Peng Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Peng Li has a total of 73 co-authors. Here are the top co-authors by number of shared publications:
Jie Zhou
Tencent Inc
8 shared publicationsMaosong Sun
Department of Computer Science, Tsinghua University, Beijing, China
6 shared publicationsXu Han
5 shared publications- Zhiyuan Liu
Tsinghua University
5 shared publications - Yang Liu
Department of Computer Science and Technology, Tsinghua University, Beijing, China
4 shared publications Pengcheng Yang
Key Lab of Computational Linguistics, School of EECS, Peking University
3 shared publications
Latest Publications
- Open Access
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 3507-3516
EditEval: Towards Comprehensive and Automatic Evaluation for Text-guided Video Editing
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 3507-3516
Open AccessEditEval: Towards Comprehensive and Automatic Evaluation for Text-guided Video Editing
- Open Access
- 18 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ACM Transactions on Audio, Speech, and Language Processing
- Vol. 32 (23299290) , 3631-3643
Exploring Universal Intrinsic Task Subspace for Few-Shot Learning via Prompt Tuning
Why can pre-trained language models (PLMs) learn universal representations and effectively adapt to broad NLP tasks differing a lot superficially? In this work, we empirically find evidence indicating that the adaptations of PLMs to various few-shot tasks can be reparameterized as optimizing only a few free parameters in a unified low-dimensional intrinsic task subspace , which may help us understand why PLMs could easily adapt to various NLP tasks with small-scale data. To find such a subspace and examine its universality, we propose an analysis pipeline called intrinsic prompt tuning (IPT). Specifically, we resort to the recent success of prompt tuning and decompose the soft prompts of multiple NLP tasks into the same low-dimensional nonlinear subspace, then we learn to adapt the PLM to unseen data or tasks by only tuning parameters in this subspace. In the experiments, we study diverse few-shot NLP tasks and surprisingly find that in a 250-dimensional subspace found with 100 tasks, by only tuning 250 free parameters, we can recover 97% and 83% of the full prompt tuning performance for 100 seen tasks (using different training data) and 20 unseen tasks, respectively, showing great generalization ability of the found intrinsic task subspace. Besides being an analysis tool, IPTcould further help us improve the prompt tuning stability.
- 18 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ACM Transactions on Audio, Speech, and Language Processing
- Vol. 32 (23299290) , 3631-3643
Open AccessExploring Universal Intrinsic Task Subspace for Few-Shot Learning via Prompt Tuning
Why can pre-trained language models (PLMs) learn universal representations and effectively adapt to broad NLP tasks differing a lot superficially? In this work, we empirically find evidence indicating that the adaptations of PLMs to various few-shot tasks can be reparameterized as optimizing only a few free parameters in a unified low-dimensional intrinsic task subspace , which may help us understand why PLMs could easily adapt to various NLP tasks with small-scale data. To find such a subspace and examine its universality, we propose an analysis pipeline called intrinsic prompt tuning (IPT). Specifically, we resort to the recent success of prompt tuning and decompose the soft prompts of multiple NLP tasks into the same low-dimensional nonlinear subspace, then we learn to adapt the PLM to unseen data or tasks by only tuning parameters in this subspace. In the experiments, we study diverse few-shot NLP tasks and surprisingly find that in a 250-dimensional subspace found with 100 tasks, by only tuning 250 free parameters, we can recover 97% and 83% of the full prompt tuning performance for 100 seen tasks (using different training data) and 20 unseen tasks, respectively, showing great generalization ability of the found intrinsic task subspace. Besides being an analysis tool, IPTcould further help us improve the prompt tuning stability.
- 7 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ACM Transactions on Audio, Speech, and Language Processing
- Vol. 32 (23299290) , 3002-3013
Black-Box Prompt Tuning With Subspace Learning
Black-box prompt tuning employs derivative-free optimization algorithms to learn prompts within low-dimensional subspaces rather than back-propagating through the network of Large Language Models (LLMs). Recent studies reveal that black-box prompt tuning lacks versatility across tasks and LLMs, which we believe is related to the suboptimal choice of subspaces. In this paper, we introduce B lack-box prompt tuning with S ubspace L earning (BSL) to enhance the versatility of black-box prompt tuning. Based on the assumption that nearly optimal prompts for similar tasks reside in a common subspace, we propose identifying such subspaces through meta-learning on a collection of similar source tasks. Consequently, for a target task that shares similarities with the source tasks, we expect that optimizing within the identified subspace can yield a prompt that performs well on the target task. Experimental results confirm that our BSL framework consistently achieves competitive performance across various downstream tasks and LLMs.
- 7 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ACM Transactions on Audio, Speech, and Language Processing
- Vol. 32 (23299290) , 3002-3013
Black-Box Prompt Tuning With Subspace Learning
Black-box prompt tuning employs derivative-free optimization algorithms to learn prompts within low-dimensional subspaces rather than back-propagating through the network of Large Language Models (LLMs). Recent studies reveal that black-box prompt tuning lacks versatility across tasks and LLMs, which we believe is related to the suboptimal choice of subspaces. In this paper, we introduce B lack-box prompt tuning with S ubspace L earning (BSL) to enhance the versatility of black-box prompt tuning. Based on the assumption that nearly optimal prompts for similar tasks reside in a common subspace, we propose identifying such subspaces through meta-learning on a collection of similar source tasks. Consequently, for a target task that shares similarities with the source tasks, we expect that optimizing within the identified subspace can yield a prompt that performs well on the target task. Experimental results confirm that our BSL framework consistently achieves competitive performance across various downstream tasks and LLMs.
- 10 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ACM Transactions on Audio, Speech, and Language Processing
- Vol. 32 (23299290) , 486-496
Gradual Syntactic Label Replacement for Language Model Pre-Training
Pre-training serves as a foundation of recent NLP models, where language modeling tasks are performed over large texts. Typical models like BERT and GPT take the corpus as a whole and treat each word equally for language modeling. However, recent works show that the naturally existing frequency bias in the raw corpus may limit the power of the language model. In this paper, we propose a multi-stage training strategy that gradually increases the training vocabulary by modifying the training data. Specifically, we leverage the syntactic structure as a bridge for infrequent words and replace them with the corresponding syntactic labels, then we recover their original lexical surface for further training. Such strategy results in an easy-to-hard curriculum learning process, where the model learns the most common words and some basic syntax concepts, before recognizing a large number of uncommon words via their specific usages and the previously learned category knowledge. Experimental results show that such a method can improve the performance of both discriminative and generative pre-trained language models on benchmarks and various downstream tasks.
- 10 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ACM Transactions on Audio, Speech, and Language Processing
- Vol. 32 (23299290) , 486-496
Gradual Syntactic Label Replacement for Language Model Pre-Training
Pre-training serves as a foundation of recent NLP models, where language modeling tasks are performed over large texts. Typical models like BERT and GPT take the corpus as a whole and treat each word equally for language modeling. However, recent works show that the naturally existing frequency bias in the raw corpus may limit the power of the language model. In this paper, we propose a multi-stage training strategy that gradually increases the training vocabulary by modifying the training data. Specifically, we leverage the syntactic structure as a bridge for infrequent words and replace them with the corresponding syntactic labels, then we recover their original lexical surface for further training. Such strategy results in an easy-to-hard curriculum learning process, where the model learns the most common words and some basic syntax concepts, before recognizing a large number of uncommon words via their specific usages and the previously learned category knowledge. Experimental results show that such a method can improve the performance of both discriminative and generative pre-trained language models on benchmarks and various downstream tasks.
- Open Access
- 4 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 1722-1732
Learning to Relate to Previous Turns in Conversational Search
Conversational search allows a user to interact with a search system in multiple turns. A query is strongly dependent on the conversation context. An effective way to improve retrieval effectiveness is to expand the current query with historical queries. However, not all the previous queries are related to, and useful for expanding the current query. In this paper, we propose a new method to select relevant historical queries that are useful for the current query. To cope with the lack of labeled training data, we use a pseudo-labeling approach to annotate useful historical queries based on their impact on the retrieval results. The pseudo-labeled data are used to train a selection model. We further propose a multi-task learning framework to jointly train the selector and the retriever during fine-tuning, allowing us to mitigate the possible inconsistency between the pseudo labels and the changed retriever. Extensive experiments on four conversational search datasets demonstrate the effectiveness and broad applicability of our method compared with several strong baselines.
- 4 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 1722-1732
Open AccessLearning to Relate to Previous Turns in Conversational Search
Conversational search allows a user to interact with a search system in multiple turns. A query is strongly dependent on the conversation context. An effective way to improve retrieval effectiveness is to expand the current query with historical queries. However, not all the previous queries are related to, and useful for expanding the current query. In this paper, we propose a new method to select relevant historical queries that are useful for the current query. To cope with the lack of labeled training data, we use a pseudo-labeling approach to annotate useful historical queries based on their impact on the retrieval results. The pseudo-labeled data are used to train a selection model. We further propose a multi-task learning framework to jointly train the selector and the retriever during fine-tuning, allowing us to mitigate the possible inconsistency between the pseudo labels and the changed retriever. Extensive experiments on four conversational search datasets demonstrate the effectiveness and broad applicability of our method compared with several strong baselines.
- Open Access
- 1 January 2023
- Published by KEAI Publishing in AI Open
- Vol. 4, 98-110
Restricted orthogonal gradient projection for continual learning
- 1 January 2023
- Published by KEAI Publishing in AI Open
- Vol. 4, 98-110
Open AccessRestricted orthogonal gradient projection for continual learning
- Open Access
- 1 January 2023
- Published by KEAI Publishing in AI Open
- Vol. 4, 56-63
AdaDS: Adaptive data selection for accelerating pre-trained language model knowledge distillation
- 1 January 2023
- Published by KEAI Publishing in AI Open
- Vol. 4, 56-63
Open AccessAdaDS: Adaptive data selection for accelerating pre-trained language model knowledge distillation
- 3 April 2020
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 34 (05) , 7504-7511
DMRM: A Dual-Channel Multi-Hop Reasoning Model for Visual Dialog
Visual Dialog is a vision-language task that requires an AI agent to engage in a conversation with humans grounded in an image. It remains a challenging task since it requires the agent to fully understand a given question before making an appropriate response not only from the textual dialog history, but also from the visually-grounded information. While previous models typically leverage single-hop reasoning or single-channel reasoning to deal with this complex multimodal reasoning task, which is intuitively insufficient. In this paper, we thus propose a novel and more powerful Dual-channel Multi-hop Reasoning Model for Visual Dialog, named DMRM. DMRM synchronously captures information from the dialog history and the image to enrich the semantic representation of the question by exploiting dual-channel reasoning. Specifically, DMRM maintains a dual channel to obtain the question- and history-aware image features and the question- and image-aware dialog history features by a mulit-hop reasoning process in each channel. Additionally, we also design an effective multimodal attention to further enhance the decoder to generate more accurate responses. Experimental results on the VisDial v0.9 and v1.0 datasets demonstrate that the proposed model is effective and outperforms compared models by a significant margin.
- 3 April 2020
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 34 (05) , 7504-7511
DMRM: A Dual-Channel Multi-Hop Reasoning Model for Visual Dialog
Visual Dialog is a vision-language task that requires an AI agent to engage in a conversation with humans grounded in an image. It remains a challenging task since it requires the agent to fully understand a given question before making an appropriate response not only from the textual dialog history, but also from the visually-grounded information. While previous models typically leverage single-hop reasoning or single-channel reasoning to deal with this complex multimodal reasoning task, which is intuitively insufficient. In this paper, we thus propose a novel and more powerful Dual-channel Multi-hop Reasoning Model for Visual Dialog, named DMRM. DMRM synchronously captures information from the dialog history and the image to enrich the semantic representation of the question by exploiting dual-channel reasoning. Specifically, DMRM maintains a dual channel to obtain the question- and history-aware image features and the question- and image-aware dialog history features by a mulit-hop reasoning process in each channel. Additionally, we also design an effective multimodal attention to further enhance the decoder to generate more accurate responses. Experimental results on the VisDial v0.9 and v1.0 datasets demonstrate that the proposed model is effective and outperforms compared models by a significant margin.
- Open Access
- 1 January 2020
- Published by Association for Computational Linguistics (ACL)
- p. 1652-1671
MAVEN: A Massive General Domain Event Detection Dataset
Xiaozhi Wang, Ziqi Wang, Xu Han, Wangyi Jiang, Rong Han, Zhiyuan Liu, Juanzi Li, Peng Li, Yankai Lin, Jie Zhou. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP). 2020.
- 1 January 2020
- Published by Association for Computational Linguistics (ACL)
- p. 1652-1671
Open AccessMAVEN: A Massive General Domain Event Detection Dataset
Xiaozhi Wang, Ziqi Wang, Xu Han, Wangyi Jiang, Rong Han, Zhiyuan Liu, Juanzi Li, Peng Li, Yankai Lin, Jie Zhou. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP). 2020.
- 1 August 2019
- Published by International Joint Conferences on Artificial Intelligence
- p. 5116-5122
A Dual Reinforcement Learning Framework for Unsupervised Text Style Transfer
Unsupervised text style transfer aims to transfer the underlying style of text but keep its main content unchanged without parallel data. Most existing methods typically follow two steps: first separating the content from the original style, and then fusing the content with the desired style. However, the separation in the first step is challenging because the content and style interact in subtle ways in natural language. Therefore, in this paper, we propose a dual reinforcement learning framework to directly transfer the style of the text via a one-step mapping model, without any separation of content and style. Specifically, we consider the learning of the source-to-target and target-to-source mappings as a dual task, and two rewards are designed based on such a dual structure to reflect the style accuracy and content preservation, respectively. In this way, the two one-step mapping models can be trained via reinforcement learning, without any use of parallel data. Automatic evaluations show that our model outperforms the state-of-the-art systems by a large margin, especially with more than 10 BLEU points improvement averaged on two benchmark datasets. Human evaluations also validate the effectiveness of our model in terms of style accuracy, content preservation and fluency. Our code and data, including outputs of all baselines and our model are available at https://github.com/luofuli/DualRL.
- 1 August 2019
- Published by International Joint Conferences on Artificial Intelligence
- p. 5116-5122
A Dual Reinforcement Learning Framework for Unsupervised Text Style Transfer
Unsupervised text style transfer aims to transfer the underlying style of text but keep its main content unchanged without parallel data. Most existing methods typically follow two steps: first separating the content from the original style, and then fusing the content with the desired style. However, the separation in the first step is challenging because the content and style interact in subtle ways in natural language. Therefore, in this paper, we propose a dual reinforcement learning framework to directly transfer the style of the text via a one-step mapping model, without any separation of content and style. Specifically, we consider the learning of the source-to-target and target-to-source mappings as a dual task, and two rewards are designed based on such a dual structure to reflect the style accuracy and content preservation, respectively. In this way, the two one-step mapping models can be trained via reinforcement learning, without any use of parallel data. Automatic evaluations show that our model outperforms the state-of-the-art systems by a large margin, especially with more than 10 BLEU points improvement averaged on two benchmark datasets. Human evaluations also validate the effectiveness of our model in terms of style accuracy, content preservation and fluency. Our code and data, including outputs of all baselines and our model are available at https://github.com/luofuli/DualRL.