March 2026 — page 1 of 2
- AI models secretly know when they're wrong—but can't admit it· AI
- Simple Terminal Agents Outperform Complex AI Systems for Business Automation· Agentic AI
- OpenTools: Community-Built Framework Tackles the Blind Spot in AI Agent Reliability· AI
- New Framework Generates Better Debug Logs by Learning from Program Execution· AI
- New tool helps people critically examine their AI therapy conversations· AI
- LLM Agents Dramatically Improve Bug Detection in Software Commits· Software Engineering
- New framework teaches AI radiology models when to express doubt about findings· AI
- New Benchmark Exposes How Well LLMs Actually Understand User Interests· AI
- Framework reveals many AI-generated code specifications are behaviorally weak despite passing verification· AI
- New methodology addresses AI coding's hidden verification gap· AI & Software Development
- AI Model Trained with Reinforcement Learning Improves Drug Poisoning Diagnosis in Emergencies· AI
- New research challenges AI language models' bias against dialects and minority languages· AI
- Research shows interconnecting requirements boosts software prioritization from user feedback· Software Engineering
- New framework detects AI hallucinations in retrieval systems with fine-grained claim verification· AI
- Study reveals why malware detectors fail in the NPM ecosystem and what works best· Security
- Large-scale study reveals AI-generated code looks more like human work than lab tests suggested· AI
- Research shows understanding author intent improves AI-generated scientific reports· AI
- New AI-Powered Tool Catches Memory Leaks Standard Code Analyzers Miss· Software Engineering
- Study: AI-Generated Code Poses Growing Burden on Developers and Teams· AI
- Researchers Develop Framework to Hunt Malicious AI Agent Skills Across Public Registries· Agentic AI
- Etna brings rigorous testing to property-based evaluation across multiple languages· Software Engineering
- New Benchmark Tests AI Agents on Website Building Tasks· AI
- Perplexity Misleads: New Distillation Method Shows Log-Likelihood Hides Real Generation Quality Gaps· AI
- Open Source Projects Adopt Varied Strategies to Manage AI-Generated Contributions· AI
- New benchmark reveals why code agents struggle with real-world feature implementation· AI
- New AI System Neutralizes Speaker Accents in Real-Time While Preserving Voice Identity· AI
- Study reveals exploit effectiveness across Java library versions, discovers 796 missing vulnerabilities· Security
- New AI system detects millions in smart contract vulnerabilities before they're exploited· Agentic AI
- New method detects near-verbatim memorization in AI language models with far less computation· AI
- TheBotCompany: LLM-powered agent teams tackle long-term software development· Agentic AI
- Encoder quality, not decoder complexity, drives code-error classification performance· AI
- New Framework Tackles Root Cause Analysis in Microservices When Data Goes Missing· AI
- Study Reveals LLMs Struggle to Gauge Their Own Accuracy More Than Previously Thought· AI
- New Method Optimizes Multilingual Speech Translation by Analyzing Training Patterns· AI
- New framework improves machine translation by learning when context helps· AI
- How AI Models Can Improve Themselves Without Human Input· AI
- Hidden bias in depression detection: When interviewer scripts fool AI models· AI
- Interpretable AI System Combines YOLO Object Detection with Kolmogorov-Arnold Networks for Trustworthy Vision· AI
- New AI System Uses Runtime Tracing to Fix Complex Software Bugs More Efficiently· AI
- New Framework Treats Brain Signals as Semantic Sketches Rather Than Full Sentences· AI
- New Framework Bridges Cryptographic Inventory and Architecture Design for Secure Migration· Security
- LLM Agent Automates Proof Discovery for Complex Program Verification· Agentic AI
- New Framework Separates Speaker Identity from Deepfake Detection in Synthetic Speech· AI
- New Tool Detects Hidden Code Clones Across 2 Million AI Agent Skills· AI
- New Distillation Technique Speeds Up Language Models Without Sacrificing Quality· AI
- Multi-agent coordination framework boosts AI software engineering accuracy by up to 26%· Agentic AI
- New Method Maps LLM Weaknesses by Capturing Prompt Specificity, Not Just Similarity· AI
- Researchers Expose Finetuning Loophole: LLMs Can Regurgitate Entire Copyrighted Books· AI
- New AI Framework Learns from Failures to Self-Improve at Research Coding· Agentic AI
- New routing method uses LLM internal activations to predict which model handles tasks best· AI
- Study reveals AI models struggle to predict how chatbots change human beliefs· AI Safety
- New Method Speeds Up AI Interpretability Analysis with Near-Instant Attribution· AI
- AI models spot unreliable speakers but fail to act on the insight· AI
- New Framework Helps AI Systems Debug Complex Compiler Bugs More Effectively· Agentic AI
- New benchmark tests AI assistants on messy, real-world tasks with stateful environments· AI
- Research Uncovers Blind Spot in Python Security: Vendored Libraries Evade Detection· Security
- Why AI Text Gets Caught: Decoding Methods Block Human-Like Word Choices· AI
- Better cross-lingual alignment doesn't guarantee stronger transfer performance, researchers find· AI
- LLM-based code review tools can be tricked into missing security flaws via metadata manipulation· AI Security
- New Method Cuts Memory Use for Long-Context LLM Inference Without Sacrificing Quality· AI
- New Method Strengthens Safety of Multimodal AI Models in Extended Conversations· AI
- New AI Method Breaks Through Code Verification Bottleneck with Hierarchical Proof Search· AI
- New Benchmark Exposes How Far LLMs Fall Short in Formal Rust Verification· AI
- Researchers develop metrics to detect how LLMs drift in multi-turn conversations· AI
- New Framework Reveals Major Gaps in AI Agent Safety Testing· AI Safety
- Researchers Expose Hidden Pedagogical Risks in AI Tutoring Systems· AI
- Study reveals how LLMs secretly compute confidence before users ask· AI
- Systematic Review Maps Large Language Models' Role in Software Log Analysis· AI
- New framework reduces LLM fine-tuning costs by half using knowledge-aware active learning· AI
- Reframing AI Prompts as Software Requirements: A Research Agenda· AI
- AI Agents Transform Social Platforms With Unequal Participation and Flattened Discourse· AI
- New Method Uses Prompt Programming to Reduce Cultural Bias in AI Language Models· AI
- Holotron-12B: Fast Multimodal Agent for Production Computer Use· Agentic AI
- New framework detects poisoned code-generation AI models with 97% accuracy· AI Security
- New technique helps retrieval systems learn faster by selectively pruning training data· AI
- Worm attack spreads across LLM agent networks—autonomous agents vulnerable to self-propagating malware· Agentic AI
- AI coaching helps people express empathy more effectively, study finds· AI
- New Method Trains LLMs to Master Private Code Libraries Through Synthetic Data· AI
- How Scientific Software Developers Leave Debt—and Why It Lingers· Software Engineering
- New AI Model Teaches Robots to Reason Through Task Progress Like Human Supervisors· Robotics
- New Benchmark Tests AI Coding Agents on Real-World Optimization at Repository Scale· Agentic AI
- Researchers compare four formalisms for programming robot missions· Robotics
- Study Reveals How LLMs Generate Plausible Wrong Answers in Educational Quizzes· AI
- Researchers Map Fragmented Engineering Datasets to Speed Design Research· AI
- Study Maps the Value and Controversy of ISTQB Software Testing Certifications· Software Engineering
- New benchmark tests how well AI language models learn rare programming languages· AI
- Dialogue AI Systems Need Better Evaluation Beyond Surface-Level Metrics, Study Shows· AI
- New Training Method Lets AI Systems Learn to Judge and Generate High-Impact Research Ideas· AI
- New Benchmark Tests Audio-Language Models on Persian Challenges· AI
- LLM-Powered Text Annotation Amplifies Racial Biases Across Thousands of Tasks· AI
- Study reveals why LLMs fail to select correct answers they can internally represent· AI
- Study Maps How Large Audio Language Models Fuse Sound and Text Information· AI
- New Method Detects and Fixes Bias in AI Text Generation On-the-Fly· AI
- Study Reveals How Hardware Imperfections Degrade Large Language Model Reasoning on Memristor Chips· AI
- Study quantifies how deep learning model reuse is reshaping scientific research· AI
- New Method Improves AI's Ability to Generate CAD Code by Selecting Better Examples· AI
- Why AI Struggles With Fiction Despite Training on Books: The Narrative Generation Paradox· AI
- New benchmark reveals AI coding agents struggle with long-term software maintenance· Agentic AI
- New benchmark reveals embedding models struggle with long-term memory retrieval tasks· AI
- Translationese Explained as Cognitive Response to Task Difficulty· Language & AI