While a few high-quality bias benchmark datasets exist to address stereotypes in Language Models (LMs), a notable lack of focus remains on body image stereotypes. To bridge this gap, we propose BIStereo, a suite to uncover LMs’ biases towards people of certain physical appearance characteristics, namely, skin complexion, body shape, height, attire, and a miscellaneous category including hair texture, eye color, and more. Our dataset comprises 40k sentence pairs designed to assess LMs’ biased preference for certain body types. We further include 60k premise-hypothesis pairs designed to comprehensively assess LMs’ preference for fair skin tone. Additionally, we curate 553 tuples consisting of a body image descriptor, gender, and a stereotypical attribute, validated by a diverse pool of annotators for physical appearance stereotypes.We propose a metric, TriSentBias, that captures the biased preferences of LMs towards a certain body type over others. Using BIStereo, we assess the presence of body image biases in ten different language models, revealing significant biases in models Muril, XLMR, Llama3, and Gemma. We further evaluate the LMs through downstream NLI and Analogy tasks.Our NLI experiments highlight notable patterns in the LMs that align with the well-documented cognitive bias in humans known as the Halo Effect.

Large Language Models (LLMs) have achieved remarkable success on question answering (QA) tasks, yet they often encode harmful biases that compromise fairness and trustworthiness. Most existing bias mitigation approaches are restricted to predefined categories, limiting their ability to address novel or context-specific emergent biases. To bridge this gap, we tackle the novel problem of open-set bias detection and mitigation in text-based QA. We introduce _OpenBiasBench_, a comprehensive benchmark designed to evaluate biases across a wide range of categories and subgroups, encompassing both known and previously unseen biases. Additionally, we propose _Open-DeBias_, a novel, data-efficient, and parameter-efficient debiasing method that leverages adapter modules to mitigate existing social and stereotypical biases while generalizing to unseen ones. Compared to the state-of-the-art BMBI method, Open-DeBias improves QA accuracy on BBQ dataset by nearly **48%** on ambiguous subsets and **6%** on disambiguated ones, using adapters fine-tuned on just a small fraction of the training data. Remarkably, the same adapters, in a zero-shot transfer to Korean BBQ, achieve **84% accuracy**, demonstrating robust language-agnostic generalization. Through extensive evaluation, we also validate the effectiveness of Open-DeBias across a broad range of NLP tasks, including StereoSet and CrowS-Pairs, highlighting its robustness, multilingual strength, and suitability for general-purpose, open-domain bias mitigation. The project page is available at: [https://sites.google.com/view/open-debias25](https://sites.google.com/view/open-debias25)

With an evergrowing number of LLMs reporting superlative performance for English, their ability to perform equitably for different dialects of English (i.e., dialect robustness) needs to be ascertained. Specifically, we use English language (US English or Indian English) conversations between humans who play the word-guessing game of ‘taboo‘. We formulate two evaluative tasks: target word prediction (TWP) (i.e., predict the masked target word in a conversation) and target word selection (TWS) (i.e., select the most likely masked target word in a conversation, from among a set of candidate words). Extending MD3, an existing dialectic dataset of taboo-playing conversations, we introduce M-MD3, a target-word-masked version of MD3 with the en-US and en-IN subsets. We create two subsets: en-MV (where en-US is transformed to include dialectal information) and en-TR (where dialectal information is removed from en-IN). We evaluate three multilingual LLMs–one open source (Llama3) and two closed-source (GPT-4/3.5). LLMs perform significantly better for US English than Indian English for both TWP and TWS tasks, for all settings, exhibiting marginalisation against the Indian dialect of English. While GPT-based models perform the best, the comparatively smaller models work more equitably after fine-tuning. Our evaluation methodology exhibits a novel and reproducible way to examine attributes of language models using pre-existing dialogue datasets with language varieties. Dialect being an artifact of one’s culture, this paper demonstrates the gap in the performance of multilingual LLMs for communities that do not use a mainstream dialect.

In the landscape of natural language processing (NLP), addressing the challenges of bias and hallucination is paramount to ensuring the ethical and unbiased development of Large Language Models (LLMs). This tutorial delves into the intricate dimensions of LLMs, shedding light on the critical importance of understanding and mitigating the profound impacts of bias and hallucination. Divided into two parts, the first part delves deep into the complexity of bias propagation in LLM development, where we dissect its origins and far-reaching impacts. We then present innovative methodologies for mitigating diverse forms of bias, including dynamic word embeddings and robust benchmarking strategies. The second part of the tutorial discusses hallucination - a prevalent issue in generative AI systems such as LLMs. Through advanced data-driven techniques, we decode its intricate effects and complexities, followed factually-driven mitigation strategies. Furthermore, we shed light on the pivotal role of human cognitive behavior in the context of hallucination, drawing insights from cognitive data, including human eye-tracking data. Ultimately, this cutting-edge tutorial serves as a guiding light, equipping participants with indispensable tools and insights to navigate the ethical complexities of LLMs, thus paving the way for the development of unbiased and ethically robust NLP systems.