Yi-Lin (Pascal) Tuan

pascaltuan at gmail.com

I’m a researcher of machine learning algorithms, specializing in conversational AI, reinforcement learning, model interpretability. My research includes:

Learning Algorithms to improve efficiency and efficacy: Atom Modeling (2024), Knowledge-Grounded Reinforcement Learning (2023), Maximizing ATE (2022-2024)
Reinforcement Learning with Learned Reward Models for Text Generation: PPO for Text Generation (2018), Stepwise Rewarding GAN for Text Generation (2018), Safety and Helpfulness Control (2024)
Knowledge Augmented Chatbots: Dynamic Knowledge Graph-Grounded Dialogue Model (2019), Differential Knowledge Graph Dialogue Model (2022), Knowledge Injection (2020)
Text Generative Model Explanation and Evaluation: Language Model Generation Explanation (2021), Interpretable Knowledge Augmented Dialogue Model (2022), Quality Estimation with Synthetic Data (2021)

I received a PhD in Computer Science from the University of California Santa Barbara, where I worked with Professor William Yang Wang. I received a BS in Electrical Engineering from National Taiwan University, where I fortunately worked with professors Lin-shan Lee, Hung-yi Lee, and Yun-Nung Chen on language modeling, deep learning algorithms, and speech processing.

Education

	PhD, Computer Science, University of California Santa Barbara	2019-2024
	BS, Electrical Engineering, National Taiwan University	2013-2017

Experience

	Graduate Student Researcher, University of California Santa Barbara	2019-2024
	Research Intern, Meta GenAI	2023
	Research Intern, Meta AI	2022
	Applied Scientist Intern, Amazon Alexa AI	2021
	Research Intern, Facebook AI	2020
	Research Assitant, National Taiwan University	2017-2019

Research Papers

A Gradient Analysis Framework for Rewarding Good and Penalizing Bad Examples in Language Models

Yi-Lin Tuan, and William Yang Wang

preprint, 2024

arXiv Bib

@article{tuan2024rgpb,
  title = {A Gradient Analysis Framework for Rewarding Good and Penalizing Bad Examples in Language Models},
  author = {Tuan, Yi-Lin and Wang, William Yang},
  journal = {preprint},
  year = {2024},
}

Towards Safety and Helpfulness Balanced Responses via Controllable Large Language Models

Yi-Lin Tuan, Xilun Chen, Eric Michael Smith, Louis Martin, Soumya Batra, Asli Celikyilmaz, William Yang Wang, and Daniel M. Bikel

preprint, 2024

arXiv Bib

@article{tuan2024towards,
  title = {Towards Safety and Helpfulness Balanced Responses via Controllable Large Language Models},
  author = {Tuan, Yi-Lin and Chen, Xilun and Smith, Eric Michael and Martin, Louis and Batra, Soumya and Celikyilmaz, Asli and Wang, William Yang and Bikel, Daniel M.},
  journal = {preprint},
  year = {2024},
}

Dynamic Latent Separation for Deep Learning

Yi-Lin Tuan, Zih-Yun Chiu, and William Yang Wang

preprint, 2024

arXiv Bib

@article{tuan2024dynamic,
  title = {Dynamic Latent Separation for Deep Learning},
  author = {Tuan, Yi-Lin and Chiu, Zih-Yun and Wang, William Yang},
  journal = {preprint},
  year = {2024},
}

Flexible Attention-Based Multi-Policy Fusion for Efficient Deep Reinforcement Learning

Zih-Yun Chiu*, Yi-Lin Tuan*, William Yang Wang, and Michael C Yip

In NeurIPS, 2023

arXiv Bib Code

@inproceedings{chiu2022knowledge,
  title = {Flexible Attention-Based Multi-Policy Fusion for Efficient Deep Reinforcement Learning},
  author = {Chiu*, Zih-Yun and Tuan*, Yi-Lin and Wang, William Yang and Yip, Michael C},
  booktitle = {NeurIPS},
  year = {2023}
}

CausalDialogue: Modeling Utterance-level Causality in Conversations

Yi-Lin Tuan, Alon Albalak, Wenda Xu, Michael Saxon, Connor Pryor, Lise Getoor, and William Yang Wang

In ACL-Findings, 2023

arXiv Bib Code

@inproceedings{tuan2023causaldialogue,
  title = {CausalDialogue: Modeling Utterance-level Causality in Conversations},
  author = {Tuan, Yi-Lin and Albalak, Alon and Xu, Wenda and Saxon, Michael and Pryor, Connor and Getoor, Lise and Wang, William Yang},
  booktitle = {ACL-Findings},
  year = {2023}
}

Not All Errors are Equal: Learning Text Generation Metrics using Stratified Error Synthesis

Wenda Xu, Yi-Lin Tuan, Yujie Lu, Michael Saxon, Lei Li, and William Yang Wang

In EMNLP-Findings, 2022

arXiv Bib Code

@inproceedings{xu-etal-2022-errors,
  title = {Not All Errors are Equal: Learning Text Generation Metrics using Stratified Error Synthesis},
  author = {Xu, Wenda and Tuan, Yi-Lin and Lu, Yujie and Saxon, Michael and Li, Lei and Wang, William Yang},
  booktitle = {EMNLP-Findings},
  year = {2022},
  address = {Abu Dhabi, United Arab Emirates},
  publisher = {Association for Computational Linguistics},
  url = {https://aclanthology.org/2022.findings-emnlp.489},
  pages = {6559--6574}
}

FETA: A Benchmark for Few-Sample Task Transfer in Open-Domain Dialogue

Alon Albalak, Yi-Lin Tuan, Pegah Jandaghi, Connor Pryor, Luke Yoffe, Deepak Ramachandran, Lise Getoor, Jay Pujara, and William Yang Wang

In EMNLP, 2022

arXiv Bib Code

@inproceedings{Albalak2022FETAAB,
  title = {FETA: A Benchmark for Few-Sample Task Transfer in Open-Domain Dialogue},
  author = {Albalak, Alon and Tuan, Yi-Lin and Jandaghi, Pegah and Pryor, Connor and Yoffe, Luke and Ramachandran, Deepak and Getoor, Lise and Pujara, Jay and Wang, William Yang},
  booktitle = {EMNLP},
  year = {2022}
}

Deep Learning–Based Perceptual Stimulus Encoder for Bionic Vision

Lucas Relic, Bowen Zhang, Yi-Lin Tuan, and Michael Beyeler

In Augmented Humans, 2022

arXiv Bib

@incollection{relic2022deep,
  author = {Relic, Lucas and Zhang, Bowen and Tuan, Yi-Lin and Beyeler, Michael},
  booktitle = {Augmented Humans},
  pages = {323--325},
  year = {2022}
}

D-REX: Dialogue Relation Extraction with Explanations

Alon Albalak, Varun Embar, Yi-Lin Tuan, Lise Getoor, and William Yang Wang

In NLP for Conversational AI workshop in ACL, 2022

arXiv Bib

@inproceedings{albalak2022d,
  title = {D-REX: Dialogue Relation Extraction with Explanations},
  author = {Albalak, Alon and Embar, Varun and Tuan, Yi-Lin and Getoor, Lise and Wang, William Yang},
  booktitle = {NLP for Conversational AI workshop in ACL},
  pages = {34--46},
  year = {2022}
}

HybriDialogue: An Information-Seeking Dialogue Dataset Grounded on Tabular and Textual Data

Kai Nakamura, Sharon Levy, Yi-Lin Tuan, Wenhu Chen, and William Yang Wang

In ACL-Findings, 2022

arXiv Bib

@inproceedings{nakamura2022hybridialogue,
  title = {HybriDialogue: An Information-Seeking Dialogue Dataset Grounded on Tabular and Textual Data},
  author = {Nakamura, Kai and Levy, Sharon and Tuan, Yi-Lin and Chen, Wenhu and Wang, William Yang},
  booktitle = {ACL-Findings},
  pages = {481--492},
  year = {2022}
}

Towards Large-Scale Interpretable Knowledge Graph Reasoning for Dialogue Systems

Yi-Lin Tuan, Sajjad Beygi, Maryam Fazel-Zarandi, Qiaozi Gao, Alessandra Cervone, and William Yang Wang

In ACL-Findings, 2022

arXiv Bib Code

@inproceedings{tuan2022towards,
  title = {Towards Large-Scale Interpretable Knowledge Graph Reasoning for Dialogue Systems},
  author = {Tuan, Yi-Lin and Beygi, Sajjad and Fazel-Zarandi, Maryam and Gao, Qiaozi and Cervone, Alessandra and Wang, William Yang},
  booktitle = {ACL-Findings},
  pages = {383--395},
  year = {2022}
}

Local explanation of dialogue response generation

Yi-Lin Tuan, Connor Pryor, Wenhu Chen, Lise Getoor, and William Yang Wang

NeuIPS, 2021

arXiv Bib Code

@article{tuan2021local,
  title = {Local explanation of dialogue response generation},
  author = {Tuan, Yi-Lin and Pryor, Connor and Chen, Wenhu and Getoor, Lise and Wang, William Yang},
  journal = {NeuIPS},
  volume = {34},
  pages = {404--416},
  year = {2021}
}

Quality Estimation without Human-labeled Data

Yi-Lin Tuan, Ahmed El-Kishky, Adithya Renduchintala, Vishrav Chaudhary, Francisco Guzmán, and Lucia Specia

In EACL, 2021

arXiv Bib

@inproceedings{tuan2021quality,
  title = {Quality Estimation without Human-labeled Data},
  author = {Tuan, Yi-Lin and El-Kishky, Ahmed and Renduchintala, Adithya and Chaudhary, Vishrav and Guzmán, Francisco and Specia, Lucia},
  booktitle = {EACL},
  pages = {619--625},
  year = {2021}
}

Parallelized reverse curriculum generation

Zih-Yun Chiu, Yi-Lin Tuan, Hung-yi Lee, and Li-Chen Fu

preprint, 2021

arXiv Bib

@article{chiu2021parallelized,
  title = {Parallelized reverse curriculum generation},
  author = {Chiu, Zih-Yun and Tuan, Yi-Lin and Lee, Hung-yi and Fu, Li-Chen},
  journal = {preprint},
  year = {2021}
}

Knowledge injection into dialogue generation via language models

Yi-Lin Tuan, Wei Wei, and William Yang Wang

preprint, 2020

arXiv Bib

@article{tuan2020knowledge,
  title = {Knowledge injection into dialogue generation via language models},
  author = {Tuan, Yi-Lin and Wei, Wei and Wang, William Yang},
  journal = {preprint},
  year = {2020}
}

DyKgChat: Benchmarking Dialogue Generation Grounding on Dynamic Knowledge Graphs

Yi-Lin Tuan, Yun-Nung Chen, and Hung-Yi Lee

In EMNLP, 2019

arXiv Bib Code

@inproceedings{tuan2019dykgchat,
  title = {DyKgChat: Benchmarking Dialogue Generation Grounding on Dynamic Knowledge Graphs},
  author = {Tuan, Yi-Lin and Chen, Yun-Nung and Lee, Hung-Yi},
  booktitle = {EMNLP},
  pages = {1855--1865},
  year = {2019}
}

Personalized Dialogue Response Generation Learned from Monologues.

Feng-Guang Su*, Aliyah R Hsu*, Yi-Lin Tuan, and Hung-Yi Lee

In Interspeech, 2019

Bib PDF Code

@inproceedings{su2019personalized,
  title = {Personalized Dialogue Response Generation Learned from Monologues.},
  author = {Su*, Feng-Guang and Hsu*, Aliyah R and Tuan, Yi-Lin and Lee, Hung-Yi},
  booktitle = {Interspeech},
  pages = {4160--4164},
  year = {2019}
}

Proximal policy optimization and its dynamic version for sequence generation

Yi-Lin Tuan*, Jinzhi Zhang*, Yujia Li, and Hung-yi Lee

preprint, 2018

arXiv Bib

@article{tuan2018proximal,
  title = {Proximal policy optimization and its dynamic version for sequence generation},
  author = {Tuan*, Yi-Lin and Zhang*, Jinzhi and Li, Yujia and Lee, Hung-yi},
  journal = {preprint},
  year = {2018}
}

Improving conditional sequence generative adversarial networks by stepwise evaluation

Yi-Lin Tuan, and Hung-Yi Lee

IEEE/ACM Transactions on Audio, Speech, and Language Processing (Journal), 2018

arXiv Bib Code

@article{tuan2019improving,
  title = {Improving conditional sequence generative adversarial networks by stepwise evaluation},
  author = {Tuan, Yi-Lin and Lee, Hung-Yi},
  journal = {IEEE/ACM Transactions on Audio, Speech, and Language Processing (Journal)},
  volume = {27},
  number = {4},
  pages = {788--798},
  year = {2018},
  publisher = {IEEE}
}

Transcribing lyrics from commercial song audio: The first step towards singing content processing

Che-Ping Tsai*, Yi-Lin Tuan*, and Lin-shan Lee

In ICASSP, 2018

arXiv Bib Code

@inproceedings{tsai2018transcribing,
  title = {Transcribing lyrics from commercial song audio: The first step towards singing content processing},
  author = {Tsai*, Che-Ping and Tuan*, Yi-Lin and Lee, Lin-shan},
  booktitle = {ICASSP},
  pages = {5749--5753},
  year = {2018},
  organization = {IEEE}
}

miscs

I like to write songs, play the guitar, learn languages, and explore new things!