Yenting Lin林彥廷

Research Scientist, Google DeepMind

I work on agentic post-training for long-horizon knowledge work.

Most recently, Gemini 3.8 Live Extended Thinking. Before that, open models for Traditional Chinese.

Latest

September 2026

Gemini 3.8 Live Extended Thinking

Real-time voice that reasons and calls tools in the background, without breaking the conversation.

Read the announcement

Experience

Google DeepMindAgentic post-training, long-horizon tasks, Gemini Live
Meta GenAIReasoning with stepwise feedback
NVIDIA ResearchError correction for multimodal LMs
Amazon Alexa AIFactuality evaluation and synthetic data

Selected Publications

Step-KTO: Optimizing Mathematical Reasoning through Stepwise Binary Feedback

arXiv 2025
Measuring Taiwanese Mandarin Language Understanding

COLM 2024
LLM-Eval: Unified Multi-Dimensional Automatic Evaluation for Open-Domain Conversations with LLMs

NLP4ConvAI 2023
Selective In-Context Data Augmentation for Intent Detection using Pointwise V-Information

EACL 2023
Knowledge-Grounded Conversational Data Augmentation with Generative Conversational Networks

SIGDIAL 2022
SalesBot: Transitioning from Chit-Chat to Task-Oriented Dialogues

ACL 2022

Open Source

Traditional Chinese LLMs

Open-weight models for Traditional Chinese, from pretraining data to evaluation.
Mistral-Small-Reasoning

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