AI news this week: 45 source-linked stories
A dated list of 45 AI stories with links to their original sources.
Stories published this week, linked to their original sources. The underlying claims have not been independently verified.
Beyond Surface Style: Aligning Multi-Turn User Simulators with Behavioral Consistency
Faithful user simulation is fundamental to building, evaluating, and improving interactive AI at scale.
PAWS: Policy-driven Agentic World Simulation
Policy interventions propagate through public communication, institutional decisions, and stakeholder responses, yet datasets for financial multi-agent simulation rarely connect these processes to temporally aligned historical evidence.
Training Object Permanence in World Models
Object permanence and solidity are hallmarks of human cognitive priors.
TW3Cast: A Frozen Router of Lightly Fine-Tuned Foundation Models for Time-Series Forecasting on GIFT-Eval, Selected Entirely on the Training Split
TW3Cast is a
BaseCamp — An Agentic AI Framework for Automating DNA Sequencing Data Pipelines
DNA sequencing
When Should Forecasting Agents Reason? Behavioral Stress Tests for Reliability Routing
Forecasting agents increasingly combine language-model reasoning, retrieval, ensembling, and calibration, but it remains unclear when each behavior should be trusted.
Adversarial Closed-Loop Curriculum for Evolving Role-Playing Agents
Role-playing agents based on large language models have been widely applied in areas such as personalized assistance and social simulation.
TWIST: A Proposed Benchmark for Intervention Quality in Conversational Memory, with a Human-Validated Draft-Alignment
TWIST is
Driving Epidemic Models with AI Agents: the Epydemix Agent Framework
Artificial Intelligence agents based on large language models provide convenient natural language interfaces to scientific software, but reliability is not automatic.
Pistis Technical Report
We introduce the Pistis model family, comprising 27B- and 9B-parameter multimodal large language models built on Qwen3.6 and Qwen3.5, respectively, and developed through a general and scalable post-training framework.
Reinforcement Learning with Decomposed Subtasks
Reinforcement
Escaping Python Dependency Hell: A Hybrid Replay-and-Repair Pipeline for Python Dependency Resolution
Dependency conflicts in Python ecosystems arise from incompatible version constraints, missing packages, and undocumented compatibility relationships, causing many real-world code snippets to fail at execution.
Training Intelligent Voice Assistant Wakeup with Controllable Synthetic Conversations
This paper introduces a novel
TwinCheck: Evidence-Grounded Negative-Twin Verification for Stateful Tool Agents
TwinCheck
Harness as a Language: A Minimalist Agent Framework With Maximal Expressivity
JAZ
Policy-as-Skill: Governed LLM Decision Support with Evidence, Deterministic Control, and Audit
Policy-
Which Objectives Need a Dial? Predicting Objective Conflict and Covering Trade-offs in Steerable Pluralistic Alignment
Multi-Objective Direct Preference
Building Socio-Affective Artificial Intelligence for Interactive Multi-Agent Simulations
The objective of this article is to provide design principles and a software architecture for enabling interaction between humans and multiple agents in simulated dynamic worlds.
Math Reasoning in LLMs is Organized by Approach, Not Topic
Mathematical reasoning benchmarks are typically organized by topic, but language models may organize their internal computation by reusable reasoning approach instead.
Are Stated Reasoning Steps Causally Load-Bearing?
Chain-of-thought (CoT) monitoring assumes that the reasoning a model writes reflects the computation that directly produces its answer.
Propose, Don’t Judge: An Anytime-Valid Referee for LLM Agents That Mine Investment Factors
Language-model
Silent Failures in Agent-Tool Interaction: An Audit of ToolUniverse
Agentic AI systems are increasingly adopting automated pipelines that integrate multiple tools.
Goal-driven Variant Categorization
Process discovery rarely yields a single coherent process structure.
MAWILE: Multi-Axis Workbench for Inspecting LLM Evaluators
Large language model (LLM) judges provide a flexible and scalable method for evaluating model and agent outputs, but their verdicts can be sensitive to incidental changes in the evaluated response, judge instructions, and scoring rubric.
Social Influence and the Allocation of Scientific Attention in AI Populations
AI systems are becoming participants in the evaluation and use of scientific research.
Didactic knowledge or Clinical Cases? How Data Types Shape Medical Large Language Models
Medical large language models are commonly trained on mixtures of didactic data (e.g., textbooks) and clinical data (e.g., patient records), yet how these data types differentially shape model capabilities remains unclear.
The Wisdom of Artificial Deliberative Crowds
The aggregation of many lay estimates often outperforms individual expert judgment, a phenomenon known as the wisdom of crowds.
EvidenT: Building Trustworthy Enterprise Assistants through Evidence Groundedness and Traceability
Enterprise AI assistants must produce responses that are verifiable and traceable to source evidence.
Grab and OpenAI bring practical AI skills to Southeast Asia
OpenAI and Grab launch GO Forward with AI, a regional programme helping 30,000 partners build practical AI skills across Southeast Asia.
At AI Day Singapore, NVIDIA and Partners Showcase AI Advancements Across Southeast Asia
NVIDIA AI
IntLawNER: A Named Entity Recognition Dataset and Benchmark in International Law
International law provides the normative framework through which states coordinate action, regulate armed conflict, and protect human rights, yet its texts remain without token-level named entity recognition (NER) resources.
AutoGym: Blueprint-First Generation of Verifiable Agent Gyms
Researchers introduce AutoGym, a framework for generating agent-training tasks with executable environments and verifiers.
Replication Without Persistence in Hosted LLMs: Measurement Sensitivity in Action-Time Belief Evaluation
Behavioural evaluations of hosted language models can vary because the evaluated service, the measurement instrument, or both differ across runs.
An Affordable AI-Integrated Smart Cane for Multimodal Mobility Assistance of Visually Impaired Users
Visual impairment affects over 2.2 billion people worldwide, yet conventional white canes cannot detect elevated hazards or provide semantic environmental context.
Agreement Overstates Evidence: Error Dependence in LLM Judge Consensus
Researchers report correlated errors among LLM judges, so agreement between judges may overstate the evidence.
PAANI : On Device Visual Evidence Fusion and Explainable Guidance for River Robot Simulation
Mobile river monitoring robots must interpret obstacles and water boundaries that geographic waypoints alone cannot describe.
Better prompt caching for GPT-6
OpenAI says GPT-6 prompt caching improves cache reuse and adds diagnostics and explicit cache breakpoints.
Anthropic introduces Claude Opus 5.5
Anthropic announced Claude Opus 5.5 on September 22. It says typical workloads cost 40% less than Opus 5; the claim comes from Anthropic’s own tests.
NVIDIA Isaac ROS 5.0 Advances Agentic, Open Source Robotics Development
NVIDIA Isaac ROS
OpenAI introduces GPT-6 Sol and GPT-6 Luna
OpenAI announced GPT-6 Sol and Luna on September 22. The company says both are available through its API and Codex, with different capability and cost profiles.
5 Companies Using NVIDIA AI for Clean Energy
NVIDIA is highlighting
AI Security Is an Engineering Problem — How to Solve It at Every Layer of the Agent Stack
AI security is an
From Enablement to Execution, Egypt’s AI Ecosystem Reaches Production Scale
Egypt’s AI builders
NVIDIA Launches DSX Ready to Qualify Power and Cooling Products for AI Factories
Every AI factory needs power and cooling that fit its computing architecture.
Why Deploying Physical AI at Scale Demands Safety at Every Layer
Physical AI is moving rapidly from research to large-scale deployment.