Qintong Li

Qintong Li

PhD student

The University of Hong Kong

Biography

Hi! I’m a final-year PhD student in computer science at The University of Hong Kong, advised by Lingpeng Kong and BCM Kao. I have previously interned at Tencent AI Lab, working with Piji Li and Wei Bi.

My primary research interests lie in the area of natural language processing and machine learning. My research aims to build robust and explainable AI systems that can solve real-world problems, concerning both performance and social implications. Toward this goal, my work spans LLM alignment, long-range reasoning, targeting the major bottlenecks that cannot be solved simply by scaling up data to train the next generation of LLMs.

Interests

  • LLM alignment
  • Mathematical reasoning
  • Evaluation

Education

  • Ph.D. in Computer Science, 2025

    The University of Hong Kong

  • M.Sc. in Computer Science, 2021

    Shandong University

  • B.Eng. in Software Engineering, 2018

    Shandong University

Publications

Forewarned is Forearmed: Leveraging LLMs for Data Synthesis through Failure-Inducing Exploration

We presents ReverseGen, a new paradigm for generating effective synthetic data from the “failure” cases of a target model on specific tasks. We optimize a language model by rewarding it for generating instructions that cause failures in the target model while employing a selection strategy to maintain instruction diversity. This optimization objective is achieved through an iterative preference learning algorithm. (ICLR 2025)
Qintong Li, Jiahui Gao, Sheng Wang, Renjie Pi, Xueliang Zhao, Chuan Wu, Xin Jiang, Zhenguo Li, Lingpeng Kong
PDF

Exploring the Reliability of Large Language Models as Customized Evaluators for Diverse NLP Tasks

Our analysis shows that 1) LLM evaluators can generate unnecessary criteria or omit crucial criteria, resulting in a slight deviation from the experts. 2) LLM evaluators excel in general criteria, such as fluency, but face challenges with complex criteria, such as numerical reasoning. (COLING 2025)
Qintong Li, Leyang Cui, Lingpeng Kong, Wei Bi
Code

Retrieved Sequence Augmentation for Protein Representation Learning

We introduce Retrieved Sequence Augmentation(RSA) for protein representation learning without additional alignment or pre-processing. RSA links query protein sequences to a set of sequences with similar structures or properties in the database and combines these sequences for downstream prediction. (EMNLP 2024)
Chang Ma, Haiteng Zhao, Lin Zheng, Jiayi Xin, Qintong Li, Lijun Wu, Zhihong Deng, Yang Lu, Qi Liu, Lingpeng Kong
PDF Code
Retrieved Sequence Augmentation for Protein Representation Learning

GSM-Plus: A Comprehensive Benchmark for Evaluating the Robustness of LLMs as Mathematical Problem Solvers

We introduce the adversarial grade school math (GSM-Plus) dataset, an extension of GSM8K augmented with various mathematical perturbations. Our experiments on 25 LLMs and 4 prompting techniques show that while LLMs exhibit different levels of math reasoning abilities, their performances are far from robust. (ACL 2024)
Qintong Li, Leyang Cui, Xueliang Zhao, Lingpeng Kong, Wei Bi
PDF Code Dataset
GSM-Plus: A Comprehensive Benchmark for Evaluating the Robustness of LLMs as Mathematical Problem Solvers

MAGE: Machine-generated Text Detection in the Wild

Recent advances in large language models highlights the importance of deepfake text detection to avoid potential risks such as fake news propagation and plagiarism. We build a wild testbed by gathering texts from various human writings and deepfake texts generated by different LLMs. (ACL 2024)
Yafu Li, Qintong Li, Leyang Cui, Wei Bi, Longyue Wang, Linyi Yang, Shuming Shi, Yue Zhang
PDF Code
MAGE: Machine-generated Text Detection in the Wild

BBA: Bi-Modal Behavioral Alignment for Reasoning with Large Vision-Language Models

We introduce the Bi-Modal Behavioral Alignment prompting method, designed to maximize the potential of Domain-Specific Languages (DSL) in augmenting complex multi-modal reasoning tasks. BBA initiates by guiding large vision-language models to create separate reasoning chains for visual and DSL representations. (Findings of ACL 2024)
Xueliang Zhao, Xinting Huang, Tingchen Fu, Qintong Li, Shansan Gong, Lemao Liu, Wei Bi, Lingpeng Kong
PDF
BBA: Bi-Modal Behavioral Alignment for Reasoning with Large Vision-Language Models

Explanation Regeneration via Information Bottleneck

Explanations generated through single-pass prompting often lack sufficiency and conciseness, we develop an information bottleneck method to produce refined explanations that are sufficient and concise. (Findings of ACL 2023)
Qintong Li, Zhiyong Wu, Lingpeng Kong, Wei Bi
PDF Code
Explanation Regeneration via Information Bottleneck

A Cognitive Stimulation Dialogue System with Multi-source Knowledge Fusion for Elders with Cognitive Impairment

When communicating with elders with cognitive impairment, cognitive stimulation (CS) helps to maintain the cognitive health of elders. We construct a Chinese CS conversation dataset and propose a multi-source knowledge fusion method for CS dialogue. (ACL 2023)
Jiyue Jiang, Sheng Wang, Qintong Li, Lingpeng Kong, Chuan Wu
PDF Code
A Cognitive Stimulation Dialogue System with Multi-source Knowledge Fusion for Elders with Cognitive Impairment

Efficient Zero-shot Learning via Dataset Generation

We study a flexible and efficient zero-short learning method. Given a zero-shot task, we first generate a dataset from scratch using PLMs in an unsupervised manner. Then, we train a tiny task model under the supervision of the synthesized dataset. (EMNLP 2022)
Jiacheng Ye, Jiahui Gao, Qintong Li, Hang Xu, Jiangtao Feng, Zhiyong Wu, Tao Yu, Lingpeng Kong
PDF Code
Efficient Zero-shot Learning via Dataset Generation

Event Transition Planning for Open-ended Text Generation

A novel two-stage method which explicitly arranges the ensuing events in open-ended text generation. (Findings of ACL 2022)
Qintong Li, Piji Li, Wei Bi, Zhaochun Ren, Yuxuan Lai, Lingpeng Kong
PDF Code Poster Video
Event Transition Planning for Open-ended Text Generation

Knowledge Bridging for Empathetic Dialogue Generation

The first attempt to leverage external knowledge to accurately perceive and appropriately express implicit emotions in empathetic dialogue generation. (AAAI 2022)
Qintong Li, Piji Li, Zhaochun Ren, Pengjie Ren, Zhumin Chen
PDF Code Video
Knowledge Bridging for Empathetic Dialogue Generation

Abstractive Opinion Tagging

We propose a new task, abstractive opinion tagging, and a dataset, eComTag, to generate opinion tags based on large volumes of item reviews. (WSDM 2021, oral paper)
Qintong Li, Piji Li, Xinyi Li, Zhaochun Ren, Zhumin Chen, Maarten de Rijke
PDF Code Dataset Video
Abstractive Opinion Tagging

EmpDG: Multi-resolution Interactive Empathetic Dialogue Generation

A multi-resolution adversarial model – EmpDG, to generate more empathetic responses. (COLING 2020, oral paper)
Qintong Li, Hongshen Chen, Zhaochun Ren, Pengjie Ren, Zhaopeng Tu, Zhumin Chen
PDF Code Video
EmpDG: Multi-resolution Interactive Empathetic Dialogue Generation

Projects

eComTag Dataset

Our collected eComtag dataset for abstractive opinion tagging research.

Empathetic Dialogue Generation

A humanized dialogue system is expected to generate empathetic replies, which should be sensitive to the users’ expressed emotion. The task of empathetic dialogue generation is proposed to address this problem.
Empathetic Dialogue Generation

Text Summarization

Given single-documents or multi-documents, summarizing the opinions expressed of the input is a vital task in NLP. I divide the current researches into two categories: keyphrase generation and opinion summarization.
Text Summarization

Services

Sep 1, 2023 1 min read

TA of COMP7603

I serve as a teacher assistant for course Natural Language Processing (for masters) COMP3361. (2023 Fall, 2024 Fall)
Sep 1, 2021 1 min read

TA of COMP3361

I serve as a teacher assistant for course Natural Language Processing (for undergraduates) COMP3361. (2021 Fall, 2022 Fall)