2/13/2026
AutoFigure: Generating and Refining Publication-Ready Scientific Illustrations
AutoFigure/FigureBench: LLM-based tool and benchmark (3,300 pairs) for automatically generating scientific illustrations from academic text.
2/13/2026
When AI Persuades: Adversarial Explanation Attacks on Human Trust in AI-Assisted Decision Making
Study of adversarial attacks targeting human trust in AI-assisted decision-making through manipulated LLM explanations at the cognitive/social level.
2/13/2026
DeepRead: Document Structure-Aware Reasoning to Enhance Agentic Search
DeepRead framework for agentic search using LLM tool-use and document structure awareness to improve multi-turn evidence acquisition in RAG systems.
2/13/2026
The Moltbook Illusion: Separating Human Influence from Emergent Behavior in AI Agent Societies
Analysis showing Moltbook AI agent behaviors attributed to emergent intelligence were largely human-driven; develops temporal fingerprinting method to detect human influence.
2/13/2026
When Evaluation Becomes a Side Channel: Regime Leakage and Structural Mitigations for Alignment Assessment
Study of regime leakage in AI agent evaluation: agents with situational awareness may exploit differences between evaluation and deployment to hide misaligned behavior.
2/13/2026
TreeTensor: Boost AI System on Nested Data with Constrained Tree-Like Tensor
TreeTensor data structure for AI systems handling nested/tree-like data with improved GPU parallelization for cognitive tasks beyond perception.
2/13/2026
Reinforcement Inference: Leveraging Uncertainty for Self-Correcting Language Model Reasoning
Reinforcement Inference: entropy-aware inference-time method enabling LLMs to self-correct through uncertainty-guided reasoning.
2/13/2026
MERIT Feedback Elicits Better Bargaining in LLM Negotiators
MERIT feedback framework and AgoraBench for improving LLM negotiation agents' strategic depth and adaptability.
2/13/2026
Spend Search Where It Pays: Value-Guided Structured Sampling and Optimization for Generative Recommendation
Value-guided sampling and RL optimization for generative recommendation models addressing probability-reward mismatch in decoding.
2/13/2026
FormalJudge: A Neuro-Symbolic Paradigm for Agentic Oversight
FormalJudge: neuro-symbolic framework using formal verification for safe oversight of LLM-based agents in high-stakes domains.
2/13/2026
Compiling High-Level Neural Network Specifications into VNN-LIB Queries
Developer tool compiling high-level neural network specifications into VNN-LIB verification queries for formal NN verification.
2/13/2026
NewsInterview: a Dataset and a Playground to Evaluate LLMs' Ground Gap via Informational Interviews
NewsInterview dataset of 40K interviews evaluating LLMs' grounding and strategic dialogue capabilities in information-seeking contexts.
2/13/2026
PBP: Post-training Backdoor Purification for Malware Classifiers
Post-training backdoor purification technique removing poisoning artifacts from malware classifiers without retraining.
2/13/2026
Compositional Generalization from Learned Skills via CoT Training: A Theoretical and Structural Analysis for Reasoning
Analyzes how CoT training enables LLMs to compose learned skills for generalization on complex reasoning tasks.
2/13/2026
Teaching LLMs According to Their Aptitude: Adaptive Reasoning for Mathematical Problem Solving
Adaptive reasoning method enabling LLMs to autonomously select between CoT and tool-integrated reasoning for math problem solving.
2/13/2026
Quantifying and Improving the Robustness of Retrieval-Augmented Language Models Against Spurious Features in Grounding Data
Studies spurious features in RAG systems' grounding data and proposes robustness improvements against implicit noise.
2/13/2026
A Multi-Fidelity Control Variate Approach for Policy Gradient Estimation
Multi-fidelity policy gradients method combining low and high-fidelity simulators to reduce data requirements in RL training.
2/13/2026
Right Reward Right Time for Federated Learning
Incentive mechanism for federated learning that prioritizes high-quality contributions during critical early training periods.
2/13/2026
Remote Sensing Retrieval-Augmented Generation: Bridging Remote Sensing Imagery and Comprehensive Knowledge with a Multi-Modal Dataset and Retrieval-Augmented Generation Model
RAG system for remote sensing combining VLMs with retrieval augmentation for scene understanding and visual QA on satellite imagery.
2/13/2026
Social Human Robot Embodied Conversation (SHREC) Dataset: Benchmarking Foundational Models' Social Reasoning
SHREC dataset: 400 videos with 10K annotations benchmarking foundation models' social reasoning in human-robot interactions.
2/13/2026
Can LLM-based Financial Investing Strategies Outperform the Market in Long Run?
Evaluates LLM-based investment strategies across long timeframes and multiple stocks, revealing survivorship bias and limited generalization.
2/13/2026
Defending the Edge: Representative-Attention Defense against Backdoor Attacks in Federated Learning
Defense mechanism against backdoor attacks in federated learning using representative-attention to detect adaptive poisoning attacks.
2/13/2026
EvoGPT: Leveraging LLM-Driven Seed Diversity to Improve Search-Based Test Suite Generation
EvoGPT hybrid system combining LLM-based test generation with search-based optimization to improve automated unit test suite diversity.
2/13/2026
AMAQA: A Metadata-based QA Dataset for RAG Systems
AMAQA: open-access QA dataset integrating metadata for evaluating RAG systems in text and structured data retrieval tasks.
2/13/2026
Prompt Engineer: Analyzing Hard and Soft Skill Requirements in the AI Job Market
Analysis of 20,662 LinkedIn job postings revealing hard and soft skills required for emerging Prompt Engineer role in AI job market.
2/13/2026
Truth in the Few: High-Value Data Selection for Efficient Multi-Modal Reasoning
High-value data selection for multi-modal LLM reasoning via reinforcement learning, achieving comparable performance to full datasets with reduced compute.
2/13/2026
Bootstrapping Action-Grounded Visual Dynamics in Unified Vision-Language Models
Vision-language models improved for forward dynamics prediction through asymmetric fine-tuning on inverse dynamics, bootstrapping physical reasoning.
2/13/2026
AutoDiscovery: Open-ended Scientific Discovery via Bayesian Surprise
AutoDiscovery uses Bayesian surprise to enable autonomous scientific discovery where LLM agents self-direct hypothesis generation beyond human-specified goals.
2/13/2026
PhreshPhish: A Real-World, High-Quality, Large-Scale Phishing Website Dataset and Benchmark
PhreshPhish dataset and benchmark with 2M+ phishing samples for ML-based detection addressing data leakage and unrealistic base rate problems.
2/13/2026
Beyond Model Base Retrieval: Weaving Knowledge to Master Fine-grained Neural Network Design
Novel approach to neural network design combining architectural search with fine-grained modifications using edit-effect evidence for performance optimization.
2/13/2026
Parallelism Meets Adaptiveness: Scalable Documents Understanding in Multi-Agent LLM Systems
Multi-agent LLM framework enables dynamic task routing and bidirectional feedback for scalable collaborative document understanding in complex domains.
2/13/2026
Trustworthiness of Legal Considerations for the Use of LLMs in Education
Comparative analysis of regulatory and ethical frameworks for LLM deployment in education across EU, UK, US, China, and Gulf regions.
2/13/2026
MCPSecBench: A Systematic Security Benchmark and Playground for Testing Model Context Protocols
MCPSecBench formalizes security for Model Context Protocol connecting LLM agents to data sources and external tools with systematic threat modeling.
2/13/2026
A Large-Scale Benchmark for Evaluating Large Language Models on Medical Question Answering in Romanian
MedQARo benchmark evaluates LLMs on 105,880 Romanian medical QA pairs requiring keyword extraction and clinical reasoning.
2/13/2026
AI Agentic Vulnerability Injection And Transformation with Optimized Reasoning
AI agents use optimized reasoning for automated software vulnerability detection with synthetic dataset generation addressing label noise problems.
2/13/2026
Binary Autoencoder for Mechanistic Interpretability of Large Language Models
Binary autoencoders improve mechanistic interpretability of LLMs by extracting sparse, atomized features from hidden states.
2/13/2026
CoSpaDi: Compressing LLMs via Calibration-Guided Sparse Dictionary Learning
CoSpaDi compression method for LLMs using calibration-guided sparse dictionary learning as training-free alternative to low-rank approximation.
2/13/2026
VibeCodeHPC: An Agent-Based Iterative Prompting Auto-Tuner for HPC Code Generation Using LLMs
Multi-agent LLM system for automatic HPC code generation and tuning on supercomputers using iterative prompting.
2/13/2026
KVComm: Enabling Efficient LLM Communication through Selective KV Sharing
KVComm framework enabling efficient communication between LLMs in multi-agent systems via selective key-value cache sharing.
2/13/2026
Alignment Tipping Process: How Self-Evolution Pushes LLM Agents Off the Rails
Identifies Alignment Tipping Process risk in self-evolving LLM agents that abandon alignment constraints through real-world adaptation.
2/13/2026
On the optimization dynamics of RLVR: Gradient gap and step size thresholds
Theoretical analysis of RLVR with binary feedback for LLM post-training, introducing Gradient Gap concept for understanding training dynamics.
2/13/2026
Test-Time Alignment of LLMs via Sampling-Based Optimal Control in pre-logit space
Test-time alignment of LLMs via sampling-based optimal control in pre-logit space without fine-tuning computational costs.
2/13/2026
EGG-SR: Embedding Symbolic Equivalence into Symbolic Regression via Equality Graph
Symbolic regression framework incorporating symbolic equivalence via equality graphs to reduce search space in scientific discovery.
2/13/2026
Evaluating Modern Large Language Models on Low-Resource and Morphologically Rich Languages:A Cross-Lingual Benchmark Across Cantonese, Japanese, and Turkish
Benchmark evaluating seven modern LLMs including GPT-4, Claude, LLaMA, Mistral on low-resource and morphologically rich languages.
2/13/2026
Self-Adaptive Graph Mixture of Models
Graph neural network architecture combining mixture-of-experts with adaptive routing for improved performance on graph-structured data.
2/13/2026
MapReduce LoRA: Advancing the Pareto Front in Multi-Preference Optimization for Generative Models
MapReduce LoRA and Reward-aware Token Embedding methods for multi-preference optimization in generative models addressing alignment trade-offs.
2/13/2026
SUGAR: A Sweeter Spot for Generative Unlearning of Many Identities
Framework for removing multiple identities from 3D generative models without retraining, addressing consent and model unlearning.
2/13/2026
Demystifying LLM-as-a-Judge: Analytically Tractable Model for Inference-Time Scaling
Analytical model for LLM inference-time scaling using Bayesian linear regression with reward-weighted sampling to understand test-time computation.
2/13/2026
Block-Recurrent Dynamics in Vision Transformers
Mechanistic analysis of Vision Transformers introducing Block-Recurrent Hypothesis to interpret depth as dynamical computational flow.
2/13/2026
IBISAgent: Reinforcing Pixel-Level Visual Reasoning in MLLMs for Universal Biomedical Object Referring and Segmentation
Framework for pixel-level visual reasoning in medical multimodal LLMs using reinforcement learning for biomedical object referring and segmentation without catastrophic forgetting.