AI news on Thursday, June 18, 2026

SIGNIFICANCE

★ Top story · Research3d ago

Using AI to help physicians diagnose rare genetic diseases affecting children

Researchers successfully used an OpenAI reasoning model to identify rare genetic conditions in 18 pediatric patients who had previously remained undiagnosed. By analyzing complex medical records and clinical data, the system provided insights that helped doctors reach conclusions for these challenging cases. This demonstration suggests that advanced AI models may serve as effective diagnostic assistants for clinicians managing rare, difficult-to-detect disorders in children.

OOpenAI Blog CClinical Lab Products NNBC News Read original ↗

Thursday — June 18, 2026

Using AI to help physicians diagnose rare genetic diseases affecting children

Improving health intelligence in ChatGPT

Amazon in Talks to Sell Custom AI Chips in Bid to Cut Nvidia’s Dominance

Project Fetch: Phase two

Frontier Red Team

New usage analytics and updated spend controls for enterprises

Anthropic Lays Out Vision for How to Bolster AI Models’ Safety

Europe Bets Industrial AI Can Salvage Its Manufacturing Edge

How FERC’s Large-Load Interconnection Actions Help Address Grid Stress, Improve Affordability

Google Deepmind treats its own AI agents like rogue employees with office keys

US Acts to Accelerate Power Grid Hook-Ups for AI Data Centers

PM Modi meets Mistral AI CEO, co-founder Arthur Mensch in Paris

Adobe adds AI agents to Photoshop, Premiere, and more Creative Cloud apps

RODS: Reward-Driven Online Data Synthesis for Multi-Turn Tool-Use Agents

From Sparse Features to Trustworthy Proxies: Certifying SAE-Based Interpretability

Graph Grounded Cross Attention Transformer Neural Network for Structurally Constrained Full Event Sequence Generation in Predictive Process Monitoring

APT: Atomic Physical Transitions for Causal Video-Language Understanding

Possible or Definite? A Benchmark for Evaluating Diagnostic Uncertainty Preservation in Clinical Text

SAE Interventions are Unreliable: Post-Intervention Recovery of Suppressed Behavior

Sparsity Curse: Understanding RLVR Model Parameter Space from Model Merging

NAVI-Orbital: First In-Orbit Demonstration of a Zero-Shot Vision-Language Model for Autonomous Earth Observation

Do as the Romans Do: Learning Universal Behaviors from Heterogeneous Agents

From Bounding Boxes to Visual Reasoning: An On-Policy Data Annotation Tool for Vision-Language Models

CEO-Bench: Can Agents Play the Long Game?

Stealthy World Model Manipulation via Data Poisoning

Towards an Agent-First Web: Redesigning the Web for AI Agents

Bridging Creative Intent and Visual Quality: Creator-Driven Recurrent Video Generation with Agentic Feedback Loops

DRIFT: Refining Instruction Data via On-Policy Data Attribution

Gaussian Mixture Attention: Linear-Time Sequence Mixing via Probabilistic Latent Routing

Are LLMs Ready to Assist Physicians? PhysAssistBench for Interactive Doctor-Patient-EHR Assistance

PSyGenTAB: A Privacy-Preserving Framework for Synthetic Clinical Tabular Data Generation via Constrained Optimization

Taming I2V models for Image HOI Editing: A Cognitive Benchmark and Agentic Self-Correcting Framework

PACT: Preserving Anchored Cores in Task-vectors for Model Merging

Concept Modulation Models: A Unified Framework for Identifiability and Extrapolation

InTrain: Intrinsic Trainability for Zero-Cost Neural Architecture Search

DeFAb: A Verifiable Benchmark for Defeasible Abduction in Foundation Models

RegMix-D: Dynamic Data Mixing via Proxy Training Trajectories

BindEdit: Taming Attention Leakage for Precise Multi-Object Image Editing

Physics-IQ Verified

GateMem: Benchmarking Memory Governance in Multi-Principal Shared-Memory Agents

Zero-Shot Active Feature Acquisition via LLM-Elicitation

GrapNet: A Programmable Dynamic-Architecture Neural Graph Substrate

Geometric and Stochastic Analysis of Discontinuities in Sparse Mixture-of-Experts

VISUALSKILL: Multimodal Skills for Computer-Use Agents

Attribution-Guided and Coverage-Maximized Pruning for Structural MoE Compression

Data-Forcing Distillation: Restoring Diversity and Fidelity in Few-Step Video Generation

GraphPO: Graph-based Policy Optimization for Reasoning Models

Quantification of Uncertainty with Adversarial Models in Medical Image Segmentation

Robust and Interpretable Adaptation of Equivariant Materials Foundation Models via Sparsity-promoting Fine-tuning

Sumi: Open Uniform Diffusion Language Model from Scratch

DIPHINE: Diffusion-based $\Phi$-ID Neural Estimator

Strategic Feature Selection

Scaling Learning-based AEB with Massive Unlabeled Data

URDF Synthesis from RGB-D Sequences via Differentiable Joint Inference and Energy-Consistent Verification

Investigating Inductive Biases for Machine Learning Emulation of Sudden Stratospheric Warmings in Idealised Isca Simulations

Leadership as Coordination Control: Behavioral Signatures and the Recovery-Advantage Boundary in Multi-Agent LLM Teams

WorldLines: Benchmarking and Modeling Long-Horizon Stateful Embodied Agents

Learning from Your Own Mistakes: Constructing Learnable Micro-Reflective Trajectories for Self-Distillation

Visual-OPSD: Cross-Modal On-Policy Self-Distillation for Efficient Unified Multimodal Reasoning

Semantic Robustness Certification for Vision-Language Models

Attention as Frustrated Synchronization

Beyond Reward Engineering: A Data Recipe for Long-Context Reinforcement Learning

SAGE: Retain-Aware Post-Hoc Sanitization of Final Unlearning Vector

Effects of sparsity and superposition on loss in simple autoencoders

Forged Calamity: Benchmark for Cross-Domain Synthetic Disaster Detection in the Age of Diffusion

CaVe-VLM-CoT: An Interpretable Vision-Language Model Framework

What Does the Weight Norm Control in Grokking? Logit-Scale Mediation under Cross-Entropy

Motion-Focused Latent Action Enables Cross-Embodiment VLA Training from Human EgoVideos

SproutRAG: Attention-Guided Tree Search with Progressive Embeddings for Long-Document RAG

ProfiLLM: Utility-Aligned Agentic User Profiling for Industrial Ride-Hailing Dispatch

Microsoft Scout, New Enterprise Autopilot Built on OpenClaw, Announced at Build 2026

RedactionBench

Lost in a Single Vector: Improving Long-Document Retrieval with Chunk Evidence Aggregation

RTSGameBench: An RTS Benchmark for Strategic Reasoning by Vision-Language Models

Decoupling Search from Reasoning: A Vendor-Agnostic Grounding Architecture for LLM Agents

RouteJudge: An Open Platform for Reproducible and Preference-Aware LLM Routing

SenFlow: Inter-Sentence Flow Modeling for AI-Generated Text Detection in Hybrid Documents

Toward Training-Free Zero-Shot Anomaly Detection in 3D Medical Images: A Batch-Based Approach Using 2D Foundation Models

Learning from Own Solutions: Self-Conditioned Credit Assignment for Reinforcement Learning with Verifiable Rewards