The paper introduces a unified framework that simultaneously measures intrinsic (encoded) and extrinsic (expressed) gender bias in large language models using identical neutral prompts. It finds a consistent link between latent gender information and output bias, but shows that alignment via supervised fine‑tuning reduces expressed bias while leaving internal gender associations largely intact and reactivatable by adversarial prompts. The study also demonstrates that debiasing gains on structured benchmarks may not transfer to realistic tasks such as story generation.
By Nour Bouchouchi, Thibault Laugel, Xavier Renard, Christophe Marsala, Marie-Jeanne Lesot, Marcin Detyniecki
Humans possess an innate ability to understand fine-grained interpersonal relationships, which is central to everyday social interactions. Although such reasoning is inherently multimodal, it remains largely unexplored by existing multimodal large language models (MLLMs).
PIVOTSBench is a benchmark designed to assess multimodal large language models’ ability to reason about fine‑grained interpersonal relationships. It is constructed from Social‑IQ 2.0 and YouTube data and evaluates models on predicting bidirectional relationship dimensions grounded in psychology research. The benchmark also includes auxiliary tasks that test models’ capacity to identify and use critical visual cues, and it examines the impact of visual modalities, social role information, and different prediction settings on model performance.
By Shuxiang Zhang, Yiting Yin, Wenxuan Song, Yuhang Wu, Miao Liu
arXiv:2606. 07969v1 Announce Type: cross Abstract: Gender bias in AI-generated stories is a well-documented problem.
By Imani Finkley, Yuanxi Li, Melanie Walsh
arXiv:2603. 23485v2 Announce Type: replace-cross Abstract: Standard evaluation practices assume that large language model (LLM) outputs are stable when prompts are embedded in contextually equivalent discourses.
By Sagar Kumar, Ariel Flint, Luca Maria Aiello, Andrea Baronchelli
The paper investigates how in‑context learning (ICL) in large vision‑language models (LVLMs) can amplify gender bias. Using the VL‑BICLE framework, the authors show that gendered ICL demonstrations shift model bias toward the demonstrated gender, especially in tasks involving gendered language such as image captioning and pronoun prediction. They find that similarity‑based retrieval does not mitigate this bias and that replacing real images with synthetic ones from stable diffusion reduces bias without hurting caption quality.
By Tong Xiang, Noa Garcia, Yuta Nakashima
arXiv:2605.31556v2 Announce Type: replace-cross
Abstract: Alignment teaches vision-language models (VLMs) to avoid expressing demographic biases, and when gender is clearly visible they largely succe...
By Arnau Marin-Llobet, Simon Henniger, Mahzarin R. Banaji
arXiv:2601. 14569v2 Announce Type: replace-cross Abstract: Social understanding abilities are crucial for multimodal large language models (MLLMs) to interpret human social interactions.
By Leena Mathur, Bhaavanaa Thumu, Youssouf Kebe, Louis-Philippe Morency
The paper introduces CGER-Net, a context‑grounded multimodal framework designed to reduce biased gendered economic role attribution in image‑text memes. By assessing whether sufficient evidence exists for gendered reasoning, the model gates inference, abstaining when cues are ambiguous and confidently attributing roles when explicit. Evaluation on the EconMeme‑GE dataset shows CGER‑Net cuts gender overcommitment by up to 44% on ambiguous memes while preserving accuracy on clear cases, and human judges find 79% of its rationales epistemically aligned with the evidence.
By Kushal Kanwar, Dushyant Singh Chauhan, Kapil Rana, Gopendra Vikram Singh, Nils Lukas
arXiv:2512. 00807v2 Announce Type: replace Abstract: Vision-Language Models (VLMs) inherit significant social biases from their training data, notably in gender representation.
By Yujie Lin, Jiayao Ma, Qingguo Hu, Wenbo Li, Genji Li, Derek Wong, Jinsong Su
The paper introduces ViD, a vision‑dominant gender bias mitigation framework for large vision‑language models. ViD uses causal analysis of attention patterns and dual mechanisms—backdoor adjustment and refined token selection—to suppress bias while preserving reasoning and generation quality. Experiments show a 14.7% reduction in gender bias on FACET and significant improvements on MS COCO image captioning, all without extra training overhead.
By Zhipeng Zhao, Zhaoqiang Wei, Peishun Liu, Youwei Zhao, Ruichun Tang
Large language models (LLMs) are increasingly embedded in everyday life and widely used for information seeking, raising concerns about their potential to perpetuate social biases and reinforce stereotypes. In this study, we investigate gender bias in LLMs through the lens of their associations with musical instruments.