1. Research Direction Hotness Analysis

The 82 papers from this period showcase a strong focus on enhancing Large Language Models (LLMs), applying AI to specialized domains, and ensuring AI systems are safe and efficient. The following key research areas have emerged as the most active.

2. Author Relationship Graph

The author network reveals several clusters of collaboration. While most authors contributed to a single paper in this dataset, a few, like Divij Handa, Raghav Sharma, and Manan Mehta, appear on multiple papers, indicating active research streams. The graph below visualizes the key collaboration links among authors who co-authored papers in this batch. Larger nodes or more connections suggest a central role in the current research landscape.

3. Technical Innovation Summary

This period's papers introduce several notable technical and methodological innovations:

• Methodological Innovations:
  • MITS (Mutual Information Tree Search): A new framework for tree-search reasoning in LLMs that provides more reliable quality assessment of reasoning steps.
  • GuidedSampling: An inference algorithm that decouples exploration and generation to produce more diverse candidate solutions from LLMs.
  • Adversarial Agent Collaboration (ACToR): A novel GAN-inspired approach using a generator and discriminator agent for complex tasks like C-to-Rust code translation.
  • HydroFusion-LMF: A semi-supervised framework for hydrological forecasting that fuses multiple network architectures and adapts large pre-trained models.
• Key Technical Breakthroughs:
  • EvoEngineer: A system that masters automated evolution of CUDA kernel code using LLMs, tackling a critical bottleneck in AI performance.
  • Spatial CAPTCHA: A new CAPTCHA design that tests spatial reasoning, proving much more robust against modern Multimodal LLMs than traditional CAPTCHAs.
  • LLM Chemistry: A framework to quantify synergistic or antagonistic behavior between collaborating LLMs, moving beyond simple output assessment to analyze the collaborative process itself.
  • SATER (Self-Aware and Token-Efficient Routing): An intelligent model routing system that decides whether to use a powerful (expensive) LLM or a smaller (cheaper) SLM based on its own confidence score.
• Application Domain Expansion:
  • Penetration Testing: The introduction of `PentestMCP`, a toolkit for creating AI agents that can perform automated security testing.
  • Hadith Text Processing: The `Rezwan` project demonstrates the use of LLMs for large-scale processing and enrichment of a 1.2M-entry religious text corpus.
  • Biomechanical Feedback: A framework that translates 3D biomechanical data from tennis strokes into actionable, natural language feedback for players and coaches.
  • Mission-Driven Organizations: The first studies are emerging on how non-profits and humanitarian organizations are adopting AI, revealing unique challenges and opportunities.

4. Full Paper List (82 Papers)

• MITS: Enhanced Tree Search Reasoning for LLMs via Pointwise Mutual Information
• Rainbow Padding: Mitigating Early Termination in Instruction-Tuned Diffusion LLMs
• H-DDx: A Hierarchical Evaluation Framework for Differential Diagnosis
• OptAgent: Optimizing Query Rewriting for E-commerce via Multi-Agent Simulation
• Algorithm Generation via Creative Ideation
• Rare Text Semantics Were Always There in Your Diffusion Transformer
• Deep learning the sources of MJO predictability: a spectral view of learned features
• A Hybrid Co-Finetuning Approach for Visual Bug Detection in Video Games
• Deep Domain Adaptation for Turbofan Engine Remaining Useful Life Prediction: Methodologies, Evaluation and Future Trends
• PentestMCP: A Toolkit for Agentic Penetration Testing
• Cross-Modal Content Optimization for Steering Web Agent Preferences
• Explainable but Vulnerable: Adversarial Attacks on XAI Explanation in Cybersecurity Applications
• Predicting Stock Price Movement with LLM-Enhanced Tweet Emotion Analysis
• Towards Unsupervised Speech Recognition at the Syllable-Level
• Operationalizing Data Minimization for Privacy-Preserving LLM Prompting
• MonitorVLM:A Vision Language Framework for Safety Violation Detection in Mining Operations
• MedReflect: Teaching Medical LLMs to Self-Improve via Reflective Correction
• REG: A Regularization Optimizer for Robust Training Dynamics
• Mind the Goal: Data-Efficient Goal-Oriented Evaluation of Conversational Agents and Chatbots using Teacher Models
• Referring Expression Comprehension for Small Objects
• Artery-Vein Segmentation from Fundus Images using Deep Learning
• TreePrompt: Leveraging Hierarchical Few-Shot Example Selection for Improved English-Persian and English-German Translation
• Code4MeV2: a Research-oriented Code-completion Platform
• EvoEngineer: Mastering Automated CUDA Kernel Code Evolution with Large Language Models
• You Have Been LaTeXpOsEd: A Systematic Analysis of Information Leakage in Preprint Archives Using Large Language Models
• Adaptively Sampling-Reusing-Mixing Decomposed Gradients to Speed Up Sharpness Aware Minimization
• GuidedSampling: Steering LLMs Towards Diverse Candidate Solutions at Inference-Time
• Rezwan: Leveraging Large Language Models for Comprehensive Hadith Text Processing: A 1.2M Corpus Development
• Lightweight and Data-Efficient MultivariateTime Series Forecasting using Residual-Stacked Gaussian (RS-GLinear) Architecture
• Beyond Token Length: Step Pruner for Efficient and Accurate Reasoning in Large Language Models
• ReTiDe: Real-Time Denoising for Energy-Efficient Motion Picture Processing with FPGAs
• A4FN: an Agentic AI Architecture for Autonomous Flying Networks
• Small Language Models for Agentic Systems: A Survey of Architectures, Capabilities, and Deployment Trade offs
• Adaptive and Explainable AI Agents for Anomaly Detection in Critical IoT Infrastructure using LLM-Enhanced Contextual Reasoning
• Designing Empirical Studies on LLM-Based Code Generation: Towards a Reference Framework
• Spatial CAPTCHA: Generatively Benchmarking Spatial Reasoning for Human-Machine Differentiation
• AI Adoption Across Mission-Driven Organizations
• PoseGaze-AHP: A Knowledge-Based 3D Dataset for AI-Driven Ocular and Postural Diagnosis
• Multi-Modal Oral Cancer Detection Using Weighted Ensemble Convolutional Neural Networks
• Adversarial Agent Collaboration for C to Rust Translation
• Kantian-Utilitarian XAI: Meta-Explained
• Refactoring with LLMs: Bridging Human Expertise and Machine Understanding
• On the Convergence and Size Transferability of Continuous-depth Graph Neural Networks
• SPEAR: Soft Prompt Enhanced Anomaly Recognition for Time Series Data
• Towards Carbon-Aware Container Orchestration: Predicting Workload Energy Consumption with Federated Learning
• What Can You Do When You Have Zero Rewards During RL?
• Artificial-Intelligence Grading Assistance for Handwritten Components of a Calculus Exam
• Deep Learning-Based Multi-Factor Authentication: A Survey of Biometric and Smart Card Integration Approaches
• Domain-Adapted Granger Causality for Real-Time Cross-Slice Attack Attribution in 6G Networks
• LLM-Driven Rubric-Based Assessment of Algebraic Competence in Multi-Stage Block Coding Tasks with Design and Field Evaluation
• Direct Routing Gradient (DRGrad): A Personalized Information Surgery for Multi-Task Learning (MTL) Recommendations
• Enhanced Urban Traffic Management Using CCTV Surveillance Videos and Multi-Source Data Current State Prediction and Frequent Episode Mining
• Latent Mixture of Symmetries for Sample-Efficient Dynamic Learning
• Deep Reinforcement Learning for Multi-Agent Coordination
• Neon: Negative Extrapolation From Self-Training Improves Image Generation
• Understanding the Role of Training Data in Test-Time Scaling
• Can an LLM Induce a Graph? Investigating Memory Drift and Context Length
• Neural Bayesian Filtering
• Implicit Models: Expressive Power Scales with Test-Time Compute
• Does higher interpretability imply better utility? A Pairwise Analysis on Sparse Autoencoders
• EmbodiSwap for Zero-Shot Robot Imitation Learning
• Bridging the Gap Between Multimodal Foundation Models and World Models
• Cost Efficient Fairness Audit Under Partial Feedback
• Mechanistic Interpretability of Socio-Political Frames in Language Models
• 6G-Enabled Digital Twin Framework for Real-Time Cyber-Physical Systems: An Experimental Validation with Industrial Bearing Fault Detection
• Diverse Text-to-Image Generation via Contrastive Noise Optimization
• Detecting Invariant Manifolds in ReLU-Based RNNs
• Proximal Diffusion Neural Sampler
• The Hidden Game Problem
• AI-Assisted Pleural Effusion Volume Estimation from Contrast-Enhanced CT Images
• Unlocking Reasoning Capabilities in LLMs via Reinforcement Learning Exploration
• Optimal Scaling Needs Optimal Norm
• Talking Tennis: Language Feedback from 3D Biomechanical Action Recognition
• LLM Chemistry Estimation for Multi-LLM Recommendation
• Strategy Logic, Imperfect Information, and Hyperproperties
• Quantifying Risks in Multi-turn Conversation with Large Language Models
• SATER: A Self-Aware and Token-Efficient Approach to Routing and Cascading
• The Enduring Dominance of Deep Neural Networks: A Critical Analysis of the Fundamental Limitations of Quantum Machine Learning and Spiking Neural Networks
• Less Diverse, Less Safe: The Indirect But Pervasive Risk of Test-Time Scaling in Large Language Models
• HydroFusion-LMF: Semi-Supervised Multi-Network Fusion with Large-Model Adaptation for Long-Term Daily Runoff Forecasting
• LLM-Guided Evolutionary Program Synthesis for Quasi-Monte Carlo Design
• Dissecting Larval Zebrafish Hunting using Deep Reinforcement Learning Trained RNN Agents