2/18/2026
World-Model-Augmented Web Agents with Action Correction
WAC: LLM-based web agent with world model for reasoning about environment changes and predicting execution risks before action.
2/18/2026
Improving LLM Reliability through Hybrid Abstention and Adaptive Detection
Hybrid approach combining abstention and adaptive detection to improve LLM safety-utility tradeoff without static confidence thresholds.
2/18/2026
GenAI-LA: Generative AI and Learning Analytics Workshop (LAK 2026), April 27--May 1, 2026, Bergen, Norway
EduEVAL-DB dataset with 854 teacher explanations across K-12 science/language/social science for training pedagogical evaluators and AI tutors.
2/18/2026
Quantifying construct validity in large language model evaluations
Framework for evaluating construct validity of LLM benchmarks, separating actual capabilities from benchmark artifacts like contamination and annotation errors.
2/18/2026
RUVA: Personalized Transparent On-Device Graph Reasoning
On-device graph-based reasoning system for personalized AI with transparency and accountability, addressing limitations of RAG-based black-box systems.
2/18/2026
How Vision Becomes Language: A Layer-wise Information-Theoretic Analysis of Multimodal Reasoning
Layer-wise analysis of multimodal Transformers using information theory to decompose how visual and linguistic information contribute to predictions in vision-question tasks.
2/18/2026
CARE Drive A Framework for Evaluating Reason-Responsiveness of Vision Language Models in Automated Driving
Research on evaluating vision-language model reasoning in autonomous driving scenarios beyond performance metrics.
2/18/2026
PERSONA: Dynamic and Compositional Inference-Time Personality Control via Activation Vector Algebra
Presents PERSONA, a training-free framework for dynamic personality control in LLMs using activation vector algebra without fine-tuning.
2/18/2026
Recursive Concept Evolution for Compositional Reasoning in Large Language Models
Introduces recursive concept evolution method to improve compositional reasoning in LLMs on benchmarks like ARC-AGI-2 and MATH by evolving latent representations.
2/18/2026
GlobeDiff: State Diffusion Process for Partial Observability in Multi-Agent Systems
Proposes GlobeDiff, a diffusion-based approach for multi-agent coordination under partial observability, addressing limitations of belief-state and communication methods.
2/18/2026
This human study did not involve human subjects: Validating LLM simulations as behavioral evidence
Methodological study validating when LLM simulations produce valid behavioral evidence for causal inference in social science research.
2/18/2026
Enhancing Building Semantics Preservation in AI Model Training with Large Language Model Encodings
Method using LLM encodings to improve semantic representation of building objects for AI training in construction and engineering.
2/18/2026
Developing AI Agents with Simulated Data: Why, what, and how?
Framework and reference guide for simulation-based synthetic data generation techniques for training AI agents and subsymbolic models.
2/18/2026
LemonadeBench: Evaluating the Economic Intuition of Large Language Models in Simple Markets
Benchmark evaluating LLM economic intuition and decision-making through simulated lemonade stand business management tasks.
2/18/2026
EduResearchBench: A Hierarchical Atomic Task Decomposition Benchmark for Full-Lifecycle Educational Research
Evaluation benchmark with hierarchical task decomposition for assessing LLM capabilities in full-lifecycle educational research workflows.
2/18/2026
CircuChain: Disentangling Competence and Compliance in LLM Circuit Analysis
Circuit analysis framework distinguishing competence from compliance in LLM reasoning under engineering constraints and methodological conventions.
2/18/2026
Indic-TunedLens: Interpreting Multilingual Models in Indian Languages
Interpretability framework for multilingual LLMs in Indian languages using shared affine transformations for cross-lingual understanding.
2/18/2026
GRACE: an Agentic AI for Particle Physics Experiment Design and Simulation
Agentic AI system for autonomous experimental design in particle physics using natural language prompts and Monte Carlo simulation.
2/18/2026
Beyond Context Sharing: A Unified Agent Communication Protocol (ACP) for Secure, Federated, and Autonomous Agent-to-Agent (A2A) Orchestration
Unified agent communication protocol for secure, federated, decentralized autonomous agent-to-agent orchestration across platforms.
2/18/2026
CLOT: Closed-Loop Global Motion Tracking for Whole-Body Humanoid Teleoperation
Real-time whole-body humanoid teleoperation system with closed-loop global motion tracking to prevent pose drift.
2/18/2026
Safe-SDL:Establishing Safety Boundaries and Control Mechanisms for AI-Driven Self-Driving Laboratories
Safety framework for autonomous self-driving laboratories integrating AI with robotic automation for closed-loop experimentation.
2/18/2026
Structural Divergence Between AI-Agent and Human Social Networks in Moltbook
Analysis comparing interaction network structures of AI agents versus humans in co-inhabited online platform Moltbook.
2/18/2026
CGRA-DeBERTa Concept Guided Residual Augmentation Transformer for Theologically Islamic Understanding
CGRA-DeBERTa transformer with LoRA adaptations for question-answering over classical Islamic texts using concept-guided residual augmentation.
2/18/2026
Exploiting Layer-Specific Vulnerabilities to Backdoor Attack in Federated Learning
Analysis of backdoor attack vulnerabilities in federated learning systems exploiting layer-specific weaknesses during distributed model training.
2/18/2026
ScrapeGraphAI-100k: A Large-Scale Dataset for LLM-Based Web Information Extraction
Large-scale dataset of 100k real-world LLM-based web information extraction events capturing structural web context, released from ScrapeGraphAI telemetry.
2/18/2026
Weight space Detection of Backdoors in LoRA Adapters
Detection method for backdoor attacks in LoRA adapters by analyzing weight space without requiring test data or knowledge of trigger patterns.
2/18/2026
OpaqueToolsBench: Learning Nuances of Tool Behavior Through Interaction
Benchmark dataset and framework for evaluating LLM agents' ability to learn tool behavior and improve documentation for opaque real-world tools through interaction.
2/18/2026
Colosseum: Auditing Collusion in Cooperative Multi-Agent Systems
Colosseum framework for auditing collusive behavior in multi-agent systems where LLM agents communicate and coordinate through free-form language.
2/18/2026
MAVRL: Learning Reward Functions from Multiple Feedback Types with Amortized Variational Inference
Bayesian inference method for learning reward functions from heterogeneous feedback types including demonstrations, comparisons, ratings, and stops.
2/18/2026
Automatically Finding Reward Model Biases
Approach to automatically detect reward model biases in LLMs using iterative LLM proposals, addressing spurious attributes like length, hallucinations, and sycophancy.
2/18/2026
Closing the Distribution Gap in Adversarial Training for LLMs
Method to improve adversarial training for LLMs by addressing distribution gaps that leave models vulnerable to simple in-distribution adversarial examples like prompt rewriting.
2/18/2026
GenAI for Systems: Recurring Challenges and Design Principles from Software to Silicon
Cross-stack analysis of generative AI applications in computing systems design from code generation through hardware design space exploration to RTL synthesis.
2/18/2026
MyoInteract: A Framework for Fast Prototyping of Biomechanical HCI Tasks using Reinforcement Learning
Framework for prototyping biomechanical HCI tasks using reinforcement learning, improving usability and interpretability for interaction design.
2/18/2026
Decision Making under Imperfect Recall: Algorithms and Benchmarks
Benchmark suite for imperfect-recall decision problems in game theory, covering scenarios where agents have limited memory and communication.
2/18/2026
How to Train Your Long-Context Visual Document Model
Study of training long-context vision language models up to 344K tokens, analyzing pretraining, finetuning, and preference optimization with reproducible recipes.
2/18/2026
Fast and Effective On-policy Distillation from Reasoning Prefixes
On-policy distillation technique for training student LLMs using teacher supervision on token-level trajectories, improving generalization over off-policy methods with reduced computational cost.
2/18/2026
From Diagnosis to Inoculation: Building Cognitive Resistance to AI Disempowerment
AI literacy framework with learning outcomes to help users build cognitive resistance against disempowerment from AI assistant interactions including reality and value distortion.
2/18/2026
Accelerating Large-Scale Dataset Distillation via Exploration-Exploitation Optimization
Dataset distillation method using exploration-exploitation optimization to compress large-scale data into synthetic datasets while maintaining model performance and reducing training costs.
2/18/2026
Visual Persuasion: What Influences Decisions of Vision-Language Models?
Framework systematically studying visual preferences and decision factors of vision-language models.
2/18/2026
High-Fidelity Network Management for Federated AI-as-a-Service: Cross-Domain Orchestration
Network management and orchestration platform for federated AI-as-a-Service deployment.
2/18/2026
Complex-Valued Unitary Representations as Classification Heads for Improved Uncertainty Quantification in Deep Neural Networks
Quantum-inspired classification head using complex-valued unitary transformations for uncertainty quantification.
2/18/2026
AI-Paging: Lease-Based Execution Anchoring for Network-Exposed AI-as-a-Service
Network orchestration framework for AI-as-a-Service model selection and execution placement.
2/18/2026
The Information Geometry of Softmax: Probing and Steering
Analyzes geometric structure of softmax representations in neural networks for interpretability.
2/18/2026
Hybrid Federated and Split Learning for Privacy Preserving Clinical Prediction and Treatment Optimization
Hybrid federated and split learning framework for privacy-preserving clinical prediction across institutions.
2/18/2026
Sparrow: Text-Anchored Window Attention with Visual-Semantic Glimpsing for Speculative Decoding in Video LLMs
Speculative decoding optimization for video LLMs using text-anchored window attention and visual caching.
2/18/2026
On Surprising Effectiveness of Masking Updates in Adaptive Optimizers
Shows random masking in RMSProp optimizer outperforms sophisticated preconditioners for LLM training.
2/18/2026
Unforgeable Watermarks for Language Models via Robust Signatures
Watermarking scheme for LLMs preventing false attribution with robust signature guarantees.
2/18/2026
SCENE OTA-FD: Self-Centering Noncoherent Estimator for Over-the-Air Federated Distillation
Pilot-free aggregation primitive for over-the-air federated distillation using energy signaling.
2/18/2026
Prescriptive Scaling Reveals the Evolution of Language Model Capabilities
Establishes prescriptive scaling laws for foundation models given compute budgets and downstream accuracy.
2/18/2026
A Scalable Curiosity-Driven Game-Theoretic Framework for Long-Tail Multi-Label Learning in Data Mining
Game-theoretic framework for long-tail multi-label classification in large-scale data mining.