Luce ranking to efficiently reorder in-context learning examples without exhaustive search.
3/24/2026
Constrained Online Convex Optimization with Memory and Predictions
Algorithms for constrained online convex optimization with memory constraints and predictions.
3/24/2026
A Generalised Exponentiated Gradient Approach to Enhance Fairness in Binary and Multi-class Classification Tasks
Fairness improvement method using exponentiated gradient approach for multi-class classification tasks to mitigate bias.
3/24/2026
Mechanisms of Introspective Awareness
Study of introspective awareness mechanisms in LLMs, investigating whether steering detection reflects genuine circuitry or shallow heuristics.
3/24/2026
DSPA: Dynamic SAE Steering for Data-Efficient Preference Alignment
DSPA: inference-time method using sparse autoencoders for LLM preference alignment without weight updates, enabling mechanistic steering.
3/24/2026
Off-Policy Evaluation for Ranking Policies under Deterministic Logging Policies
Off-policy evaluation methods for ranking systems using offline logged data, addressing bias in inverse propensity score estimators.
3/24/2026
Learning Can Converge Stably to the Wrong Belief under Latent Reliability
Research on how learning systems can converge to incorrect solutions when feedback reliability is unobservable, addressing theoretical issues in optimization.
3/24/2026
Multinoulli Extension: A Lossless Continuous Relaxation for Partition-Constrained Subset Selection
Continuous relaxation method for partition-constrained subset selection with submodular objectives, improving query complexity over existing local-search approaches.
3/24/2026
Quotient Geometry, Effective Curvature, and Implicit Bias in Simple Shallow Neural Networks
Develops differential-geometric framework accounting for parameter redundancy in shallow neural networks via quotient geometry to measure intrinsic predictor properties.
3/24/2026
Optimizing Feature Extraction for On-device Model Inference with User Behavior Sequences
Addresses on-device ML inference bottleneck by optimizing feature extraction from user behavior sequences for low-latency mobile app execution.
3/24/2026
BOxCrete: A Bayesian Optimization Open-Source AI Model for Concrete Strength Forecasting and Mix Optimization
Open-source Bayesian optimization model for concrete strength prediction and mix design optimization, applying ML to materials science with public datasets.
3/24/2026
Sharper Generalization Bounds for Transformer
Derives sharper generalization error bounds for Transformer architectures using offset Rademacher complexity across single and multi-head, multi-layer variants.