Papers by Haozhe An
On the Influence of Gender and Race in Romantic Relationship Prediction from Large Language Models (2024.emnlp-main)
Copied to clipboard
| Challenge: | We show that models are less likely to predict romantic relationships for same-gender character pairs than different-grace character pairs. |
| Approach: | They perform name-replacement experiments to examine gender biases in large language models . they hypothesize that models mirror heteronormative biase and prejudice against interracial romantic relationships . |
| Outcome: | The results suggest that models may mirror heteronormative biases and prejudice against interracial romantic relationships in human and society. |
Learning Bias-reduced Word Embeddings Using Dictionary Definitions (2022.findings-acl)
Copied to clipboard
| Challenge: | Existing word embeddings have undesirable gender, racial, and religious biases . DD-GloVe is a train-time debiasing algorithm that uses dictionary definitions based on word definitions. |
| Approach: | They propose a dictionary-guided loss function that encourages word embeddings to be similar to their relatively neutral dictionary definition representations. |
| Outcome: | The proposed algorithm can learn word embeddings by leveraging dictionary definitions. |
Susu Box or Piggy Bank: Assessing Cultural Commonsense Knowledge between Ghana and the US (2024.emnlp-main)
Copied to clipboard
| Challenge: | Recent work has highlighted the culturally-contingent nature of commonsense knowledge . a multi-stage process is used to evaluate the commonsence of English LLMs . |
| Approach: | They propose a test set of 525 multiple-choice questions to evaluate commonsense knowledge of English LLMs in Ghana and the u.s. They use existing commonsensible datasets to rewrite them in a multi-stage process. |
| Outcome: | The proposed model improves on the culturally-contingent commonsense knowledge of English LLMs in Ghana and the United States. |
Coarse-to-Fine Dual Encoders are Better Frame Identification Learners (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Recent efforts to model frame definitions lack sufficient representation learning of definitions or lack efficient frame modeling. |
| Approach: | They propose a frame-target-encoder architecture that uses coarse-to-fine learning to model alignment between frames and targets. |
| Outcome: | The proposed framework outperforms existing models by 0.93 overall scores and 1.53 R@1 without lf. |
On the Mutual Influence of Gender and Occupation in LLM Representations (2025.acl-long)
Copied to clipboard
| Challenge: | We examine LLM representations of gender for first names in various occupational contexts to study how occupations and the gender perception of first names influence each other mutually. |
| Approach: | They examine LLM representations of gender for first names in various occupational contexts and examine how occupations and the gender perception of first names influence each other mutually. |
| Outcome: | The representations shift with the occupational context and are influenced by stereotypically feminine or masculine occupations. |
Rethinking Semantic Parsing for Large Language Models: Enhancing LLM Performance with Semantic Hints (2025.acl-short)
Copied to clipboard
| Challenge: | Semantic Parsing improves performance of smaller models, but it is unclear whether it extends similarly to large language models. |
| Approach: | They propose a prompting approach that embeds semantic hints within the prompt to improve LLM performance. |
| Outcome: | The proposed approach improves LLMs’ performance across various tasks, highlighting the potential of integrating semantic information to improve LLM capabilities. |
Aligning Large Language Models to Follow Instructions and Hallucinate Less via Effective Data Filtering (2025.acl-long)
Copied to clipboard
Shuzheng Si, Haozhe Zhao, Gang Chen, Cheng Gao, Yuzhuo Bai, Zhitong Wang, Kaikai An, Kangyang Luo, Chen Qian, Fanchao Qi, Baobao Chang, Maosong Sun
| Challenge: | Existing studies show that training LLMs on data containing unfamiliar knowledge during instruction tuning can encourage hallucinations. |
| Approach: | They propose a framework that measures how familiar the LLM is with instruction data and introduce an expert-aligned reward model to ensure the quality of selected samples. |
| Outcome: | The proposed framework reduces hallucinations while maintaining a competitive ability to follow instructions. |
Mitigating Language-Level Performance Disparity in mPLMs via Teacher Language Selection and Cross-lingual Self-Distillation (2024.naacl-long)
Copied to clipboard
| Challenge: | Large-scale multilingual pretrained language models (mPLMs) yield impressive performance on cross-language tasks, yet significant performance disparities exist across different languages within the same mPLm. |
| Approach: | They propose to leverage the learned knowledge from well-performing languages to guide under-performing ones within the same mPLM. |
| Outcome: | The proposed model shows that it can guide under-performing languages while minimizing language-level performance disparities across different mPLMs. |
Do Large Language Models Discriminate in Hiring Decisions on the Basis of Race, Ethnicity, and Gender? (2024.acl-short)
Copied to clipboard
| Challenge: | We study whether large language models exhibit race- and gender-based name discrimination in hiring decisions . |
| Approach: | They propose templatic prompts to LLMs to write an email to a named job applicant informing them of a hiring decision. |
| Outcome: | The proposed model generates an acceptance or rejection email based on the applicant's first name . |
Improving the Robustness of Distantly-Supervised Named Entity Recognition via Uncertainty-Aware Teacher Learning and Student-Student Collaborative Learning (2024.findings-acl)
Copied to clipboard
| Challenge: | Named Entity Recognition (NER) methods require a substantial quantity of high-quality annotation for training models. |
| Approach: | They propose a method to reduce the number of incorrect pseudo labels in self-training . they propose 'uncertainty-aware teacher learning' and 'student-student collaboration' |
| Outcome: | The proposed method is superior to state-of-the-art DS-NER denoising methods. |
GATEAU: Selecting Influential Samples for Long Context Alignment (2025.emnlp-main)
Copied to clipboard
Shuzheng Si, Haozhe Zhao, Gang Chen, Yunshui Li, Kangyang Luo, Chuancheng Lv, Kaikai An, Fanchao Qi, Baobao Chang, Maosong Sun
| Challenge: | Existing studies have attempted to scale up the available data volume by synthesizing long instruction-following samples, but a lack of a well-defined strategy for ensuring data quality may introduce low-quality samples and restrict the model’s performance. |
| Approach: | They propose a framework to identify influential samples enriched with long-range dependency relations that can be used to align large language models to handle instructions with extremely long contexts. |
| Outcome: | The proposed framework identifies samples with long-range dependency relations and shows that the model trained on these samples exhibits better instruction-following and long-context understanding capabilities. |
SODAPOP: Open-Ended Discovery of Social Biases in Social Commonsense Reasoning Models (2023.eacl-main)
Copied to clipboard
| Challenge: | Existing diagnostic tests for detecting social biases in NLP models only detect stereotypic associations pre-specified by the designer. |
| Approach: | They propose an approach for automatic social bias discovery in social commonsense question-answering by substituting names associated with different demographic groups and generating many distractor answers from a masked language model. |
| Outcome: | The proposed approach uncovers model’s stereotypic associations between demographic groups and an open set of words. |
Nichelle and Nancy: The Influence of Demographic Attributes and Tokenization Length on First Name Biases (2023.acl-short)
Copied to clipboard
| Challenge: | Prior research has shown that social commonsense reasoning models exhibit biases along dimensions of race, ethnicity, and gender. |
| Approach: | They conduct first name substitution experiments to measure the influence of demographic attributes of a name and name tokenization length on models' disparate behavior. |
| Outcome: | The results show that demographic attributes of a name and name tokenization length are factors that systematically affect the behavior of social commonsense reasoning models. |