September 2025 — page 2 of 2
- Withdrawn Paper Proposes Self-Knowledge Method to Filter Noisy Retrieved Content in RAG Systems· AI
- Machine Learning Model Designs Nanoporous Materials with Built-In Interpretability· AI
- New Method Reduces Gender Bias in LLM Decision-Making Systems· AI
- Japanese Riddles Expose Critical Reasoning Gap in AI Self-Evaluation· AI
- New AI Method Improves Binary Decompilation by Reconstructing Hidden Loop Structures· AI
- Researchers discover optimal transport method as effective audio spoofing attack on voice verification systems· AI Security
- Geometric Framework Uses Embedding Space to Spot and Fix AI Hallucinations· AI
- Silent Build Failures Masquerade as Success in Industrial CI Systems· Software Engineering
- Automated Testing Tool Helps Catch Silent Failures in Machine Learning Notebooks· AI
- New architecture enables AI models to seamlessly blend reasoning with precise numerical computation· Agentic AI
- GraphMend Tool Automatically Fixes PyTorch 2 Compilation Breaks, Delivering 5x Average Speedup· AI
- Withdrawn paper proposed integrating causal reasoning into AI retrieval systems· AI
- Researchers Test Whether LLMs Can Truly Explain Malware Behavior—and Find They Often Fall Short· AI
- New positional encoding method overcomes long-context limits of popular transformer technique· AI
- New benchmark reveals steering methods destabilize LLM safety across multiple dimensions· AI
- Researchers Expose Fairness Gaps in AI Toxicity Classifiers Using Mechanistic Interpretability· AI
- DoubleAgents: Aligning AI Assistants With Evolving User Needs in Team Workflows· Agentic AI
- Research maps the defining features and engineering blueprints of AI-native software systems· AI
- Researchers Develop Detection Methods to Catch Malicious Overfitting Attacks in Federated Learning· AI Security
- New Framework Cuts AI Model Overhead for Converting UI Designs into Code by Nearly Half· AI
- New study reveals how developers experience vibe coding with AI assistants· AI
- New ML method helps mobile robots quickly find optimal positions to grab objects· Robotics
- Transformer-based spatial transformers bring flexibility and stability to rotation-invariant image analysis· AI
- Diffusion LLMs Show Promise as Alternative to Autoregressive Code Generation· AI
- First Comprehensive Survey Maps Privacy Risks in Recommender System Attacks· AI Security
- New analytic method sidesteps backpropagation for continual learning, cuts energy use by 20×· AI
- Enterprise AI expectations shifted from team collaboration to individual assistants between 2023 and 2025· AI
- New distillation method lets smaller AI models match capabilities of much larger ones· Agentic AI
- New Method Adapts Speech Recognition Models to Specialized Language Using Text Alone· AI
- ReFine: A Two-Part Framework for Generating Synthetic Tabular Data from Limited Records· Machine Learning
- Offline RL learns to schedule jobs from imperfect data—and random solutions outperform expert heuristics· AI
- Standard Personality Tests Fail to Predict How LLMs Actually Behave With Users· AI
- Study reveals how developers use LLMs for coding—and where AI-generated code stumbles· AI
- Study reveals widespread license violations as AI models move from Hugging Face to GitHub· AI
- AI Framework Cuts Brain Imaging Scan Time in Half While Maintaining Quality· Medical AI
- Shopee's Compass-v3 AI Model Masters Multilingual E-Commerce Across Southeast Asia· AI
- Medical AI models retain troubling amounts of training data, study finds· AI
- New Model Combines Recurrence and Transformers to Search Across Mixed-Media Collections· AI
- AI-Guided Quantum Network Routing Handles Real-World Interference and Uncertainty· Quantum Computing
- New benchmark dataset helps AI extract structured knowledge from full NLP research papers· AI
- New Benchmark Reveals How LLMs Fail to Align with Values Across Different Cultures and Regions· AI
- New Foundation Model Boosts Healthcare Predictions by Learning from Age and Sex· AI
- Researchers Build Framework for Translating Literary Works into Romanian Using Open-Source Models· AI
- OpenCoderRank brings AI-tailored coding exams to resource-limited settings· AI
- Model-Based Training Cuts Robot Learning Time by 12% Without Sacrificing Performance· Robotics
- Reconstruction Alignment Boosts Unified Multimodal Models with Self-Supervised Learning· AI
- Lightweight AI models show promise for private, on-device health monitoring· AI
- Extra Computation at Test-Time Doesn't Boost Knowledge Tasks—Yet· AI
- LLM Study Traces 150 Years of German Migration Rhetoric, From Postwar Inclusion to Recent Backlash· AI
- LLMs Generate Realistic Travel Diaries Using Census Data, Matching Classical Models· AI
- Training Randomness in AI Brain Imaging Models Reveals Hidden Uncertainty—and a Data-Boost Opportunity· AI
- Magnetic microrobots navigate complex biological environments autonomously· Robotics
- New Benchmark Exposes How AI Models Lose Information When Converting Between Images and Text· AI
- RL-Powered Flight Planning Cuts Route Optimization Time in Half· AI
- New AI Method Detects Bearing Faults in Machinery Using Real-Time Vibration Analysis· AI
- FedProx Outperforms Other Federated Learning Strategies for Hospital Mortality Prediction· Machine Learning
- Language models struggle with simplicity: Study reveals LLMs fail at Occam's Razor reasoning· AI
- New method trains AI agents to learn from both successes and failures without human-designed rewards· AI
- Iterative Feedback Strengthens LLM Diagram Generation from Natural Language· AI
- Neural networks enable modular robots to reconfigure using only local neighbor data· Agentic AI
- New fingerprinting method helps protect large language models from unauthorized copying· AI
- LMEnt Toolkit Opens Black Box of How Language Models Learn World Knowledge· AI
- Steering Vectors Reduce AI Evaluators' Bias Toward Their Own Outputs· AI
- Framework Uses Structured Ontologies to Build Better Geometry Lesson Plans· Education Technology
- Researchers Develop Framework to Measure How Well AI Annotations Match Expert Judgment· AI
- Large Language Models Struggle to Track Team Understanding in Conversation· AI
- Most Video QA Benchmarks Don't Actually Need Temporal Grounding, Study Shows· AI
- New Framework Uses Logic and Counterfactuals to Formally Verify AI System Explainability· AI
- Vis-CoT: Interactive Framework for Human-in-the-Loop LLM Reasoning (Withdrawn)· AI