Back Rohan Paul

Rohan Paul· @rohanpaul_ai · X· 2026-11-01 04:13 · Original publication time AI score54

LongChang tests AI scientific discovery in simulated worlds with altered physical lawsMachine translation

Automatically verified and published · Generated and evidence-checked automatically; not reviewed by a human.

AI introduction

The LongChang team evaluated 12 large language models across 14 simulated worlds with altered physical laws. The team reported a 16.7% discovery rate for GPT-6-Astra and 0% for the other 11 models; a fivefold increase in exploration budget brought no improvement.

Original post · AI translation

Rohan Paul@rohanpaul_ai

Translation · refer to the original

@XuXander24218 Grad students in physics education could use these scenes to teach controlled experiments

ReplyXander Xu@XuXander24218

Can AI discover physical laws that humanity has never known? Excited to share LongChang (龙场), our team's work evaluating frontier AI models on autonomous scientific discovery. Named after the place where Chinese philosopher Wang Yangming (王阳明) attained enlightenment, LongChang challenges 12 LLMs to explore 14 simulated worlds with fundamentally altered physical laws. 📊 Results: - GPT-6-Astra: 16.7% discovery rate - All other 11 models: 0% - 5× more exploration budget brought no improvement. Can AI discover what it doesn't even know to ask? 🌐 Try our interactive demo directly on our website! We'd love to hear what you think. 📄 Paper: https://t.co/kLzkxN5g8r 💻 Code: https://t.co/tzfNqEdoxC first author @ZhenYang_CS

View this post on X ↗

04:13 · 2026-11-01 · UTC+8

Found an error? Send a correction