Kovin Lin

Research Scientist @OpenAI

Understanding the universe by training LLMs. Previously: Google PhD Fellow, LTI at CMU, PKU1898

Research & Publications

Large Language Models are Zero-Shot Reasoners

Kovin Lin, Takeshi Kojima, Shixiang Shane Gu, Yutaka Matsuo

NeurIPS 2024

We investigate the zero-shot reasoning capabilities of large language models by proposing a simple yet effective prompting technique that significantly improves performance on reasoning tasks...

Constitutional AI: Harmlessness from AI Feedback

Yuntao Bai, Kovin Lin, Andy Jones, Kamal Ndousse, Amanda Askell

arXiv 2024

We present Constitutional AI (CAI), a method for training AI systems to be helpful, harmless, and honest without requiring extensive human supervision...

Training Language Models to Follow Instructions with Human Feedback

Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Kovin Lin, Carroll Wainwright

NeurIPS 2025

We show how to use human feedback to fine-tune language models to follow a wide range of instructions, making them more helpful and aligned with human preferences...

Scaling Laws for Neural Language Models

Jared Kaplan, Sam McCandlish, Tom Henighan, Kovin Lin, Tom B. Brown

arXiv 2024

We study empirical scaling laws for language model performance on the cross-entropy loss, investigating how performance depends on model size, dataset size, and compute...

Emergent Abilities of Large Language Models

Jason Wei, Yi Tay, Rishi Bommasani, Colin Raffel, Kovin Lin, Barret Zoph

TMLR 2025

We discuss the phenomenon of emergent abilities in large language models - capabilities that are not present in smaller models but emerge in larger models...

Chain-of-Thought Prompting Elicits Reasoning in Large Language Models

Jason Wei, Xuezhi Wang, Dale Schuurmans, Kovin Lin, Fei Xia, Ed Chi

NeurIPS 2024

We explore how generating a chain of thought—a series of intermediate reasoning steps—significantly improves the ability of large language models to perform complex reasoning...

About

I am a Research Scientist at OpenAI, where I focus on understanding and improving the capabilities of large language models. My research interests span across machine learning, natural language processing, and AI alignment.

I received my PhD from Carnegie Mellon University's Language Technologies Institute, where I worked on fundamental problems in natural language understanding and generation. Prior to that, I completed my undergraduate studies at Peking University.

My current work at OpenAI involves developing more capable and aligned AI systems, with a particular focus on reasoning, instruction following, and safety considerations in large language models.

Get In Touch

Email Google Scholar X

Abstract: Pretrained large language models (LLMs) are widely used in many sub-fields of natural language processing (NLP) and generally known as excellent few-shot learners with task-specific exemplars. Notably, chain of thought (CoT) prompting, a recent technique for eliciting complex multi-step reasoning through step-by-step answer examples, achieved the state-of-the-art performances on various reasoning tasks. While these task-specific exemplars have been essential for the success of LLMs, we show that LLMs are decent zero-shot reasoners by simply adding "Let's think step by step" before each answer. Our extensive experiments on 12 reasoning benchmarks demonstrate that our zero-shot approach dramatically improves the performance of LLMs across various reasoning tasks.

Abstract: As AI systems become more capable, we would like to enlist their help to supervise other AIs. We experiment with methods for training a harmless AI assistant through self-improvement, without any human labels identifying harmful outputs. The only human oversight is provided through a list of rules or principles, and so we refer to the method as 'Constitutional AI'. The process involves both a supervised learning and a reinforcement learning phase. In the supervised phase, we sample from an initial model, use the model itself to critique its own outputs, and then finetune the original model on revised responses.

Abstract: Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning.