2/19/2026
Transformers can do Bayesian Clustering
Transformer-based model for Bayesian clustering on datasets with missing values. Unsupervised learning extension of prior-data fitting.
2/19/2026
Synthesizing High-Quality Visual Question Answering from Medical Documents with Generator-Verifier LMMs
Framework for synthesizing medical VQA datasets from biomedical literature using generator-verifier multimodal models. LLM application.
2/19/2026
Shrinking the Variance: Shrinkage Baselines for Reinforcement Learning with Verifiable Rewards
Variance reduction techniques for RL with verifiable rewards in large reasoning models. Applies to LLM post-training with policy gradients.
2/19/2026
Q3R: Quadratic Reweighted Rank Regularizer for Effective Low-Rank Training
Q3R regularizer for low-rank training and fine-tuning of large deep learning models. Parameter-efficient training method.
2/19/2026
Watch Out for the Lifespan: Evaluating Backdoor Attacks Against Federated Model Adaptation
Security analysis of backdoor attacks in federated learning with LoRA fine-tuning. Covers distributed ML safety.
2/19/2026
High entropy leads to symmetry equivariant policies in Dec-POMDPs
Theoretical analysis of entropy regularization in Dec-POMDPs proving policy convergence to equivariant solutions. Machine learning theory.
2/19/2026
Adaptive Aggregation with Two Gains in QFL
Adaptive aggregation method for quantum federated learning handling client quality variation, teleportation fidelity, and device instability.
2/19/2026
Out-of-Distribution Detection in Molecular Complexes via Diffusion Models for Irregular Graphs
Out-of-distribution detection for 3D molecular graphs using diffusion models. Addresses OOD challenges for irregular graph structures in molecular complexes.
2/19/2026
Communication Compression for Distributed Learning with Aggregate and Server-Guided Feedback
Communication compression techniques for federated learning using biased compression with error feedback to handle asymmetric bandwidth constraints.
2/19/2026
Imitation Learning for Combinatorial Optimisation under Uncertainty
Imitation learning framework for combinatorial optimization under uncertainty. Studies how expert quality affects policy approximation in sequential decision problems.
2/19/2026
Mixture-of-Experts as Soft Clustering: A Dual Jacobian-PCA Spectral Geometry Perspective
Research analyzing Mixture-of-Experts architectures through geometric lens using Dual Jacobian-PCA spectral probe to understand routing and representation geometry.
2/19/2026
StableQAT: Stable Quantization-Aware Training at Ultra-Low Bitwidths
Quantization-aware training framework for ultra-low bitwidth deployment of large models, improving stability over straight-through estimators.
2/19/2026
Reinforcement Unlearning via Group Relative Policy Optimization
Method for removing memorized sensitive data from LLMs via group relative policy optimization, addressing GDPR/EU AI Act compliance without full retraining.
2/19/2026
GEPC: Group-Equivariant Posterior Consistency for Out-of-Distribution Detection in Diffusion Models
Out-of-distribution detection for diffusion models using group equivariance properties learned from in-distribution convolutional data.
2/19/2026
Cardinality-Preserving Attention Channels for Graph Transformers in Molecular Property Prediction
Graph transformer with cardinality-preserving attention channels for molecular property prediction with limited labeled data.
2/19/2026
Quant VideoGen: Auto-Regressive Long Video Generation via 2-Bit KV-Cache Quantization
System optimization for autoregressive video generation using 2-bit KV-cache quantization to reduce memory from 30GB and enable hardware deployment.
2/19/2026
Adaptive Exploration for Latent-State Bandits
Multi-armed bandit algorithms for sequential decision-making with hidden, time-varying states and unobserved confounders.
2/19/2026
Feature salience -- not task-informativeness -- drives machine learning model explanations
Study showing feature salience rather than task-informativeness drives XAI model explanations, questioning assumption about importance attribution.
2/19/2026
Features as Rewards: Scalable Supervision for Open-Ended Tasks via Interpretability
Using interpretable LLM features as supervision signal for RL-based open-ended tasks, demonstrated on hallucination reduction.
2/19/2026
Boundary Point Jailbreaking of Black-Box LLMs
Novel automated jailbreak attack method that evades classifier-based safeguards in frontier LLMs without requiring white/grey-box access.
2/19/2026
Closing the Distribution Gap in Adversarial Training for LLMs
Research on fundamental limitations of adversarial training for LLMs, showing models remain vulnerable to distribution-shifted attacks like prompt rewriting.
2/19/2026
Evaluating Language Model Agency through Negotiations
Framework evaluating language model agency through multi-turn negotiation games, testing six public LLMs against real-world interactive scenarios.
2/19/2026
Ctrl-GenAug: Controllable Generative Augmentation for Medical Sequence Classification
Diffusion-based data augmentation method for medical sequence classification with controllable semantic and sequential generation.
2/19/2026
Autoassociative Learning of Structural Representations for Modeling and Classification in Medical Imaging
Deep learning approach for medical imaging using autoassociative learning to capture discrete object categories instead of continuous features.
2/19/2026
LMSeg: Unleashing the Power of Large-Scale Models for Open-Vocabulary Semantic Segmentation
Open-vocabulary semantic segmentation using large-scale vision-language models like CLIP to recognize unseen objects in images.
2/19/2026
VerifiableFL: Verifiable Claims for Federated Learning using Exclaves
Federated learning system using secure enclaves to provide verifiable claims about model training data and algorithms, preventing malicious deviation.
2/19/2026
Adaptive Rank Allocation for Federated Parameter-Efficient Fine-Tuning of Language Models
Research on parameter-efficient fine-tuning of language models in federated settings, addressing privacy and efficiency for distributed training on resource-constrained devices.
2/19/2026
Large Language Models for Water Distribution Systems Modeling and Decision-Making
LLM agent framework enabling natural language interaction with water distribution system simulator EPANET.
2/19/2026
View Invariant Learning for Vision-Language Navigation in Continuous Environments
Method for view-invariant learning in vision-language navigation for embodied AI agents.
2/19/2026
PoeTone: A Framework for Constrained Generation of Structured Chinese Songci with LLMs
Framework for constrained generation of Chinese Songci poetry with LLMs using structural and tonal constraints.
2/19/2026
Language and Experience: A Computational Model of Social Learning in Complex Tasks
Computational model of social learning combining linguistic guidance with sensorimotor experience for AI systems.
2/19/2026
Statistical Inference Leveraging Synthetic Data with Distribution-Free Guarantees
Framework for combining synthetic and real data in statistical inference with distribution-free guarantees.
2/19/2026
Conditionally Whitened Generative Models for Probabilistic Time Series Forecasting
Generative framework for probabilistic multivariate time series forecasting using conditional whitening.
2/19/2026
Multilingual Routing in Mixture-of-Experts
Analysis of expert routing patterns in multilingual Mixture-of-Experts language models across languages.
2/19/2026
Lossless Vocabulary Reduction for Auto-Regressive Language Models
Method for reducing vocabulary size in auto-regressive language models while maintaining lossless compression.
2/19/2026
Precise Attribute Intensity Control in Large Language Models via Targeted Representation Editing
Technique for precise control over attribute intensity in LLM outputs via targeted representation editing.
2/19/2026
Randomized Masked Finetuning: An Efficient Way to Mitigate Memorization of PIIs in LLMs
Randomized Masked Fine-Tuning reduces PII memorization in LLMs during fine-tuning while maintaining performance. Privacy-preserving technique.
2/19/2026
Transformers for Tabular Data: A Training Perspective of Self-Attention via Optimal Transport
Self-attention training analysis via optimal transport theory for tabular classification. Theoretical perspective on transformers.
2/19/2026
KANEL\'E: Kolmogorov-Arnold Networks for Efficient LUT-based Evaluation
KANELÉ: Kolmogorov-Arnold Networks optimized for FPGA lookup table deployment. Efficient neural network inference framework.
2/19/2026
High-dimensional learning dynamics of multi-pass Stochastic Gradient Descent in multi-index models
Theoretical analysis of SGD learning dynamics in high-dimensional multi-index models. Fundamental ML research.
2/19/2026
SEISMO: Increasing Sample Efficiency in Molecular Optimization with a Trajectory-Aware LLM Agent
SEISMO: LLM agent for sample-efficient molecular optimization using trajectory awareness. Applies agents to chemistry.
2/19/2026
Logarithmic-time Schedules for Scaling Language Models with Momentum
Time-varying AdamW schedules (beta, weight-decay) for language model training exploiting power-law data structure. Improves LLM training efficiency.
2/19/2026
Protean Compiler: An Agile Framework to Drive Fine-grain Phase Ordering
Compiler optimization using machine learning for phase ordering decisions. Developer tools, not AI-focused.
2/19/2026
Vision and Language: Novel Representations and Artificial intelligence for Driving Scene Safety Assessment and Autonomous Vehicle Planning
Vision-language models for autonomous driving safety assessment and planning. Applies VLMs to scene understanding and decision-making.
2/19/2026
From Collapse to Improvement: Statistical Perspectives on the Evolutionary Dynamics of Iterative Training on Contaminated Sources
Statistical analysis of model collapse in iterative training with synthetic data. Shows conditions for improvement despite contamination.
2/19/2026
When Models Examine Themselves: Vocabulary-Activation Correspondence in Self-Referential Processing
Investigation of whether self-examination language in LLMs reflects computation or confabulation. Analyzes LLM interpretability via activation patterns.
2/19/2026
Learning Gradient Flow: Using Equation Discovery to Accelerate Engineering Optimization
Equation discovery to learn gradient descent dynamics and accelerate optimization without computing gradients. ML acceleration.
2/19/2026
BPP: Long-Context Robot Imitation Learning by Focusing on Key History Frames
Robot policy learning that handles long observation histories by selecting key frames. Addresses spurious correlations in imitation learning.
2/19/2026
Weight space Detection of Backdoors in LoRA Adapters
Detection method for backdoor attacks in LoRA adapters without running inference. Addresses security in open-source LLM fine-tuning.
2/19/2026
Show HN: I built a pre-push safety net for AI-generated code
Ship Safe: open-source developer tool that scans AI-generated code for security risks before pushing to repositories.