Searcharxiv⌕ Search

arXiv subjects

Akram Elbouanani

Publications and source records attributed to Akram Elbouanani.

5 recordsLinked to original sources

Portrait Fitness as a Pretext: Probing Implicit Demographic Bias in VLMs

We audit face-related bias in VLMs using BALPORTRAITS, a new balanced dataset of 9,000 fictional portraits across ethnicity, gender, and age segments. Sixteen VLMs estimate portrait fitness scores for occupations, brands, activities, and banking or insurance products within an advertising-casting framework designed to avoid direct inference of protected attributes. Item-level estimations are aggregated with scenario-specific prestige scores to compare rankings across demographic groups. While the magnitude and direction of these biases vary across scenarios, model families, and demographic dimensions, with age, ethnicity, and gender contributing differently depending on the task, we find demographic dispersion across all evaluated models, with a majority of pairwise segment differences reaching statistical significance. This study introduces a controlled protocol to assess implicit face-related associations and demonstrates that single-scenario or single-attribute audits are insufficient.

cs.CV↗

User Attributes and Political Ideology Shape LLM Receptiveness in Online Political Debates

Online political discussions increasingly involve AI-mediated participation, as users invoke LLMs to assess contested claims, arbitrate disputes, and respond to opposing viewpoints. As these systems become participants in online exchanges, they may alter which speakers and argumentative strategies receive validation within the discussion. We study this interactional dimension by reconstructing 5,376 Debate.org debates as LLM conversations and running nearly 150,000 controlled multi-turn conversations that vary user political orientation, expertise, rhetorical strategy, and defended position, including conversations with competing users. Across naturalistic and controlled debates, we find that LLMs are systematically more receptive to left-leaning than right-leaning users. The asymmetry persists when users with different political orientations defend the same position and when opposing users appear in the same conversation, indicating that it is not fully explained by stance agreement or simple accommodation to the current speaker. Repeated interaction increases receptiveness without eliminating this ordering; expertise also matters, while rhetorical differences are smaller. The pattern appears across multiple models, but is not fixed: fine-tuning on different political corpora can largely erase or reverse it. These results show that LLMs can respond systematically differently to participants in the same political debate, making such interactional asymmetries important for understanding their emerging role in online public discourse.

cs.AI↗

A Scalable Entity-Based Framework for Auditing Bias in Large Language Models

Existing approaches to bias evaluation in large language models (LLMs) trade ecological validity for statistical control, relying either on artificial prompts that poorly reflect real-world use or on naturalistic tasks that lack scale and rigor. We introduce a scalable bias-auditing framework that uses named entities as controlled probes to measure systematic disparities in model behavior. Synthetic data enables us to construct diverse, controlled inputs, and we show that it reliably reproduces bias patterns observed in natural text, supporting its use for large-scale analysis. Using this framework, we conduct the largest bias audit to date, comprising 1.9 billion data points across multiple entity types, tasks, languages, models, and prompting strategies. We find consistent patterns: models penalize right-wing politicians and favor left-wing politicians, prefer Western and wealthier countries over the Global South, favor Western companies, and penalize firms in the defense and pharmaceutical sectors. While instruction tuning reduces bias, increasing model scale amplifies it, and prompting in Chinese or Russian does not mitigate Western-aligned preferences. These findings highlight the need for systematic bias auditing before deploying LLMs in high-stakes applications. Our framework is extensible to other domains and tasks, and we make it publicly available to support future work.

cs.CL↗

CEA-LIST at CheckThat! 2025: Evaluating LLMs as Detectors of Bias and Opinion in Text

This paper presents a competitive approach to multilingual subjectivity detection using large language models (LLMs) with few-shot prompting. We participated in Task 1: Subjectivity of the CheckThat! 2025 evaluation campaign. We show that LLMs, when paired with carefully designed prompts, can match or outperform fine-tuned smaller language models (SLMs), particularly in noisy or low-quality data settings. Despite experimenting with advanced prompt engineering techniques, such as debating LLMs and various example selection strategies, we found limited benefit beyond well-crafted standard few-shot prompts. Our system achieved top rankings across multiple languages in the CheckThat! 2025 subjectivity detection task, including first place in Arabic and Polish, and top-four finishes in Italian, English, German, and multilingual tracks. Notably, our method proved especially robust on the Arabic dataset, likely due to its resilience to annotation inconsistencies. These findings highlight the effectiveness and adaptability of LLM-based few-shot learning for multilingual sentiment tasks, offering a strong alternative to traditional fine-tuning, particularly when labeled data is scarce or inconsistent.

cs.CL↗

Analyzing Political Bias in LLMs via Target-Oriented Sentiment Classification

Political biases encoded by LLMs might have detrimental effects on downstream applications. Existing bias analysis methods rely on small-size intermediate tasks (questionnaire answering or political content generation) and rely on the LLMs themselves for analysis, thus propagating bias. We propose a new approach leveraging the observation that LLM sentiment predictions vary with the target entity in the same sentence. We define an entropy-based inconsistency metric to encode this prediction variability. We insert 1319 demographically and politically diverse politician names in 450 political sentences and predict target-oriented sentiment using seven models in six widely spoken languages. We observe inconsistencies in all tested combinations and aggregate them in a statistically robust analysis at different granularity levels. We observe positive and negative bias toward left and far-right politicians and positive correlations between politicians with similar alignment. Bias intensity is higher for Western languages than for others. Larger models exhibit stronger and more consistent biases and reduce discrepancies between similar languages. We partially mitigate LLM unreliability in target-oriented sentiment classification (TSC) by replacing politician names with fictional but plausible counterparts.

cs.CL↗