Yuxiao Dong (@ericdongyx) profile picture

Associate Prof. of Tsinghua CS reasoning & vision & agent for #llms prev. @MetaAI @MSFTResearch

Joined October 2011
Photo by Yuxiao Dong (@ericdongyx)
Yuxiao Dong retweeted
Nobody calls their models Small, Medium, Large these days. OpenAI: Luna, Terra, Sol, Astra Google: Hydrogen, Argon, Carbon Anthropic: Haiku, Sonnet, Opus, Fable AI labs must have gotten this grand idea from Starbucks: Tall, Grande, Venti.
164
169
34
4,791
172,321
😕😕
Brute force computation is not enough for intelligence. Clever new methods are needed. The science of AI is far from finished. IMHO
56
Yuxiao Dong retweeted
I think people will generally underestimate how much technological acceleration is directly attributable to deepseek. Because of them, it is going to proliferate. Really, it's because of them. Thank you, for everything
109
395
56
6,693
219,721
Not just China but the whole world (or humanity, as AI CEOs like to call it) is fortunate to have this group of people who quietly put their heads down, tackle the problems that need solving, and share their work with every single one of us 🙏❤️
🚀 Introducing DeepInfer, our in-house inference framework, now fully open source! Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable. 🔗 github.com/deepseek-ai/DeepI… 1/4
1
141
Yuxiao Dong retweeted
OpenAI's Navier-Stokes effort used ~10,000 concurrent agents and ~130B output tokens. It made us rethink the scale ahead for RL infrastructure. slime v0.4.0 is a first step toward greater scale, with four architectural changes: • Focus: synchronous training for algorithmic exploration and correctness; fully async training for throughput at scale. • Distributed rollout: spread orchestration across cluster CPUs, with independent generation pools and event loops. • Persistent storage: introduce straw, our new storage middleware for rollout and training data, providing durable queues and shared tensor storage. • Independent recovery: restart Megatron while healthy SGLang engines stay running, preserving accepted rollout work. More on the architecture below 👇 github.com/THUDM/slime/relea… 1/5
6
15
3
131
12,342
Yuxiao Dong retweeted
Btw GLM 5.3 Flash (had codex go find the results, could be off)
🤖 Made with AI
11
8
6
139
105,444
Yuxiao Dong retweeted
Internet may collapse if GPT-7 is so good it understands all the vagueposts
7
3
283
12,230
respect
Nearly half a year of silence. We spent it studying one problem: how far RL can scale. MiMo-V2.6 is in the middle of its RL run right now. Three things we scaled: compute (~2B tokens per step, 1568 prompts × 16 rollouts, fully async), environments and harnesses (multi-task agentic RL, mixed across multiple harnesses in one run), and grader compute (agentic in-group credit assignment, with test-case and rubric-based rewards). We'll open-source the details piece by piece over the coming weeks. Streaming the run: mimo.xiaomi.com/rl/
1
1
199
Yuxiao Dong retweeted
Introducing ApprenticeBench: computer use + continual learning on a real job. We show Fable 5.1 and GPT-6 Astra can now continually learn on a job and surpass human professionals. A decisive step change in AI's job readiness. No FDEs. Agents deploy themselves into the job. 🧵
35
167
51
1,224
233,462
many more
Prediction: 50% of Neolabs will pivot into the RLaaS market
2
228
Yuxiao Dong retweeted
Add some 🧃 to muse spark 1.2 :) Try this agent for your daily professional jobs!
Replying to @ArtificialAnlys
Muse Spark 1.2 ranks #5 on GDPval-AA v2, our benchmark of agentic real-world knowledge work, at 1631 Elo - behind e.g. Claude Opus 5 (max, 1852), GPT-5.6 Sol (max, 1730), and Kimi K3 (1685), and ahead of e.g. Claude Opus 4.8 (max, 1588). Muse Spark 1.1 scored 1371 at its launch last month
9
8
85
11,625
Yuxiao Dong retweeted
An internal version of Astra, @OpenAI’s next major model family, solved 10 major open problems in mathematics, quantum complexity, and theoretical computer science. We believe it will be a major step for scientific reasoning. openai.com/index/ten-advance…
10 proofs from our next major model Astra on long-standing open problems in mathematics and theoretical computer science (also including new circuit lower bounds for computing the permanent!) GPT-5.6 has already enabled so much exciting work in math and science. Can’t wait to see what comes next!
826
2,508
1,516
17,397
13,385,746
Yuxiao Dong retweeted
We got attacked by secret unreleased proprietary models and defended ourselves with an open model, more precisely the @nvidia quantized version of GLM 5.2 coming from @Zai_org. Banning any open model would hurt first cyber security defenders, startups, small companies, researchers and everyone who's not a frontier lab and need on-prem affordable controlable models to compete and protect themselves. Let's not do that!
156
567
97
3,788
396,668
Yuxiao Dong retweeted
🔥💯
1,021
1,768
147
32,583
6,310,518
from outside to inside, i started to understand why someone I respect is proud to be friends with Wenfeng, and has started quoting Wenfeng in talks 🫥
New Wenfeng just dropped
1
2
34
7,852
Yuxiao Dong retweeted
So proud of our security team! They caught, contained & publicly disclosed an attack unlike anything we've seen before, and did it at record speed. Also massively grateful to @Zai_org: they shared GLM5.2 as open weights (for free!) with the world and it became a key part of our defense. This is day one for cybersecurity in the age of agents & we're all learning that secrecy is not the answer & that all defenders (not just a few selected ones) everywhere need more powerful models without restrictions, especially open ones!
Hardest IR of my career: one narrow objective, endless parallel paths, machine speed. One takeaway, we fought back with open models, in the open. AI security won’t be solved by one company in secret. Open source puts these tools in every defender’s hands
175
677
139
4,840
505,806
muse 1.1 is not bad, so 3.6F as a flash tier one is ok?
gemini who? 🏎️💨
2
314
Yuxiao Dong retweeted
Here’s another example: Hugging Face tried using American frontier models to analyze an AI-powered cyber attack. But the guardrails blocked requests containing real exploit payloads so they switched to GLM 5.2 running locally. The guardrails actually impaired defensive security.
Replying to @DavidSacks
We had this experience ourselves this week! Very scary to be guardrailed as a defender when you know attackers are likely bypassing
67
142
27
2,033
277,591