technique

LALS (Latent Association Leaning Score)

techniqueactiveprovisionallals-latent-association-leaning-score--dc45232f·1 events·first seen 16d ago

Aliases: LALS (Latent Association Leaning Score)

Co-occurring entities

gender bias in VLMs Vision-Language Models visual-token activation probing post-training alignment occupation-gender stereotype dataset

More like this (12)

Latent Adversarial Robustification (LAR)Latent Context Language Models Learning with Opponent-Learning Awareness (LOLA)Graph-LLaDA LoRA (Low-Rank Adaptation)Multi-head Latent Attention (MLA)Latent Consistency Models latent dynamical systems Locality-Sensitive Hashing Attention Operator Learning Latent World Recovery for Multimodal Learning with Missing Modalities latent reasoning

Recent events (1)

6arXiv · cs.CL·16d ago·source ↗

Vision-Language Models Suppress Female Representations Under Ambiguous Input

This paper investigates gender bias in vision-language models (VLMs) when inputs are ambiguous (e.g., workers in full gear or seen from behind), finding that models default to male outputs even for strongly female-stereotyped occupations. The authors introduce LALS (Latent Association Leaning Score), a zero-shot metric that probes internal visual-token activations to measure concept associations across layers. Across 15 occupations, 800+ ambiguous images, and four VLMs, they find a systematic decoupling: models internally encode female associations but suppress them before generation, with male signals amplifying end-to-end while female signals peak mid-network and are filtered out. Cultural visual cues like clothing color further modulate these internal associations.

Evaluation and Benchmarking AI Safety Research gender bias in VLMs Vision-Language Models visual-token activation probing +5 more