Welton Wang works on Sora and robotics at OpenAI and has built AI software projects including ModerateHatespeech, BetterFoodRecs and The Bipartisan Press’s Political Bias API. He has studied computer science at Stanford and writes about evaluating language models.

Unclaimed profile

Welton Wang works on Sora and robotics at OpenAI and has built AI software projects including ModerateHatespeech, BetterFoodRecs and The Bipartisan Press’s Political Bias API. He has studied computer science at Stanford and writes about evaluating language models.

Career

Robotics Researcher at OpenAI until Present
weltonwang.com

Thoughts on Adversarial Evals

Welton Wang examines how adversarially filtered benchmarks can distort comparisons and the real-world meaning of model performance.

Open ↗

Welton's Collections

Welton's Lists

In these lists

Cosigns Given