2/19/2026
Amortized Predictability-aware Training Framework for Time Series Forecasting and Classification
Amortized training framework addressing low-predictability samples in time series forecasting and classification.
2/19/2026
Factored Latent Action World Models
Factored latent action world models learning from action-free video for scalable controllable video generation.
2/19/2026
Regret and Sample Complexity of Online Q-Learning via Concentration of Stochastic Approximation with Time-Inhomogeneous Markov Chains
High-probability regret bounds for online Q-learning in infinite-horizon MDPs without optimism or bonus terms.
2/19/2026
Fast KV Compaction via Attention Matching
KV cache compaction method via attention matching to reduce memory overhead for long-context LLM inference.
2/19/2026
A Graph Meta-Network for Learning on Kolmogorov-Arnold Networks
Graph meta-network architecture for weight-space models to predict neural network accuracy on Kolmogorov-Arnold networks.
2/19/2026
HAWX: A Hardware-Aware FrameWork for Fast and Scalable ApproXimation of DNNs
Hardware-aware framework for DNN approximation using multi-level sensitivity scoring and heterogeneous approximation blocks.
2/19/2026
The Implicit Bias of Adam and Muon on Smooth Homogeneous Neural Networks
Theoretical analysis of implicit bias in momentum-based optimizers (Adam, Muon, MomentumGD) on homogeneous neural networks.
2/19/2026
Improved Bounds for Reward-Agnostic and Reward-Free Exploration
Theoretical analysis of reward-free and reward-agnostic exploration in MDPs with improved regret bounds.
2/19/2026
Easy Data Unlearning Bench
Unified benchmarking suite for evaluating machine unlearning algorithms with KLoM metric and precomputed model ensembles.
2/19/2026
Intra-Fairness Dynamics: The Bias Spillover Effect in Targeted LLM Alignment
Studies bias spillover in LLM fairness alignment, showing how single-attribute debiasing can worsen disparities in other dimensions.
2/19/2026
Beyond SGD, Without SVD: Proximal Subspace Iteration LoRA with Diagonal Fractional K-FAC
LoRSum improves LoRA fine-tuning efficiency using proximal subspace iteration with diagonal K-FAC, reducing memory overhead.
2/19/2026
Synthesis and Verification of Transformer Programs
Develops verification techniques for C-RASP language capturing transformer-expressible concepts using model checkers and SMT-solvers.
2/19/2026
Interpretability-by-Design with Accurate Locally Additive Models and Conditional Feature Effects
Proposes conditionally additive local models balancing interpretability of GAMs with accuracy by adding conditional feature interactions.
2/19/2026
Capacity-constrained demand response in smart grids using deep reinforcement learning
Applies deep reinforcement learning to capacity-constrained demand response for smart grid management and congestion prevention.
2/19/2026
FEKAN: Feature-Enriched Kolmogorov-Arnold Networks
Introduces FEKAN, feature-enriched Kolmogorov-Arnold networks with improved scalability and convergence over existing KAN variants.
2/19/2026
Transfer Learning of Linear Regression with Multiple Pretrained Models: Benefiting from More Pretrained Models via Overparameterization Debiasing
Studies transfer learning for linear regression using multiple overparameterized pretrained models with debiasing.
2/19/2026
Vulnerability Analysis of Safe Reinforcement Learning via Inverse Constrained Reinforcement Learning
Analyzes vulnerabilities of safe reinforcement learning via inverse constrained RL without requiring policy gradient access.
2/19/2026
A Scalable Approach to Solving Simulation-Based Network Security Games
Introduces MetaDOAR meta-controller augmenting double oracle paradigm with filtering and caching for multi-agent RL on cyber networks.
2/19/2026
Steering diffusion models with quadratic rewards: a fine-grained analysis
Provides fine-grained analysis of steering diffusion models with quadratic rewards at inference time for downstream tasks.
2/19/2026
AIFL: A Global Daily Streamflow Forecasting Model Using Deterministic LSTM Pre-trained on ERA5-Land and Fine-tuned on IFS
Proposes LSTM model pretrained on ERA5-Land reanalysis and finetuned on IFS for global streamflow forecasting.
2/19/2026
Sequential Membership Inference Attacks
Develops sequential membership inference attacks exploiting model dynamics across multiple updates for privacy auditing.
2/19/2026
Almost Sure Convergence of Differential Temporal Difference Learning for Average Reward Markov Decision Processes
Proves almost sure convergence of differential temporal difference learning for average reward reinforcement learning.
2/19/2026
Optimizer choice matters for the emergence of Neural Collapse
Analyzes role of optimizer choice in emergence of neural collapse patterns during deep neural network training.
2/19/2026
Protecting the Undeleted in Machine Unlearning
Demonstrates privacy risks in machine unlearning: reconstruction attacks show remaining data can be exposed when perfect retraining is pursued.
2/19/2026
Causality is Key for Interpretability Claims to Generalise
Argues causal inference is necessary for valid interpretability claims in LLMs, critiquing non-generalizable findings and unsupported causal interpretations.
2/19/2026
Do Personality Traits Interfere? Geometric Limitations of Steering in Large Language Models
Studies geometric constraints on personality trait steering in LLMs (LLaMA, Mistral), examining whether Big Five traits can be independently controlled via steering vectors.
2/19/2026
Building Safe and Deployable Clinical Natural Language Processing under Temporal Leakage Constraints
Clinical NLP safety analysis addressing temporal leakage in discharge planning models and deployment risk mitigation.
2/19/2026
MARVL: Multi-Stage Guidance for Robotic Manipulation via Vision-Language Models
MARVL uses vision-language models for multi-stage reward design in robotic manipulation RL, addressing spatial grounding and task semantics.
2/19/2026
P-RAG: Prompt-Enhanced Parametric RAG with LoRA and Selective CoT for Biomedical and Multi-Hop QA
P-RAG combines parametric and retrieval-augmented generation with LoRA and selective chain-of-thought for biomedical QA.
2/19/2026
Learning to Drive in New Cities Without Human Demonstrations
Self-play multi-agent reinforcement learning for autonomous driving policy adaptation to new cities without human demonstrations.
2/19/2026
Quality-constrained Entropy Maximization Policy Optimization for LLM Diversity
Quality-constrained entropy maximization framework for LLM fine-tuning balancing output diversity and alignment quality.
2/19/2026
World Action Models are Zero-shot Policies
DreamZero world action model using video diffusion for zero-shot robotic control by learning physical dynamics and generalizing to unseen environments.
2/19/2026
Visual Memory Injection Attacks for Multi-Turn Conversations
Security analysis of vision-language models in multi-turn conversations, exploring injection attacks via manipulated images.
2/19/2026
Can Vision-Language Models See Squares? Text-Recognition Mediates Spatial Reasoning Across Three Model Families
Study examining spatial reasoning limitations in vision-language models when localizing cells in binary grids without textual cues.
2/19/2026
MadEvolve: Evolutionary Optimization of Cosmological Algorithms with Large Language Models
MadEvolve framework uses LLMs to discover and optimize scientific algorithms for cosmology problems via iterative code modification and parameter tuning.
2/19/2026
ReLoop: Structured Modeling and Behavioral Verification for Reliable LLM-Based Optimization
ReLoop addresses silent failures in LLM-based optimization code through structured generation and behavioral verification, closing feasibility-correctness gaps up to 90%.
2/19/2026
MAEB: Massive Audio Embedding Benchmark
Benchmark evaluating 50+ audio embedding models across 30 tasks in speech, music, and audio-text reasoning in 100+ languages.
2/19/2026
Towards Efficient Constraint Handling in Neural Solvers for Routing Problems
Construct-and-Refine method for efficient constraint handling in neural solvers for complex routing problems.
2/19/2026
Heuristic Search as Language-Guided Program Optimization
Language-guided program optimization using LLMs for automated heuristic design in combinatorial optimization problems.
2/19/2026
CLAA: Cross-Layer Attention Aggregation for Accelerating LLM Prefill
CLAA cross-layer attention aggregation technique to accelerate LLM prefill stage using token ranking.
2/19/2026
Harnessing Implicit Cooperation: A Multi-Agent Reinforcement Learning Approach Towards Decentralized Local Energy Markets
Implicit cooperation MARL approach enabling decentralized agent coordination in local energy markets without direct communication.
2/19/2026
MARLEM: A Multi-Agent Reinforcement Learning Simulation Framework for Implicit Cooperation in Decentralized Local Energy Markets
MARLEM open-source multi-agent RL simulation framework for studying implicit cooperation in decentralized energy markets.
2/19/2026
The Limits of Long-Context Reasoning in Automated Bug Fixing
Systematic evaluation of LLM long-context reasoning limits in automated bug fixing using SWE-bench Verified.
2/19/2026
LGQ: Learning Discretization Geometry for Scalable and Stable Image Tokenization
LGQ discretization geometry learning approach for scalable and stable image tokenization in visual generation.
2/19/2026
Evolutionary Context Search for Automated Skill Acquisition
Evolutionary Context Search enables LLMs to acquire new knowledge post-deployment via improved retrieval-augmented generation.
2/19/2026
Rethinking ANN-based Retrieval: Multifaceted Learnable Index for Large-scale Recommendation System
Multifaceted learnable index approach for ANN-based retrieval in large-scale recommendation systems.
2/19/2026
Empirical Cumulative Distribution Function Clustering for LLM-based Agent System Analysis
ECDF clustering method for analyzing quality and distributional characteristics of LLM-based agent system responses.
2/19/2026
CHAI: CacHe Attention Inference for text2video
CHAI uses cross-inference caching to accelerate text-to-video diffusion model inference while maintaining quality.
2/19/2026
Local adapt-then-combine algorithms for distributed nonsmooth optimization: Achieving provable communication acceleration
FlexATC distributed optimization framework for nonsmooth problems with communication efficiency over networks.
2/19/2026
Learning Personalized Agents from Human Feedback
PAHF framework for training AI agents to learn and adapt to individual user preferences through continuous human feedback.