2/19/2026
Generalized Leverage Score for Scalable Assessment of Privacy Vulnerability
Generalized leverage score method for assessing privacy vulnerability and membership inference attack risk without model retraining.
2/19/2026
From Tool Orchestration to Code Execution: A Study of MCP Design Choices
Study of Model Context Protocol design choices for tool orchestration and code execution in multi-server agent systems.
2/19/2026
DocSplit: A Comprehensive Benchmark Dataset and Evaluation Approach for Document Packet Recognition and Splitting
DocSplit introduces first benchmark dataset for document packet splitting task in heterogeneous multi-page document understanding.
2/19/2026
From Reflection to Repair: A Scoping Review of Dataset Documentation Tools
Scoping review of 59 publications examining motivations and adoption barriers for dataset documentation tools in AI development.
2/19/2026
B-DENSE: Branching For Dense Ensemble Network Learning
B-DENSE proposes branching method for dense ensemble learning in diffusion models to improve inference speed and preserve structure.
2/19/2026
ReLoop: Structured Modeling and Behavioral Verification for Reliable LLM-Based Optimization
ReLoop addresses silent failures in LLM-based optimization code generation using structured modeling and behavioral verification.
2/19/2026
Anatomy of Capability Emergence: Scale-Invariant Representation Collapse and Top-Down Reorganization in Neural Networks
Study tracks capability emergence mechanisms across neural network scales using five geometric measures on 120+ emergence events.
2/19/2026
MAEB: Massive Audio Embedding Benchmark
MAEB benchmark evaluates 50+ audio embedding models across 30 tasks in speech, music, and cross-modal audio-text reasoning.
2/19/2026
MedProbCLIP: Probabilistic Adaptation of Vision-Language Foundation Model for Reliable Radiograph-Report Retrieval
MedProbCLIP applies probabilistic vision-language models for chest X-ray and radiology report retrieval in biomedical applications.
2/19/2026
Transforming GenAI Policy to Prompting Instruction: An RCT of Scalable Prompting Interventions in a CS1 Course
RCT study (N=979) evaluates prompting instruction interventions teaching students to use GenAI as tutor for better learning outcomes.
2/19/2026
AI-CARE: Carbon-Aware Reporting Evaluation Metric for AI Models
ML-CARE proposes a carbon-aware evaluation metric for AI models to measure environmental cost alongside standard performance benchmarks.
2/19/2026
Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training
Theoretical analysis of how generative AI systems perform when trained on contaminated data containing AI-generated content mixed with human-generated content.
2/19/2026
Language Statistics and False Belief Reasoning: Evidence from 41 Open-Weight LMs
Study of theory-of-mind reasoning in 41 open-weight language models, testing false belief tasks to understand mental state reasoning capabilities across diverse models.
2/19/2026
Updating Parametric Knowledge with Context Distillation Retains Post-Training Capabilities
Method for updating LLM knowledge from new documents while retaining post-training capabilities like instruction-following and reasoning through context distillation.
2/19/2026
Federated Graph AGI for Cross-Border Insider Threat Intelligence in Government Financial Schemes
Federated learning approach for detecting cross-border insider threats in government financial systems using graph-based reasoning on distributed, privacy-sensitive data.
2/19/2026
Surrogate-Based Prevalence Measurement for Large-Scale A/B Testing
Surrogate-based measurement technique using LLM labeling for cost-effective prevalence estimation in large-scale A/B testing.
2/19/2026
Rethinking ANN-based Retrieval: Multifaceted Learnable Index for Large-scale Recommendation System
Learnable index approach for approximate nearest neighbor search in large-scale recommendation systems with joint embedding and indexing.
2/19/2026
Retrieval Collapses When AI Pollutes the Web
Analysis of ecosystem-level failure in retrieval systems when AI-generated content pollutes search results and RAG training data.
2/19/2026
Human-AI Collaboration in Large Language Model-Integrated Building Energy Management Systems: The Role of User Domain Knowledge and AI Literacy
Empirical study examining how user domain knowledge and AI literacy affect interaction with LLM-integrated building energy management systems.
2/19/2026
Balancing Faithfulness and Performance in Reasoning via Multi-Listener Soft Execution
Multi-listener reinforcement learning approach to improve faithfulness of chain-of-thought reasoning in LLMs while maintaining performance.
2/19/2026
HiPER: Hierarchical Reinforcement Learning with Explicit Credit Assignment for Large Language Model Agents
Hierarchical reinforcement learning framework with explicit credit assignment for LLM agents on long-horizon multi-turn tasks.
2/19/2026
Conjugate Learning Theory: Uncovering the Mechanisms of Trainability and Generalization in Deep Neural Networks
Theoretical framework using convex conjugate duality to analyze trainability and generalization in deep neural networks with SGD.
2/19/2026
Beyond Learning: A Training-Free Alternative to Model Adaptation
Training-free approach to adapt language models by identifying and leveraging local module activations without retraining.
2/19/2026
Long-Tail Knowledge in Large Language Models: Taxonomy, Mechanisms, Interventions and Implications
Taxonomy and analysis of how LLMs handle long-tail knowledge from infrequent, domain-specific, cultural, and temporal domains.
2/19/2026
Graph neural network for colliding particles with an application to sea ice floe modeling
Graph Neural Network approach for sea ice modeling using collision physics, positioning ice pieces as nodes and interactions as edges.
2/19/2026
Are LLMs Ready to Replace Bangla Annotators?
Study evaluating LLMs as zero-shot annotators for Bangla hate speech detection, examining reliability and bias in low-resource language settings.
2/19/2026
Generative AI Usage of University Students: Navigating Between Education and Business
Qualitative study of generative AI usage among part-time university students in education and professional contexts.
2/19/2026
The Weight of a Bit: EMFI Sensitivity Analysis of Embedded Deep Learning Models
Analysis of electromagnetic fault injection attacks on embedded neural network models across different number representations.
2/19/2026
A Graph Meta-Network for Learning on Kolmogorov-Arnold Networks
Graph meta-network architecture for weight-space models that predict neural network accuracy on new datasets.
2/19/2026
A Self-Supervised Approach for Enhanced Feature Representations in Object Detection Tasks
Self-supervised learning approach for improving feature representations in object detection without labeled data.
2/19/2026
HAWX: A Hardware-Aware FrameWork for Fast and Scalable ApproXimation of DNNs
HAWX hardware-aware framework for DNN approximation using multi-level sensitivity scoring to guide AxC block integration across abstraction levels.
2/19/2026
Automated Histopathology Report Generation via Pyramidal Feature Extraction and the UNI Foundation Model
Framework combining pathology foundation model with transformer decoder for automated histopathology report generation from whole slide images.
2/19/2026
Designing Production-Scale OCR for India: Multilingual and Domain-Specific Systems
Multilingual OCR system for India using vision-language models through Chitrapathak series, balancing linguistic diversity and deployment constraints.
2/19/2026
Intra-Fairness Dynamics: The Bias Spillover Effect in Targeted LLM Alignment
Studies bias spillover effect in LLM fairness alignment, showing single-attribute mitigation can exacerbate disparities in untargeted dimensions.
2/19/2026
RoboGene: Boosting VLA Pre-training via Diversity-Driven Agentic Framework for Real-World Task Generation
RoboGene uses agentic framework for automated task generation and curation to improve vision-language-action model pre-training data diversity.
2/19/2026
IndicEval: A Bilingual Indian Educational Evaluation Framework for Large Language Models
IndicEval benchmarks LLMs on authentic UPSC, JEE, NEET exam questions in English and Hindi across STEM and humanities.
2/19/2026
Team of Thoughts: Efficient Test-time Scaling of Agentic Systems through Orchestrated Tool Calling
Team-of-Thoughts orchestrates heterogeneous agent models via tool calling to optimize multi-agent system performance through complementary capabilities.
2/19/2026
Learning to Learn from Language Feedback with Social Meta-Learning
Proposes social meta-learning approach enabling LLMs to proactively solicit and learn from conversational corrective feedback.
2/19/2026
From Growing to Looping: A Unified View of Iterative Computation in LLMs
Unifies looping and depth-growing in LLMs mechanistically, showing both exhibit convergent depth-wise signatures and improved reasoning.
2/19/2026
Interpretability-by-Design with Accurate Locally Additive Models and Conditional Feature Effects
CALMs balance interpretability of generalized additive models with accuracy by adding conditional pairwise interactions.
2/19/2026
Recursive language models for jailbreak detection: a procedural defense for tool-augmented agents
RLM-JB framework uses recursive language models for jailbreak detection in agentic systems executing tools over untrusted content.
2/19/2026
MerLean: An Agentic Framework for Autoformalization in Quantum Computation
MerLean automates formalization of quantum computation papers from LaTeX to Lean 4 code, producing 2,050 declarations from 114 statements.
2/19/2026
AIFL: A Global Daily Streamflow Forecasting Model Using Deterministic LSTM Pre-trained on ERA5-Land and Fine-tuned on IFS
LSTM-based deterministic model for global streamflow forecasting pre-trained on ERA5-Land reanalysis data and fine-tuned on operational forecasts.
2/19/2026
DataJoint 2.0: A Computational Substrate for Agentic Scientific Workflows
DataJoint 2.0 introduces relational workflow model for scientific data pipelines with provenance tracking and transactional guarantees, enabling agentic scientific workflows.
2/19/2026
FlowPrefill: Decoupling Preemption from Prefill Scheduling Granularity to Mitigate Head-of-Line Blocking in LLM Serving
FlowPrefill decouples preemption from prefill scheduling to reduce head-of-line blocking in LLM serving systems. Improves time-to-first-token latency.
2/19/2026
Explainable AI: Context-Aware Layer-Wise Integrated Gradients for Explaining Transformer Models
Layer-wise integrated gradients method for explaining Transformer predictions with context-awareness. Improves interpretability of deep neural models.
2/19/2026
Who can we trust? LLM-as-a-jury for Comparative Assessment
Proposes using LLMs as evaluators for comparative assessment with improved reliability modeling. Addresses bias and inconsistency in LLM judge performance.
2/19/2026
Causal and Compositional Abstraction
Formalizes causal abstraction between low and high-level models using natural transformations. Addresses interpretability and robustness in AI systems.
2/19/2026
A Systematic Evaluation of Sample-Level Tokenization Strategies for MEG Foundation Models
Systematic evaluation of tokenization strategies for MEG neuroimaging foundation models. Compares discretization approaches for continuous neural time series data.
2/19/2026
Almost Sure Convergence of Differential Temporal Difference Learning for Average Reward Markov Decision Processes
Convergence analysis for differential temporal difference learning in average reward reinforcement learning settings.