Tong Niu
AI researcher. I build large language models, retrieval systems, and speech models — and ship them.
I'm Tong (I also go by Ben; Niu sounds like “new”). Over the past nine years I have worked on LLMs, retrieval, and natural language generation: most recently an on-device speech model at Archer Aviation, before that vision-language models at Meta Superintelligence Labs, and for six years at Salesforce AI Research, where I built JudgeRank (#1 on BRIGHT), co-developed the SFR-Embedding model (#1 on MTEB), and generated a regression test suite with LLMs that saves millions of dollars per release.
Lately I also prove small theorems in combinatorics for fun. Earlier I wrote two books and founded an English school. This site is where I think out loud.
Latest thoughts
All thoughts →- Saw a hummingbird
Saw a hummingbird on my run today, and it suddenly hit me: the Golden Snitch in Harry Potter must have been inspired by this bird.
- The last little wing
A little bug flying past me, a dumpster someone climbs right into: to truly do what you love, you sometimes have to pay a heavier price.
- Don't just look at their current use case
The reason to join a company is not just what they do at the moment: it is the underlying mission and direction that matter the most.
- Current model training doesn't seem to have a sleep phase
We draw a lot of analogies from the human brain when we train neural networks. But one thing seems to be lacking — sleep, or deep rest.
- Quantization is based on the idea of a histogram
I suddenly realized that quantization is based on the idea of a histogram.
News
All news →This site is rebuilt from scratch
New home for my thoughts, publications and patents, and CV. First post: Quantization is just a histogram.