26 upvotes
YANAMI releases multimodal chat dataset for stateful AI interaction
What is YANAMI: A Large-Scale Multimodal Dataset and Benchmark for Multiturn Stateful Interaction?
YANAMI is a large-scale dataset that pairs images with multi-turn conversational histories, requiring models to remember and act on earlier turns. It also includes a benchmark that scores how well unified multimodal models can understand, generate, and edit visual content across a dialogue.
Who it is for
Anyone building or evaluating chat-style vision-language models can use YANAMI to train and test true stateful interaction.