April 2026 — page 2 of 2
- New benchmark tests AI's ability to generate software architectures from requirements· AI
- Reasoning Chains Can Mislead LLM Judges Into Accepting Wrong Answers· AI
- New Benchmark Reveals Reward Models Struggle to Align with Individual User Preferences· AI
- New Benchmark Reveals LLMs Struggle With End-to-End Software Generation· AI
- How to Actually Pick the Best Open Reasoning Model for Your Use Case· AI
- Persona steering can degrade AI essay quality and bias automated grading, study finds· AI
- New Benchmark Reveals Critical Weaknesses in AI Agent Safety Systems· AI Safety
- Multi-Agent System Translates Code Across Programming Languages at Scale· AI
- AI Diagnostic Agent Learns from Experience Like Clinicians, Boosting Accuracy 92%· Agentic AI
- Fine-Grained Confidence Scoring Improves How LLMs Handle Code Repair Tasks· AI
- New Framework Shows Why Complex Software Can Still Run Predictably· Software Engineering
- LLMs Favor Their Own Output When Judging Other Models, Study Finds· AI Evaluation
- LLMs learn to interrupt mid-generation to cut multi-agent chatter in half· Agentic AI
- Vision-Language Models Break Their Own Rules, Unlike Humans· AI
- What Users Really Want from AI Models: Study Reveals Top Pain Points on Hugging Face· AI
- AI System Synthesizes Software Specifications Using Reinforcement Learning from Test Failures· AI
- Transformers Learn Analogical Reasoning Through Intermediate Copy Tasks· AI
- New Framework Improves AI Code Generation by Testing and Fixing Programs Iteratively· AI
- Small amounts of speech data dramatically improve LLM-based speech recognition in new domains· AI
- Lightweight Risk-Signal Framework Brings Mathematical Forecasting to Spreadsheets· Software Engineering
- New taxonomy maps 65 root causes of software energy waste to tackle computing's environmental toll· Software Engineering
- MIRAGE: LLMs Simulate Microservice Dependencies in Real-Time for Better Testing· AI
- EffiPair: Using Performance Comparison to Guide LLM Code Optimization· AI
- New approach locks LLM outputs as deterministic code to boost reliability in enterprise workflows· Agentic AI
- LLMs Show Promise on Numerical Stability, but Lag Behind Traditional Tools in Scientific Code· AI
- New AI Method Tackles Medical Misdiagnosis by Ruling Out Plausible Alternatives· AI
- Researchers Build LLM-Powered Retail Simulator to Test Sales Strategies· AI
- New Method Cuts Reasoning AI Token Usage in Half by Stopping When Confidence Peaks· AI
- New Algorithm Creates Lean Model Portfolios to Balance Competing AI Safety Goals· AI
- Uncertainty-Aware Planning Improves Conversational AI's Ability to Meet Goals in Fewer Turns· Agentic AI
- Combee Accelerates Self-Improving AI Agents with Parallel Prompt Learning· Agentic AI
- New LLM-Based Tool Detects Hidden Ad Regulation Violations in Mobile Apps· AI
- Machine learning models predict chained vulnerabilities in software supply chains· Security
- New distillation method trains compact language models 10x faster without sacrificing quality· AI
- How Knowing Your AI Peer's Flaws Can Improve Group Decision-Making· AI
- Study reveals LLMs struggle to detect flaws when code changes but documentation stays the same· AI
- New System Guides AI to Fix Code Bugs by Observing Program Behavior in Action· AI
- Study Questions Whether Tool Use Actually Improves Web Agent Performance· AI
- Simple Vocabulary Masking Technique Restores Diversity in Self-Teaching AI Models· AI
- Large language models might engage in associative thinking, researchers argue· AI
- Researchers Propose Smart Router to Select Best AI Model for Each Prompt's Diversity· AI
- How AI Coding Agents Reshape Developer Documentation—and Break Traditional Analytics· AI
- New Benchmark Exposes Critical Gaps in Binary Code Similarity Detection· Research
- Automated Production Benchmarks Bridge Gap Between Lab and Real-World AI Coding Agents· AI
- LLMs capture social reasoning patterns but struggle with magnitude accuracy· AI
- AI Agent Automates Polymer Simulations, Achieving Mixed Success on Predicting Material Properties· Agentic AI
- New benchmark reveals implicit biases in LLMs remain 6x higher than explicit biases· AI
- Researchers Expose Test Suite Gaps in Major Bug-Repair Benchmark· AI & Software Engineering
- Multi-LLM revision gains depend on task type, not just model strength· AI
- Study reveals how developers converse with AI coding assistants in real-world projects· AI
- New framework uses targeted questions to resolve ambiguity in AI-generated cloud infrastructure code· AI
- New Method Cuts LLM Benchmark Costs by 85% While Predicting Model Performance· AI
- AI Models Learn to Cheat: Researchers Develop New Defense Against Reward Hacking· AI Alignment
- New LLM System Automatically Improves Android Bug Reports with Missing Details· AI
- Fine-grained citations may harm AI model performance, research finds· AI
- New Framework Tells LLMs When to Stop Multi-Turn Reasoning Tasks· AI