Potsawee (Punpun) Manakul

Potsawee (Punpun) Manakul

Research Scientist · Speech & Multimodal AI

About Me

Hello! I'm PunpunAbout Thai names 🇹🇭
In Thailand, everyone has a nickname given at birth, which is used daily at school, work, and among friends. My nickname is Punpun.

Official/First names are often used in formal settings and in documents, publications, etc. — mine is Potsawee. I go by either name, but most friends just call me Punpun., a Visiting Researcher at the Stanford NLP Group, working with Prof. Diyi Yang. I work on audio/speech LLMs, including training foundation models and studying scaling properties (Project SODA). Also, I'm interested in multimodal AI safety and speech/audio models for low-resource languages.

From late 2023 to early 2026, I was a founding team member & first AI researcher at Typhoon (SCB 10X), an AI lab advancing Thai AI models (LLM, ASR, TTS) where I led the development of audio/speech models.

I completed my PhD in the Speech Group at Cambridge in May 2024, advised by Prof. Mark Gales. My research focused primarily on NLP—particularly text summarization and LLM factuality. Notable PhD project includes SelfCheckGPT (1500+ citations & 600+ GitHub stars). During PhD, I had a research internship at Amazon. Before this, I did my undergraduate in Engineering, also at Cambridge.

Research Areas

News

Selected Projects & Publications

Scaling Open Discrete Audio (SODA) Potsawee Manakul, Woody Haosheng Gan, Martijn Bartelds, Guangzhi Sun, William Held, Diyi Yang Preprint Feb 2026
🏠 Project Page | 📄 Paper | 📈 Experiment Log | Code | 🤗 Models & Data | 🎮 Demo
AudioJudge: Understanding What Works in Large Audio Model Based Speech Evaluation Potsawee Manakul*, Woody Haosheng Gan*, Michael J. Ryan, Ali Sartaz Khan, Warit Sirichotedumrong, Kunat Pipatanakul, William Held, Diyi Yang EACL 2026
📄 Paper | Code
Enhancing Low-Resource Language and Instruction Following Capabilities of Audio Language Models Potsawee Manakul, Guangzhi Sun, Warit Sirichotedumrong, Kasima Tharnpipitchai, Kunat Pipatanakul Interspeech 2025
📄 Paper | Code | 🤗 Model
SelfCheckGPT: Zero-Resource Black-Box Hallucination Detection for Generative Large Language Models Potsawee Manakul, Adian Liusie, Mark Gales EMNLP 2023
🏠 Project Page & Code GitHub stars | PyPI Downloads
→ Featured in the Washington Post about AI Hallucination (May 2023)
→ Featured in DeepLearning.AI course on LLM Safety
→ Featured in reading list of Stanford CS 329T (Fall 2023)
→ Invited talk at AI CoP Seminar @ U.S. SEC on LLM Hallucination (Aug 2024)
→ 1500+ citations
MQAG: Multiple-choice Question Answering and Generation for Assessing Information Consistency Potsawee Manakul, Adian Liusie, Mark Gales AACL 2023 (Area Chair Paper Award)
Code | 🤗 Models
Long-Span Summarization via Local Attention and Content Selection Potsawee Manakul, Mark Gales ACL 2021
Code & Models
→ System winning 1st place in the Spotify Podcast Summarisation task 2020
Abstractive Spoken Document Summarization using Hierarchical Model with Multi-stage Attention Diversity Potsawee Manakul, Mark Gales, Linlin Wang Interspeech 2020 (Best Student Paper Award Finalist)
Code
→ View full list on Google Scholar

Honours and Awards

Reviewing & Academic Service

Tracker