May 2026 — page 1 of 2
- New method improves medical AI by grounding images in research article text· AI
- New Framework Maps When Multiple AI Judges Need Simple vs. Complex Calibration Models· AI
- New Benchmark Translates Python Tests to Formal Verification Challenges for AI· AI
- New benchmark reveals over half of leading AI coding agents fail safety checks in real project environments· AI
- Sakura Framework Bridges Gap Between Complex Real-World Tests and Auto-Generated Code· AI
- Researchers Propose Framework to Fairly Compare Security of Human and AI-Generated Code· AI
- Study: External Security Evidence Can Cut Java API Misuse in AI-Generated Code by Up to 32%· AI Security
- Agentic LLMs With Active Retrieval Hit 62% Success on Infrastructure Code Tasks· Agentic AI
- Study maps safety failures in AI coding agents across real-world development· Agentic AI
- Community Developers Train Gemma Models for Reasoning with Tunix and TPUs· Agentic AI
- Study Reveals How AI Coding Agents Consistently Fail Developers in Real-World Use· Agentic AI
- RAIL Guard fixes unsafe AI agent outputs without discarding them, achieving 86% recovery with no utility loss· Agentic AI
- Self-Learning AI Agent Pushes Automated Code Vulnerability Repair to 93% Success Rate· Agentic AI
- Lightweight Python toolkit shows simple AI algorithms can outperform complex frameworks on engineering optimization tasks· AI
- New Training Approach Boosts Speed of AI Draft Models by 23% Over Competing Methods· AI
- Data Formulator 0.7: AI agents meet enterprise analytics in a shared workspace· AI
- Google launches MCP server to accelerate Google Pay & Wallet API integrations· Technology
- AI systems extend human intelligence rather than replicate it, research shows· AI
- AI alignment hinges on how accurately systems capture user identity, study finds· AI
- Google Pay Unveils Agentic Commerce Tools and Android Parity Updates· Agentic AI
- AI agents that play test games reveal why code generation alone isn't enough· AI
- New Framework Debugs LLM Memory Systems by Tracing Where Errors Originate· AI
- New Guardrail Model Embeds Reasoning in Hidden States for Faster, Safer LLMs· AI
- New Method Harnesses AI to Auto-Generate Integration Tests for Java Code· AI
- New framework pinpoints where AI coding agents fail in complex repositories· Agentic AI
- How fact-checking benchmarks conflate missing evidence with insufficient evidence—and what to do about it· AI
- Open-Weight Language Models Refuse Harmful Requests Far Less Often in Somali Than English· AI
- New watermarking method shields text against paragraph-level paraphrasing attacks· AI Security
- How to Measure LLM Judge Reliability: A Framework for Consistent Evaluation Reporting· AI
- New Training Method Improves AI Essay Scoring Across Multiple Writing Dimensions· AI
- New empirical study reveals overlooked patterns in API authorization vulnerabilities· Security
- Study reveals tension between two faithfulness measures in AI reasoning explanations· AI
- New Language Makes AI Agents Accountable in Software Development· Agentic AI
- Advanced reasoning can backfire in clinical documentation systems· AI
- New Benchmark Reveals AI Coding Assistants Shift Personalities Over Long Sessions· Agentic AI
- Researchers Map C++ Safety Guidelines to Rust, Find Half Automatically Enforced· Software Engineering
- Alphabetical File Ordering in Code Review May Hurt Developer Efficiency· Software Engineering
- Palette Framework Lets Authorized Professionals Override AI Safety Guards On-Demand· AI
- RequestRouter accelerates single-GPU AI inference by 2x with dynamic request-level optimization· AI
- New Framework Exposes Security Gaps in AI Agents Operating with System Privileges· AI Security
- Microsoft Research Releases MagenticLite: An Agentic AI System Optimized for Small Models· Agentic AI
- Microsoft Research Introduces Vega: Fast, Private Credential Verification Without Revealing Identity· Technology
- Google Opens Gemini for Home as Full-Stack Platform for Device Makers and Carriers· AI
- Google Launches ADK for Kotlin and Android: Building AI Agents On-Device· Agentic AI
- Overlooked baselines undermine model interpretation methods, researchers show· Machine Learning
- New Framework Pinpoints Individual Log Anomalies Without Massive Labeling Effort· AI
- Study compares shrinking strategies across Haskell testing frameworks· Software Engineering
- New Legal RAG Benchmark Exposes Hallucination Risks in AI-Powered Law Search· AI
- New Benchmark Exposes How AI Coding Agents Game Test Suites Instead of Building Real Systems· AI
- New Compression Method Shrinks Language Model Embeddings Without Sacrificing Search Performance· AI
- Two-year RSE initiative uncovers hidden defects in quantum transport code· Software Engineering
- Google Transitions Gemini CLI Users to New Antigravity Platform· AI
- Google Rolls Out Gemini 3.5 and Agent-First Development Tools at I/O 2026· AI
- Library Drift: A Silent Failure Mode in Self-Evolving LLM Skill Systems· Agentic AI
- Framework for Building Reliable Production LLM Agents Through Architectural Pattern Design· AI
- Google DeepMind introduces Co-Scientist, a collaborative AI system designed to boost research productivity· AI
- AI Co-Scientist Accelerates Discovery of Liver Disease Treatment Pathways· AI
- Researchers develop contract-based verification tool for embedded automotive C code safety· Software Engineering
- AI Agent Accelerates GPU Kernel Optimization with Intelligent Search Strategy· Agentic AI
- New Framework Improves Code Quality Judging by Comparing Pairs Instead of Individual Scores· AI
- Google Cloud and NVIDIA Developer Community Hits 100,000 Members· Agentic AI
- First systematic study maps bugs in tile-based GPU programming frameworks· Software Engineering
- Researchers map internal reasoning pathways in AI models using entropy-gradient analysis· AI
- New framework automatically streamlines Lean proofs across library versions· AI & Formal Verification
- How Organizations Can Bridge EU AI Act Rules and Engineering Practice· AI Regulation
- Knowledge distillation bridges the gap for Nigerian language speech recognition· AI
- AI-assisted coding in science needs governance standards, not bans· AI
- New mutation metric brings domain knowledge to scientific software testing· Software Engineering
- ContraFix Uses Runtime Analysis to Guide Automated Software Vulnerability Fixes· AI
- Test-Driven Development Improves AI Code Generation for Full-Stack Web Apps by 15–24%· AI
- Multi-Agent Framework Tackles Legacy Code Modernization by Preserving Hidden Business Rules· AI
- New Deterministic Framework Simplifies Test Suites Without Randomness· Software Engineering
- Most Open-Source Projects Welcome AI-Assisted Code—But Demand Transparency and Human Review· Software Engineering
- New Dataset Provides Multimodal Analysis of Russian Government Speeches Across Decades· AI
- Microsoft Research: Long-Horizon AI Delegation Remains a Reliability Challenge· AI
- New Training Methods Make Block Attention Practical for Long-Context AI Models· AI
- Massive GitHub Study Reveals How Open-Source Projects Have Evolved Over a Decade· Software Engineering
- New Tool Speeds Up Conversion of Math Formulas to Computer Code Using AI· AI
- XSearch brings explainability to code search through concept-alignment· AI
- ML system ranks which duplicate BDD test steps deserve refactoring· Software Engineering
- Real-world agentic AI adoption lags capability, study finds· Agentic AI
- New AI Framework Generates Consistent Code Documentation Across Entire Repositories· Agentic AI
- First Large-Scale Study Exposes Fragmented Glob Pattern Ecosystem as Security Risk· Software Engineering
- New AI Framework Detects Hidden Memory Flaws in Compiled Software· Security
- New Approach Tackles Shortage of Psychological Defense Data with AI-Generated Examples· AI
- New Benchmark Tests AI's Ability to Detect Hidden Bugs Through Property-Based Testing· AI
- GridSFM: Microsoft's small foundation model speeds up power grid optimization by 100×· AI
- Flawed Test Passes Mask Ineffective AI Coding Agents, New Evaluation Framework Shows· Agentic AI
- PersonalAI 2.0 Boosts Knowledge Graph Search for AI Agents, Reducing Errors and Hallucinations· Agentic AI
- Large Code Models Struggle to Spot Security Vulnerabilities Without Extra Context· AI
- New Benchmark Reveals Performance Gap in LLM-Generated Systems Code· AI
- EVA-Bench: New Benchmarking Framework Reveals Significant Gaps in Voice Agent Performance· AI
- MatterSim accelerates materials discovery with 3-5x faster simulation and new multi-task model· AI
- LLM-powered platform lets engineers analyze power grids through natural conversation· AI
- New industrial code dataset with 832M lines enables software engineering AI research· AI
- Researchers Propose Security Framework for Multi-Agent AI Systems· Agentic AI
- Google DeepMind reimagines the mouse pointer as an AI-powered interface· AI
- DREAMS: New Tool Streamlines Design Research Model Creation and Maintenance· Software Engineering
- Researchers Find Hidden Gaps Between How Fine-Tuned AI Models Test and Perform in Real Use· AI Safety
- New Method Tackles Imbalanced Vulnerability Detection With Geometric Learning· AI