Multimodal Continuous Reasoning via Asymmetric Mutual Variational Learning

Multimodal Continuous Reasoning via Asymmetric Mutual Variational Learning

Asymmetric Mutual Variational Learning addresses train-inference mismatch in multimodal reasoning by using bidirectional calibration to prevent answer leakage and improve latent-sp…

Hugging Face · Daily Papers ·Shijie Li, Yilin Gao · ·▲ 12 upvotes

Este artigo está em destaque na seleção diária de papers do Hugging Face, curada pela comunidade de pesquisa em IA.

Autores: Shijie Li, Yilin Gao, Siyuan Yang, Tieyuan Chen, Chaofan Gan, Zhihao He

  • 12 upvotes da comunidade
  • Temas: Multimodal Large Language Models, language-space bottleneck, continuous latent reasoning, train-inference mismatch, variational training, posterior

Resumo

Resumo original (em inglês), extraído do paper:

Asymmetric Mutual Variational Learning addresses train-inference mismatch in multimodal reasoning by using bidirectional calibration to prevent answer leakage and improve latent-space stability.

Onde ler

compartilhar: