VIBE: Voice-Induced open-ended Bias Evaluation for Large Audio-Language Models via Real-World Speech
Large Audio-Language Models exhibit systematic generative biases in realistic scenarios when evaluated through open-ended tasks using human-recorded speech, with bias magnitude var…
Hugging Face · Daily Papers
·Yi-Cheng Lin, Yusuke Hirota
·
Este artigo está em destaque na seleção diária de papers do Hugging Face, curada pela comunidade de pesquisa em IA.
Autores: Yi-Cheng Lin, Yusuke Hirota, Sung-Feng Huang, Hung-yi Lee
- 0 upvotes da comunidade
- Temas: Large Audio-Language Models, generative bias, open-ended tasks, human-recorded speech, stereotypical associations, distributional shifts
Resumo
Resumo original (em inglês), extraído do paper:
Large Audio-Language Models exhibit systematic generative biases in realistic scenarios when evaluated through open-ended tasks using human-recorded speech, with bias magnitude varying significantly by task and triggered by gender and accent cues.Onde ler
// relacionados
Leia também
Blog
Meta ran ads for an app promising to nudify female politicians
Editorial
MiniMax Music 3: uma canção inteira de cinco minutos, com pesos abertos
Blog
Emotion Across Speech and Faces: Shared Affective Mechanisms in Multimodal Foundation Models
Blog