Eric Xing
@ericxing
Followers
8K
Following
617
Media
39
Statuses
438
Researcher, educator, entrepreneur, and administrator in computer science, artificial intelligence, and healthcare.
San Francisco, CA
Joined May 2010
In this paper we present the first full implementation of the Generative Latent Prediction (GLP) architecture of world modeling, that brings perception, state, action, and causality into a single, coherent world model that can plan, imagine, and reason through language,
2
15
55
While open models aren't performing well in the US, China is churning them out seemingly non-stop. Here's what I consider a unique model, a world model, out of @mbzuai in the UAE. This model is for visual simulation and is differentiated. TBD on access or open sources. The team
Introducing PAN — MBZUAI’s New World Model for Interactive Intelligence Developed by MBZUAI’s Institute of Foundation Models, PAN is built for simulation, prediction, and agentic reasoning. Unlike traditional video generators that only output frames, PAN maintains a persistent
0
4
13
Redefining the Future of AI Simulation. In a recent conversation with industry analyst Pat Moorhead, MBZUAI President and University Professor Eric Xing discussed how open and collaborative research is accelerating progress in artificial intelligence — and how models like PAN,
forbes.com
The new AI world model aims to surpass LLMs by creating simulations that generalize across domains and provide better interactivity and coherence over time.
0
2
10
A world model is more than a video generator, it understands how the world works and enables an intelligent agent to imagine, predict, and reason about how the world evolves in response to actions, and accordingly to plan and strategize. PAN employs a Generative Latent
0
1
3
I am excited to unveil PAN, a general, interactable, and long-horizon world model that invites you to step into infinitely possible real or surreal worlds unfolding from your thought experiments, so you can reason, act, and learn from such simulations anywhere and anytime.
ifm.mbzuai.ac.ae
From the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), meet PAN — an interactive world model that can see, reason, and predict how the world changes over time.
1
4
21
Great to meet @demishassabis today at @mbzuai to introduce to him our university and discuss a wide range of topics from world model to virtual cell and find lots of common grounds. In sync regarding WM toward simulating all possibilities not just pixels and GLP vs. JEPA; and
3
24
266
🌍🤝 We’re delighted to announce the launch of the MIT-MBZUAI Collaborative Research Program. #MBZUAI and MIT Schwarzman College of Computing are joining forces to launch a global collaboration to shape the future of AI. As part of this groundbreaking initiative, faculty and
1
6
30
First honorary doctorate of @mbzuai goes to @sama, witnessed by HH @𝗦𝗵𝗲𝗶𝗸𝗵 𝗞𝗵𝗮𝗹𝗲𝗱 𝗯𝗶𝗻 𝗠𝗼𝗵𝗮𝗺𝗲𝗱 𝗯𝗶𝗻 𝗭𝗮𝘆𝗲𝗱 𝗔𝗹 𝗡𝗮𝗵𝘆𝗮𝗻, and HE 𝗞𝗵𝗮𝗹𝗱𝗼𝗼𝗻 𝗞𝗵𝗮𝗹𝗶𝗳𝗮 𝗔𝗹 𝗠𝘂𝗯𝗮𝗿𝗮𝗸!
Khaled bin Mohamed bin Zayed has witnessed the inaugural Honorary Doctorate conferral ceremony of Mohamed bin Zayed University of Artificial Intelligence, awarding Sam Altman, CEO of OpenAI, in recognition of his transformative leadership in artificial intelligence. #MBZUAI
0
0
3
No sign-ups, no drama 👇
No sign-ups, no drama. 🎉 Just pure speed. See K2 Think in action and experience advanced AI reasoning at https://t.co/ltN5Yxmtvr - now with no registration required. #K2Think #AI #OpenSource #MBZUAI #G42 #Innovation
1
0
8
K2-Think is powered by the super high-throughput @CerebrasSystems, making it the fastest AI deep reasoning system on top of its leading inference quality and open source accessibility — a perfect example of what can be unleashed via academic/industry collaboration.
K2 Think is now available and runs fastest on Cerebras Inference at 2,000 TPS on Cerebras – 20x faster than GPU. Launched by @mbzuai and @G42ai, K2 Think is a world leading open-source reasoning model that beats GPT OSS 120B and DeepSeek on math and reasoning.
0
0
9
Such performance places K2-THINK at the top of all open source models on math reasoning, and is close to strong proprietary models such as o3 High which has over 1T parameters. Additionally, it is lightening fast, delivering pages of deep reasoning results in seconds. Key
0
0
5
K2-THINK 32B exhibits a micro-average score of 67.99 across all math questions, beating SOTA open source models like DeepSeek V3.1 671B (m-avg. 64.43) and GPT-OSS 120B (m-avg. 67.20) much larger, and even edging leading proprietary models like o3 & Gemini https://t.co/f1tV1LyVEX
1
2
15
K2 Think is a 32 billion parameter, open source reasoning model that punches well above its weight. Available now on Hugging Face, it’s built for advanced logic, math, and science reasoning, delivering frontier-class performance while being remarkably efficient:
huggingface.co
4
23
126