Eric Xing details PAN 2 world model: GLP architecture and 6 dimensions of physical intelligence
HeyAmit_ · x · 2026-11-11
Eric Xing shares PAN 2, built on a Generative Latent Prediction (GLP) architecture. He argues a world model must operate in an "Aristotelian world" rather than a "Platonic world," translating physical intelligence into 6 capability dimensions: obeying laws of physics, coherence over history, persistence beyond sight, conformity to constraints, invariance to viewpoints, and adherence to objective dynamics.
PAN 2 also provides 3 advanced functions: acting as an action/plan-steerable world simulator (over long sequences of actions), enabling simulative reasoning or thought experiments, and thereby supporting agents in solving complex tasks.
The retweeter's takeaway: open weights make a model usable, but open training code makes it something you can learn from — which is why he's paying attention.
More from Models
- Google poised to drop Gemini 4 Hydrogen (Flash) on second straight day of releases — gaganghotra_ · 2026-11-11
- Wave of tiny TTS releases: 8M-param Paradee hits 0.69% WER on Open TTS Leaderboard — LysandreJik · 2026-11-11
- Rumor: Gemini 4 Carbon Touted as Ultra-Cheap, Near Fable 5.2 in Capability — jason_mayes · 2026-11-11
- OpenAI mocked for reset-only response as Opus 5.5 stays on top for a month — haider1 · 2026-11-11
- CyberGate: vLLM plugin steers LLM refusals per-domain without touching weights — DebtFluffy9219 · 2026-11-11
- Interpretability study: Qwen3's 'say' and 'do' circuits are structurally disconnected, unlike OLMo — Sauers_ · 2026-11-11