Zhipeng Wang 🇺🇦

A funny thing happened this morning: I saw an email from a recruiting firm who kept following up with me on the roles they are hiring. I politely replied back given that they have been emailed me 3 times. Then quickly after that, an AI agent replied to me saying that the role is no longer moving forward .... (I did not indicate anything about my interests anyway). It's so funny that the follow-up email just came through 1 hr ago. So Are they really hiring or just bombarding people's email inbox? If you want to run a serious AI recruiting firm you'd be better to be more human like and logical.

Will be in COLM next week at SF, DM for coffee chat:)

Happy to contributed a small way into the model development, learned a lot from the Gemini pre-training Infra team, now into RSI and also the fleet reliability journey. It was tough but fun. I got the chance to push myself into uncomfortable zone and grind hard. I know it’s still a long journey going forward but really excited about the path to AGI!!!

We are back!!!

Learning Jev from @CompleteSkeptic 's presentation is super inspiring, which triggered me to re-visit some of my past research on training prefill-only LLMs. The biggest challenge for training prefill only LLMs can be summarized as "how can we capture the long-context information and "squeeze" it into a classifier"? Can long-context training actually helps model performance? The empirical findings indicate that long-context does not always lead to better performance for prefill only LLMs. We need to find the right context window, effectively combine on-policy distillation + SFT, and leverage prefill optimization + multi-item scoring kernels (kudos to the awesome work from @yundaiiiii and @qingquan_song )

aclanthology.org/2025.emnlp-ind… (EMNLP 2025 Oral). Back in 2025, we've explored the model training techniques and serving optimization enabling high-performance prefill only LLM in Recommender Systems. Inference Optimization techniques such as multi-item scoring kernel, model pruning etc. were also explored.

proceedings.mlsys.org/paper_files/pa… (MLSys 2026 Oral). We've trained, productionized and optimized the inference stack for prefill only LLMs, and deployed it for Semantic Search.

aclanthology.org/2026.findings-… (ACL 2026). You only need one single token to refine LLM safety alignment, similar idea from Jev or prefill only LLM, with much more efficient alignment algorithm.

It was a wonderful experience of exploring this particular domain, both from the ML Systems and the modeling perspective. I also remember the thought-provoking discussions with @gu_jindong_ @yundaiiiii and @qingquan_song . Science progresses periodically and the path to AGI is not straightforward. I do hope we can have multiple perspectives to think about what AGI really means, and what are the pros and cons for each path.

Also thanks @zhzHNN for the always insightful comments. Exchanging ideas with @zhzHNN is really enjoyable and full of wisdom.

Of course, thanks @CompleteSkeptic and the team for the exciting work.