February 2026 — page 1 of 2
- New Benchmark Reveals LLM Agents Struggle with Evidence-Based Diagnosis in Cloud Outages· Agentic AI
- AI Framework Flags When Project READMEs Need Updating After Code Changes· AI
- QQ: Open-Source Language Metadata Toolkit Simplifies Multilingual NLP Research· AI
- LLMs Don't Need Their Own Past Responses to Stay Accurate· AI
- New Test Reveals Why AI Struggles at Multi-Turn Teamwork· AI
- New AI System Generates Mobile App Test Cases by Extracting Business Logic· AI
- New Framework Bridges Gap Between AI Lab Performance and Real-World Impact· AI
- Researchers Publish Formal Security Framework for AI Agent Skills in Supply Chains· Agentic AI
- New Framework Automates Jailbreak Testing, Keeping LLM Safety Benchmarks Current· Agentic AI
- New Method Boosts Privacy-Preserving AI Training While Slashing Costs· AI
- New training method cuts learning variance in AI reasoning tasks by up to 80%· AI
- SkillNet: A New Infrastructure for Building and Sharing Reusable AI Agent Capabilities· Agentic AI
- Researchers retract claims of retrieval breakthrough in test-time training method after finding evaluation flaws· AI
- Study Reveals Critical Gaps in Popular AI Agent Frameworks· Agentic AI
- New AI System Automates Discovery of Hidden Dependencies Between Software Requirements· AI
- SpecMind: LLMs Learn to Refine Code Specifications Through Iterative Dialogue· AI
- New Training Method Tackles the Faithfulness Problem in AI Reasoning Explanations· AI
- New Training Method Lets AI Music Models Learn Whole Compositions at Once· AI
- Researchers Release Massive Video Reasoning Dataset with 1 Million Clips· AI
- New Framework Distinguishes Software Bugs from Hardware Noise in Quantum Processors· Quantum Computing
- Study finds test-time scaling fails to improve LLM agents in realistic deployments· AI
- Researchers Propose Unified Formal Model to Standardize How Package Managers Resolve Dependencies· Software Engineering
- New benchmark tackles design-to-code translation for e-commerce UI workflows· AI
- New benchmark exposes gaps in AI theorem provers when tackling real-world software verification· AI
- Runtime Data Boosts LLM Accuracy in Predicting Jupyter Notebook Crashes· AI
- New dataset reveals how humans intervene in AI web agents—and how to predict it· Agentic AI
- New Method Predicts Which Open-Source Projects Will Fail Due to Competition· Software Engineering
- Researchers decode how LLMs internally represent cognitive complexity across learning levels· AI
- New Tucker Decomposition Method Cuts Fine-Tuning Parameters by Hundreds While Matching LoRA Performance· Machine Learning
- HIPE-2026 Benchmark Targets Accurate Person-Place Relation Extraction in Historical Texts· AI
- New Execution Framework Improves LLM Performance in Scientific Research While Slashing Costs· Agentic AI
- New Test Method Reveals LLMs Struggle to Capture Actual Code Behavior· AI
- New framework uses hyperbolic geometry to improve multimodal emotion recognition· AI
- New Framework Lets AI Systems Selectively Delete Sensitive Data Modalities On Demand· AI
- New Benchmark Exposes Memory Gaps in AI Agents Working Across Multiple Sessions· Agentic AI
- Google DeepMind launches major AI partnerships across India's science, education, and energy sectors· AI
- Google DeepMind Launches Lyria 3: AI Music Generation Now Available in Gemini· AI
- Large language models show promise but stumble on pinpointing malicious Python code· AI
- New framework merges neural networks with symbolic reasoning for knowledge graph queries· AI
- New technique enables efficient communication between AI models with different architectures· Agentic AI
- New AI System Guides Entrepreneurs Through Iterative Business Planning With Reflection Prompts· AI
- New Method Detects Silent Errors in LLM-Generated Optimization Code· AI
- New AI Method Segments Documents by Predicting What Readers Actually Need· AI
- How Developers Configure AI Coding Agents: New Study Maps the Landscape· Agentic AI
- Roundtrip Translation Enables Text Style Transfer Without Parallel Training Data· AI
- Linear RNNs Outperform Transformers at Tracking State Changes in Code· AI
- Researchers Propose DFT-Based Method to Tame LLM Hallucinations Through Semantic Hierarchy Structures· AI
- New Layer Pruning Method Accelerates Large Language Model Deployment by 4x· AI
- Study reveals how AI models fake reasoning when given the answer first· AI
- Language statistics reveal why AI models map concepts into geometric shapes· Machine Learning
- New Dataset Brings Multimodal Peer Review Analysis Beyond Computer Science· AI
- New Framework Structures Reasoning Patterns to Improve AI Decision-Making and Transparency· AI
- New Framework Lets Non-Programmers Turn Requirement Documents Into Working Software· AI
- New Speculative Decoding Method Uses Smart Vocabulary Selection to Speed Up AI Text Generation· AI
- New Framework Distinguishes Model Confidence From Actual Capability in Large Language Models· AI
- New AI Training Method Embeds Privacy Safeguards Directly Into Language Model Agents· AI
- StackingNet enables AI models to collaborate without sharing internal data· Agentic AI
- New testing tool detects hidden bugs across mobile app features· Software Engineering
- New study finds LLMs' grammar skills drive brain-activity predictions stronger in left hemisphere· AI
- Researchers Propose Reference-Free Quality Metrics for Converting Flowchart Images into Code· AI
- AI speech screening tools falsely flag multilingual speakers as cognitively impaired at 2.5x the rate of monolingual users· AI
- New framework improves how LLMs judge text quality while catching errors· AI
- Large language models struggle to preserve urgency in crisis translation across languages· AI
- Repository Context Files Don't Boost Coding Agent Performance, Study Finds· AI
- Researchers Propose Treating Code Structurally, Not Linguistically, to Boost Multi-File AI Code Generation· AI & Software Engineering
- LLM Committees Can Improve Code Safety by Rejecting Flawed Queries· Vibecoding
- New Benchmark Reveals Multi-Agent AI Safety Gaps in Game-Theoretic Scenarios· AI Safety
- Recycling LoRAs with Adaptive Merging Shows Limited Real-World Benefit· AI
- How LLM Feedback Should Match the Learner: Study Reveals Personality Shapes Educational Response· AI
- New Framework Tests Healthcare AI Chatbots for Safety Across Patient Differences· AI
- Learned attention gates struggle against random routing when trained jointly—study reveals why· Machine Learning
- New Research Maps How Experienced Developers Actually Debug Code in the Real World· Software Engineering
- Reinforcement Learning–Trained Rewriting Policy Reduces Catastrophic Forgetting in Fine-Tuned Language Models· AI
- New Benchmark Tests How Well AI Agents Organize Knowledge, Not Just Recall It· AI
- New benchmark tests how well AI agents can replicate real-world research findings· AI
- Google DeepMind's Gemini Deep Think Advances AI-Assisted Mathematics and Physics Research· AI
- New Benchmark Exposes Gaps in AI Agents' Ability to Build Games· Agentic AI
- New Framework Reveals Gaps Between LLM Values and Actual Advisory Behavior· AI
- New Model Enables Real-Time Speech Translation Without Language-Specific Alignment Data· AI
- Hybrid Neuro-Symbolic Framework Tackles LLM Hallucinations in World Models· AI
- New benchmark tests whether neural language models need innate linguistic rules· AI
- New lightweight translation models bring machine translation efficiency to African languages· AI
- Energy Sector Workers Spot High-Value Generative AI Opportunities in Operations and Analysis· AI
- New Benchmark Tests Whether AI Can Genuinely Discover Unknown Biology· AI
- AI Models Signal Their Own Failures Before Generation—and That Can Cut Computing Costs· AI
- RL Framework Learns Smarter Token Eviction to Cut LLM Inference Memory Costs· AI
- New Plugin Helps Robot Laboratories Handle Complex Multi-Step Scientific Tasks· Agentic AI
- New training method helps AI models reason in hidden thought spaces while avoiding feature collapse· AI
- Vision Language Models Fall Short in Assessing Student Mental Models from Multimodal Answers· AI
- New AI agent Artisan automatically reproduces research results through code generation· AI
- New Method Detects and Fixes Retrieval Blind Spots in AI Search Systems· AI
- New benchmark unlocks AI-powered code repair for HarmonyOS's ArkTS language· AI
- Computational neuroscience bridges language theory and brain mechanisms· Computational Neuroscience
- Researchers Develop DeAction to Detect and Correct Misaligned Actions in AI Agents· Agentic AI
- New benchmark tests AI agents' ability to curate biology knowledge bases from scientific papers· Agentic AI
- Deep Learning Model Accelerates Light-Scattering Predictions for Advanced Nanomaterials· AI
- New AI model learns to provide emotional support through dialogue strategy optimization· AI
- Language Models Hide Knowledge of Falsehoods in Low-Dimensional Geometric Structures· AI
- New KV Cache Compression Method Cuts Long-Context Inference Costs by 85%· AI
- Adversarial AI Agents Boost Automated Test Generation for Harder-to-Find Bugs· AI