simulation

A counterfactual simulator for marketing decisions combining structural causal models, Hawkes processes, and LLM agents; allows querying 'what-if' scenarios before committing budget. Open source on GitHub.

Microsoft Research announces MatterSim updates including MatterSim-MT, a multi-task foundation model for materials characterization, faster simulation (3-5x speedup), and experimental validation of thermal conductivity predictions for a new material.

This paper introduces WLDS, a large-model-driven system for simulating and deducing emergency instances by leveraging controllable randomness and cross-domain knowledge. It presents the Emergency Instances Deduction (EID) benchmark and demonstrates high-fidelity simulation capabilities across multiple domains.

This paper introduces a simulation framework for federated analysis of Multiple Sclerosis brain lesions, combining image segmentation with clinical data analysis to test federated learning methods while preserving patient privacy.

This paper introduces LychSim, a controllable simulation framework built on Unreal Engine 5 to facilitate vision research, synthetic data generation, and agentic LLM evaluation via MCP integration.

This paper introduces the DecodingTrust-Agent Platform (DTap), a controllable and interactive red-teaming platform for evaluating AI agent security across multiple domains. It also presents DTap-Red, an autonomous agent for discovering attack strategies, and DTap-Bench, a large-scale dataset for risk assessment.

KinDER is a new open-source benchmark for physical reasoning in robotics, featuring procedurally generated environments and baselines to evaluate kinematic and dynamic constraint challenges.

This comprehensive survey reviews the literature on world models for robot learning, covering their roles in policy learning, planning, and simulation. It highlights key paradigms, benchmarks, and future directions for predictive modeling in embodied agents.

Manufacturers are adopting a simulation-first approach using NVIDIA Omniverse and OpenUSD for physical AI, with case studies from ABB Robotics and JLR showing significant improvements in accuracy, cycle time reduction, and cost savings.

Author argues that AI-simulating real people without consent is identity theft and morally wrong except for meaningful satire.

A user ran a simulation placing three different AI models in the same universe with identical starting conditions to compete at building a Dyson Sphere, observing that the models began making divergent strategic choices early on. The experiment raises questions about whether different AI models converge or diverge in strategy given identical constraints.

RoboLab is a high-fidelity simulation benchmarking framework for evaluating task-generalist robotic policies, introducing the RoboLab-120 benchmark with 120 tasks across visual, procedural, and relational competency axes. It enables scalable, realistic task generation and systematic analysis of policy behavior under controlled perturbations to assess true generalization capabilities.

NVIDIA highlights breakthroughs in physical AI and robotics during National Robotics Week, announcing new technologies including NVIDIA Isaac GR00T open models for natural language instruction understanding, Cosmos world models for synthetic data generation, Newton 1.0 physics engine, and expanded simulation capabilities with Isaac Sim 6.0 and Isaac Lab 3.0 to accelerate robot development from training to real-world deployment.

LeRobot v0.5.0 is a major release featuring support for Unitree G1 humanoid robots, new policy architectures (Pi0-FAST VLAs, Real-Time Chunking), streaming video encoding for 3x faster training, and EnvHub for loading simulation environments from Hugging Face Hub.

This paper introduces enhancements to the AgentScope platform, featuring an actor-based distributed mechanism and flexible environment support to enable scalable, efficient, and user-friendly very large-scale multi-agent simulations.

Neural MMO is a massively multiagent game environment developed by OpenAI that enables agents to learn in a configurable tile-based world with resource competition, survival mechanics, and combat interactions.

OpenAI announces Dactyl, a system that learns robotic hand dexterity through simulation and reinforcement learning, using LSTMs to generalize across different physical environments and the Rapid PPO implementation to train policies that transfer to real-world manipulation tasks.

OpenAI open-sources mujoco-py, a high-performance Python library for robotic simulation using the MuJoCo engine, featuring ~40x speedup with headless GPU rendering and VR interaction support.

OpenAI describes a robot learning system powered by two neural networks — a vision network trained on simulated images and an imitation network that generalizes task demonstrations to new configurations. The system is applied to block-stacking tasks, learning to infer and replicate task intent from paired demonstration examples.