May 2025 — page 1 of 2
- New Method Scales LLM Prompt Optimization Using Momentum-Based Sampling· AI
- How programmers visually scan code when hunting memory bugs· Software Engineering
- New AI approach generates medical CT scans directly from radiologist reports· AI
- Deep learning framework learns optimal OFDM channel filters while keeping inference simple· AI
- Study Questions GraphRAG Performance Claims, Finds Gains Less Impressive Than Reported· AI
- LLM-Guided Testing Tool Boosts Web App Coverage by up to 24%· AI
- Study Maps Where Personas Are Encoded Inside Large Language Models· AI
- One-Step Diffusion Models Achieve 2000× Speedup in Text Generation· AI
- New Training Method Achieves Better LLM Performance With Fewer Trainable Parameters· AI
- New benchmark reveals limits in keeping AI language models scientifically current· AI
- New AI Model Scales Up Chronic Disease Prediction Across Hospital Networks· AI
- Diffusion AI maps urban airflow with minimal sensors, cutting monitoring costs· AI
- Researchers Introduce MiCRo to Capture Diverse Human Preferences in AI Reward Modeling· AI
- Study-Created Emotional Content Differs From Real Posts in Key Ways· AI
- Teaching LLMs to Negotiate Like Humans: Theory of Mind Reasoning Boosts Fair Deals· AI
- New benchmark tests LLMs' ability to infer complex string-transformation rules· AI
- New Benchmark Tests How Well AI Understands Complex Social Situations· AI
- Unified Vision-Language Models Show Cross-Task Benefits Only When Visual Spaces Align· AI
- State Space Models Can Pool Document Representations to Cut Inference Costs· AI
- Framework Proposes Tiered AI Autonomy for Security Operations Centers to Enhance Human Decision-Making· Agentic AI
- New algorithm balances robot safety and path efficiency with dynamic weighting· Robotics
- Multi-agent framework uses game theory to strengthen vision-language AI reasoning· Agentic AI
- New framework trains AI agents to collaborate with unfamiliar teammates by dynamically generating challenging training partners· Agentic AI
- Large reasoning models tackle frontier mathematics with new AI Mathematician framework· AI
- New survey maps how AI copilots learn and adapt to user preferences· AI
- Quantum approach to generating synthetic tabular data shows promise against classical models· Quantum AI
- New Method Uses Kronecker Factorization to Make Sparse Autoencoders More Efficient and Interpretable· AI
- New Framework Uses Shapley Values to Fairly Credit Sources in AI Summaries· AI
- Study uncovers how diffusion models defend against adversarial attacks· AI
- Rule-Based Math Verifiers Miss Correct Answers; Model-Based Ones Fall for Reward Hacking· AI
- New Framework Boosts LLM Editing Efficiency by Compressing Context Into Special Tokens· AI
- How Language Models Foster Insight Through Dialogue: A Framework for Human-AI Co-Thinking· AI
- When Adding Retrieved Information Hurts: Study Reveals RAG's Hidden Pitfalls for LLMs· AI
- New technique enables AI safety systems to adapt on-the-fly as conditions change· Agentic AI
- Federated Healthcare AI Lets Lower-Compliance Institutions Join Without Sacrificing Accuracy· AI
- Researchers outline seven critical security gaps in multi-agent LLM systems working across organizations· AI
- Researchers develop method to combine simple reasoning steps into complex AI reasoning· AI
- New technique helps vision-language models handle high-resolution images without losing token coherence· AI
- New metric reveals how data diversity drives LLM reasoning—and a framework to generate better training data· AI
- Foundation Models and Search-Based Engineering: New Roadmap for AI-Driven Software Development· AI
- New Selective Language Modeling Approach Cuts Training Costs for Large Language Models· AI
- New metric reveals how well AI speech models actually understand audio· AI
- New Moderation Framework Trains AI Safety Filters at Scale Without Human Labeling· AI
- New AI Model Identifies Nearly 1,600 Weed Species with 91% Accuracy· Agentic AI
- New Chi-Square Wavelet Method Improves Anomaly Detection in Complex Networks· AI
- Researchers Create Dataset to Push LLMs Beyond Single-Digit Constraint Following· AI
- LLM Agents Abandon Ethics When Profit Incentives Rise, Study Finds· AI
- AI agents need open standards to prevent fragmented ecosystem, researchers argue· Agentic AI
- New compression technique slashes memory use for AI reasoning models· AI
- New Framework Improves Document AI by Progressively Focusing on Relevant Regions· AI
- Researchers propose reputation metrics to help open-source maintainers vet security risks from external contributors· Software Security
- New benchmark tackles the challenge of reliably evaluating AI systems that generate multiple types of content· AI
- New technique cuts CNN fine-tuning parameters by 80% while keeping performance intact· AI
- Knowledge-Augmented Reasoning Pushes Language Models Forward on Abstract Problem-Solving· AI
- How Familiar Falsehoods Make LLMs Overconfident· AI
- New benchmark tests how well AI models handle real-world web design tasks· AI
- New AI-Assisted Method Speeds Up Cycle Counting in Networks and Matrices· AI
- Multi-Agent Debate Framework Detects Duplicate Projects, Saving Millions in Investment· AI
- New Quantization Technique Cuts LLM Memory Use Without Training Data· AI
- Advanced reasoning models struggle with personalization—new framework aims to fix it· AI
- Game Theory Offers Framework for Understanding Human-AI Competition and Cooperation· AI
- New evaluation metric combines image, text, and semantic analysis to better judge AI-generated captions· AI
- New benchmark tests whether AI foundation models live up to clinical promises in healthcare data· AI, Healthcare
- LLM Code Judges Show Systematic Bias Against Cosmetic Code Variations· AI
- Echo-State Networks Show Promise—and Limits—in Capturing Rare Events in Chaotic Systems· AI
- Survey Maps Hidden Reasoning in AI Models—Moving Beyond Explicit Step-by-Step Chains· AI
- Internal "Belief" Signals Determine Whether LLMs Will Correct Their Own Errors· AI
- Efficient Personalized Jargon Detection Makes Technical Documents Accessible Without Heavy Training· AI
- Researchers uncover hidden attack vector: "dormant" adversarial behaviors triggered during LLM finetuning· AI Security
- Language Models Perform Better on Some Questions When Queried in Non-English Languages· AI
- AI-Guided Quantum Annealing Improves Performance on Real Hardware· Quantum Computing
- New hybrid approach combines LLMs with proof sketches to boost automated theorem proving accuracy· Agentic AI
- New framework unites geometry and neural networks for faster travel-time predictions· AI
- Why AI Reasoning Models Ignore "Think Less" Instructions—and What That Reveals About Their Minds· AI
- New technique makes AI embeddings explainable by measuring what drives similarity· AI
- Smaller LLMs Learn Complex Reasoning With Structured Knowledge Extraction Framework· Agentic AI
- New Framework Brings Transparency to Particle Swarm Optimization Algorithms· AI
- New Fact-Checking Framework Tackles LLM Reliability in Urdu· AI
- LLMs show promise as source of probabilistic knowledge for Bayesian Networks· AI
- New Mathematical Framework Extends Causal Probability Analysis Beyond Binary Cases· AI
- Researchers reveal how AI monitors can be evaded and propose ensemble defense· AI Safety
- New neural network training method replaces backpropagation with a second forward pass· AI
- New Multi-Agent Framework Splits Knowledge Graphs to Speed Up AI-Powered Search· AI
- Researchers Build Synthetic Training Grounds to Teach Robots Spatial Understanding· AI
- Researchers build first benchmark to train AI on extracting legal principles from court cases· AI
- New Evaluation Method Reveals High Vulnerability of Open-Source LLMs to Prompt Injection Attacks· AI Security
- New benchmark reveals major gaps in AI diagnostic reasoning for medical settings· AI
- New Method Combines Induction and Transduction for Better Program Synthesis· AI
- Majority-group bias in machine learning: new theory explains how standard training amplifies unfairness· AI Fairness
- Language Models Struggle as Formal Problem Solvers, Study Finds· AI
- New Framework Brings Transparency and Flexibility to Language Model Reward Models· AI
- Researchers blend reinforcement learning with symbolic reasoning to build smarter automated planners· AI
- Agentic AI System Automates Crystal Structure Analysis from Microscopy Images· Agentic AI
- New benchmark exposes gaps in discrete visual tokenizers for image generation· AI
- Large Language Models Struggle to Truly Understand Code Context, Relying on Pattern Matching Instead· AI
- GitHub Study Reveals Narrow Adoption of Chaos Engineering Tools in Open-Source Projects· Software Engineering
- New Dataset Benchmarks Training Strategies for AI Models Predicting Blood Flow in Aneurysms· AI
- New Benchmark Reveals When Vision-Language Models Should Plan vs. Ground Symbolic Reasoning· AI
- New technique trains compact AI models 1,000x faster by cloning teacher networks efficiently· AI
- Wavelet-based fine-tuning offers sparse alternative to LoRA for efficient vision model adaptation· AI