Large Language Model, Reasoning* How to improve the reasoning ability of language models (How to Learn From Mistakes on Grade-School Math Problems, https://www.arxiv.org/abs/2408.16293): Exploring how language models can benefit from learning through "data with mistakes and corrections" to better align with the human problem-solving process in mathematics.* How to understand the reasoning ability of language models (Grade-School Math and the Hidden Reasoning Process, https://arxiv.org/abs/2407.20311): Revealing the underlying mechanisms language models use to solve mathematical problems, providing insights into their reasoning capabilities.* For further details, please check our ICML 2024 tutorial (https://www.youtube.com/watch?v=yBL7J0kgldU) or refer to homepage (https://physics.allen-zhu.com)Dynamic Ads, Product-Centric Advertising* Owner of product retrieval CTR model* Owner of entity equivalence key redefinition

Pedestrian Face Heading Detection Model* Designed and implemented a face heading detection model from scratch and successfully deployed it to production.Scale Labeling System* Built a unified system to generate custom labeling tasks for pedestrians captured in trip segments, improving data labeling efficiency and scalability.

Syntactic Analysis

Chinese Named Entity Recognition