Xidong Feng
@Xidong_Feng
Followers
893
Following
257
Statuses
143
@GoogleDeepMind research scientist, working on LLM, RL, and more.
Joined June 2023
Thrilled to share that I’ll be joining @GoogleDeepMind as a Research Scientist with the Discovery Team! It’s a dream come true—8 years ago, I watched AlphaGo live in high school, and now I am lucky enough to be part of this incredible journey. Can’t wait to discover what’s next!
25
14
461
@_zhihuixie Solid work and congrats! You may also find our work related to the first stage critic training:
1
1
7
RT @ZhiyuanCS: 🚀 Call for Reviewers! 🚀 Our Workshop on Reasoning and Planning for LLMs at ICLR 2025 @iclr_conf has received an overwhelmi…
0
7
0
RT @Mengyue_Yang_: 🚨 4 Days Left! 🚨 ICLR 2025 Workshop: "World Models: Understanding, Modelling, and Scaling" is calling for submissions!…
0
19
0
RT @seohong_park: Excited to introduce flow Q-learning (FQL)! Flow Q-learning is a *simple* and scalable data-driven RL method that trains…
0
141
0
Only a week left!
🚀 Call for Papers: ICLR 2025 Workshop on World Models! 🌍🤖 📅 Submission Deadline: 10th Feb 2025 23:59 AOE 🌐 Website: We invite submissions on understanding, modeling, and scaling #WorldModels—from knowledge extraction to model-based RL, multimodal world models, and their applications in AI, robotics, and scientific discovery. 📩 Join us in shaping the future of AI-driven world modeling! #ICLR2025 #WorldModels #AI #ML #RL
0
2
11
RT @snowkylin: A Transformer for circuit design with feasibility guarantee! #ICLR2025 Introducing Circuit Transformer. It generates logic…
0
2
0
Cannot agree more. Things are even worse for overseas students. I was a self-funded UCL CS PhD and thanks to my supervisor who covered my stipend and half of my tuition fee. But I still paid 30k pounds for tuition fee-- but the funny thing is we didn't have any required courses.
Couldn't agree more. "UK Research and Innovation funding in the UK fell under the previous government from 6,835 in 2018-19 to 4,900 in 2022-23". To give a concrete example (with my @UCLCS professor hat on): 4 out of 7 @UCL_DARK PhD students were funded by the Centre for Doctoral Training (CDT) in Foundational AI at @ai_ucl. @akbirkhan @LauraRuis @_robertkirk @PaglieriDavide won Best Paper Awards at international top-tier conference, made significant contributions to AI safety, expanded our understanding of how LLMs learn to reason, and built difficult evaluations of agentic capabilities of LLMs while many other benchmarks are saturating. @UCL_DARK alumni start startups (@WecoAI), work in leading AI labs like @GoogleDeepMind, @AnthropicAI, @AIatMeta, or work in government at @AISafetyInst. @UCL_DARK wouldn't be what it is today without that CDT funding. Yet, despite the tremendous success of the @UCL Centre for Artificial Intelligence, the CDT was discontinued. @UCL_DARK now has six open positions for AI PhDs to start in Fall 2025, and it's unclear whether we will be able to make any funded offers. In turn, our lab is already significantly scaling down MSc thesis supervision, and thus not doing as much as we would like to train the next generation of AI experts. It the UK wants to have any chance at keeping up with AI, PhD funding, in addition to securing significant compute for academic research, should be their top two main priorities. Without these, the "talent" in the talent pipeline is missing. While we are at it, the starting salary for an assistant professor in the UK is in the range of £50K-60K which simply is not enough to attract international top faculty in AI to the UK. The third priority should be topping up AI postdoc and faculty salaries.
0
0
6
@aviral_kumar2 The instruction says the paper is up to 8 pages, but ICLR main conference is 9 pages and also ICML is 8 double-column pages (so basically 9 in ICLR format). It is a bit annoying to trim the paper if we resubmit our ICML submission to the workshop. Any thoughts on this?
0
0
0
Also, I think this is not like the emergence of Move 37 from AlphaZero in the game of Go. Note that AlphaZero is doing tree search on all legal moves -- so that's the largest search space. LLM leverages model generation -- if you can't sample it, you can't optimize for it. This is the pruned search space. I believe it is very unlikely to sample out of the distribution reasoning step (e.g sample back-tracing and self-correction without training on any of such data in pretraining).
0
0
1
@rm_rafailov @Grad62304977 @rosstaylor90 @Shalev_lif @madiator Exactly, for a model that has never been trained on similar data, unless you use large enough batch size - maybe you can get miracle from one out of one million trials. I dont think this is likely.
3
0
3
@rm_rafailov @Grad62304977 @Shalev_lif @madiator @rosstaylor90 Agree with this. We have no idea what the pretraining data looks like. And from what I know a lot of companies will put synthetic data into pretraining process.
1
0
5
@zdhnarsil I asked exactly the same q a few months ago hah 😆
A question about the Process reward model in a lot of LLM reasoning papers: Why almost no paper call automatic PRM dataset building (e.g., the pipeline in Math-shepherd) Monte-Carlo value function estimate (term from RL)? They are exactly the same thing.
1
1
6
RT @TheTuringPost: What are the benefits of using Natural Language Reinforcement Learning (NLRL)? • Better interpretability, because NLRL…
0
12
0
5M USD for the best open source model, impressive!
🚀 Introducing DeepSeek-V3! Biggest leap forward yet: ⚡ 60 tokens/second (3x faster than V2!) 💪 Enhanced capabilities 🛠 API compatibility intact 🌍 Fully open-source models & papers 🐋 1/n
0
0
12
RT @TheTuringPost: NLRL, or Natural Language Reinforcement Learning, is about adapting RL methods to work in the natural language field. T…
0
85
0
RT @demishassabis: It’s been an amazing last couple of weeks, hope you enjoyed our end of year extravaganza as much as we did! Just some o…
0
464
0
RT @YifeiZhou02: 🚨🚨🚨 What to do when pre-training ends? Excited to share our latest work Proposer-Agent-Evaluator (PAE), where we trained a…
0
43
0