Yenting Lin林彥廷

I'm a Research Scientist at Google DeepMind, working on audio-native large language models: post-training and making speech generation steerable. I also build open models for Traditional Chinese — I created Taiwan-LLM.

Experience

  • Google DeepMindResearch Scientist — audio LLMs, post-training, controllability
  • Meta GenAIEnhanced reasoning with stepwise feedback and long-form reasoning
  • NVIDIA ResearchError correction methods for multimodal language models
  • Amazon Alexa AIFactuality evaluation agent and synthetic data techniques

News

  • Released Step-KTO, stepwise binary feedback for mathematical reasoning.
  • Measuring Taiwanese Mandarin Language Understanding accepted to COLM 2024.
  • Released the latest Taiwan-LLM models, open-weight Traditional Chinese LLMs.
  • Released Taiwan-LLM, the first open LLM series built for Taiwan.

Selected Publications

Open Source

Taiwan-LLM

Open-weight large language models built for Taiwanese Mandarin — pretraining data, instruction tuning, and evaluation for Traditional Chinese. Widely adopted by Taiwan's research community and industry.

Mistral-Small-Reasoning

Open-weight reasoning model distilled for efficient step-by-step problem solving.