Papers by Morteza Dehghani
Social-Group-Agnostic Bias Mitigation via the Stereotype Content Model (2023.acl-long)
Copied to clipboard
Ali Omrani, Alireza Salkhordeh Ziabari, Charles Yu, Preni Golazizian, Brendan Kennedy, Mohammad Atari, Heng Ji, Morteza Dehghani
| Challenge: | Existing methods for mitigating bias require social-group-specific word pairs for each social attribute (e.g., gender) Existing approaches require only one social attribute, rendering them impractical and costly . |
| Approach: | They propose that stereotype content models capture the underlying connection between bias and stereotypes by embedding only two psychological dimensions of warmth and competence. |
| Outcome: | The proposed method performs comparably to group-specific debiasing on multiple bias benchmarks, but has theoretical and practical advantages over existing methods. |
Cost-Efficient Subjective Task Annotation and Modeling through Few-Shot Annotator Adaptation (2024.findings-emnlp)
Copied to clipboard
| Challenge: | In subjective tasks, the inclusion of diverse annotators is crucial as their unique perspectives significantly influence the annotations. |
| Approach: | They propose a framework that minimizes the annotation budget while maximizing the predictive performance for each annotator. |
| Outcome: | The proposed framework surpasses the previous SOTA in capturing the annotators’ individual perspectives with as little as 25% of the original annotation budget on two datasets. |
The Subjectivity of Respect in Police Traffic Stops: Modeling Community Perspectives in Body-Worn Camera Footage (2026.acl-long)
Copied to clipboard
Preni Golazizian, Elnaz Rahmati, Jackson Trager, Zhivar Sourati, Nona Ghazizadeh, Georgios Chochlakis, Jose J. Alcocer, Kerby Bennett, Aarya Vijay Devnani, Parsa Hejabi, Harry G. Muttram, Akshay Kiran Padte, Mehrshad Saadatinia, Chenhao Wu, Alireza Salkhordeh Ziabari, Michael Sierra-Arévalo, Nicholas Weller, Shrikanth Narayanan, Benjamin A.t. Graham, Morteza Dehghani
| Challenge: | a new study examines the perception of police-civilian traffic stops using respect ratings and free-text rationales from multiple perspectives. |
| Approach: | They propose a traffic-stop dataset annotated with respect ratings and rationales from multiple perspectives . they use a criterion-driven preference data construction framework to predict personalized respect ratings . |
| Outcome: | The proposed framework improves rating prediction performance and rationale alignment across all three annotators. |
Contextualizing Hate Speech Classifiers with Post-hoc Explanation (2020.acl-main)
Copied to clipboard
| Challenge: | Modern text classifiers struggle to learn a model of hate speech that generalizes to real-world applications. |
| Approach: | They propose a method to regularize BERT classifiers to detect bias towards identity terms by providing explanations for group identifiers and allowing models to learn from the context of group identifiers. |
| Outcome: | The proposed method limiting false positives on out-of-domain data while maintaining and improving in-domain performance. |
Flip-Flop Consistency: Unsupervised Training for Robustness to Prompt Perturbations in LLMs (2026.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) often produce inconsistent answers when faced with different phrasings of the same prompt. |
| Approach: | They propose an unsupervised method that flips the phrasings of prompts into a hard pseudo-label . they use Consensus Cross-Entropy to create a consensus, and representation alignment loss to pull lower-confidence predictors toward consensus . |
| Outcome: | The proposed method raises observed agreement by 11.62% and improves mean F1 by 8.94% on 11 datasets spanning four NLP tasks . |
Reinforced Multiple Instance Selection for Speaker Attribute Prediction (2024.naacl-long)
Copied to clipboard
Alireza Salkhordeh Ziabari, Ali Omrani, Parsa Hejabi, Preni Golazizian, Brendan Kennedy, Payam Piray, Morteza Dehghani
| Challenge: | Current methods for predicting speaker attributes take a speaker’s utterances as input and provide a prediction per speaker attribute. |
| Approach: | They propose a Multiple Instance Learning approach that uses Reinforcement Learning to predict speaker attributes using a set of utterances from social media posts and political ideologies from transcribed speeches. |
| Outcome: | The proposed approach outperforms existing methods on a range of related tasks including predicting speakers’ psychographics and demographics from social media posts and political ideologies from transcribed speeches. |
LAD-RAG: Layout-aware Dynamic RAG for Visually-Rich Document Understanding (2026.acl-long)
Copied to clipboard
Zhivar Sourati, Zheng Wang, Marianne Menglin Liu, Yazhe Hu, Mengqing Guo, Sujeeth Bharadwaj, Kyu J. Han, Tao Sheng, Sujith Ravi, Morteza Dehghani, Dan Roth
| Challenge: | Conventional retrieval-augmented generation (RAG) methods encode content in isolated chunks during ingestion, losing structural and cross-page dependencies, and retrieve a fixed number of pages at inference. |
| Approach: | They propose a Layout-Aware Dynamic RAG framework that encodes content in isolated chunks during ingestion and retrieves a fixed number of pages at inference. |
| Outcome: | Experiments on MMLongBench-Doc, LongDocURL, DUDE, and MP-DoxVQA show that LAD-RAG improves retrieval, achieving over 90% perfect recall on average without any top-k tuning, and outperforming baseline retrievers by up to 20% in recall at comparable noise levels. |
Hate Speech Classifiers Learn Normative Social Stereotypes (2023.tacl-1)
Copied to clipboard
| Challenge: | Social stereotypes negatively impact individuals’ judgments about different groups and may have a critical role in understanding language directed toward marginalized groups. |
| Approach: | They first investigate the impact of novice annotators’ stereotypes on their hate-speech-annotation behavior. Then, they examine the effect of normative stereotypes in language on the aggregated annotated judgments. |
| Outcome: | The framework provides insights into sources of bias in hate-speech moderation, informing ongoing debates regarding machine learning fairness. |
Reporting the Unreported: Event Extraction for Analyzing the Local Representation of Hate Crimes (D19-1)
Copied to clipboard
Aida Mostafazadeh Davani, Leigh Yeh, Mohammad Atari, Brendan Kennedy, Gwenyth Portillo Wightman, Elaine Gonzalez, Natalie Delong, Rhea Bhatia, Arineh Mirinjian, Xiang Ren, Morteza Dehghani
| Challenge: | Existing estimates of hate crimes in the US are under-reported relative to actual number of incidents. |
| Approach: | They propose to use event extraction and multi-instance learning to predict hate crimes in local news articles for cities without official FBI reports. |
| Outcome: | The proposed model compares to FBI reports and shows that hate crimes are under-reported in local press. |
Multilingual Entity, Relation, Event and Human Value Extraction (N19-4)
Copied to clipboard
| Challenge: | Existing systems that extract knowledge elements from multiple languages and documents do not aggregate knowledge from multiple documents and languages. |
| Approach: | They propose a multilingual knowledge extraction system that performs entity discovery and linking, relation extraction, event extraction, and coreference. |
| Outcome: | The proposed system performs entity discovery and linking, relation extraction, event extraction, and coreference. |