USTC releases ExeWorldBench, first benchmark for code-based executable 3D world modeling
ustc-community · hf · 2026-10-14
Researchers from USTC and CUHK introduce ExeWorldBench, claimed as the first benchmark to comprehensively evaluate executable world modeling via code-based 3D world construction, testing both world understanding and executable realization. It spans three progressively harder tasks — Mechanism Executability, Compositional Executability, and Intent-Driven Executable World Synthesis. Experiments on frontier models reveal substantial limitations in physically grounded, causally coherent world construction.
More from Research
- There is no such thing as an AI drug, argues researcher Patrick Kidger — PatrickKidger · 2026-10-14
- More thinking, worse scores: refusals tank closed-model evals at max reasoning effort — _lewtun · 2026-10-14
- Integer multiplication repo becomes an agent harness benchmark — and collaboration loses — iskander · 2026-10-14
- Recurrent self-conditioning: framing recent diffusion methods as depth decoupled from time — alec_helbling · 2026-10-14
- MIT Researcher Builds an Interactive Digital Twin of His Dog's Brain From Real fMRI — Sad_Measurement_7502 · 2026-10-14
- Linux kernel CVE fixes plunge from 600 to 1 per week as AI code review zeroes out bugs — MarkoMarjamaa · 2026-10-14