Synthetic Perception: Can Generated Images Unlock Latent Visual Prior for Text-Centric Reasoning?
合成感知:生成图像能否解锁潜在的视觉先验以用于以文本为中心的推理?
AI总结 本文探讨生成图像能否解锁潜在视觉先验以提升文本中心推理,通过多模态融合架构和提示工程策略实现性能提升。
Comments Accepted as a poster at the International Conference on Machine Learning (ICML 2025) NewInML Workshop