machine-learning-models

2 items

RESEARCHarXiv CS.LG·4/21/2026

SetFlow: Generating Structured Sets of Representations for Multiple Instance Learning

This work introduces SetFlow, a generative architecture that models entire Multiple Instance Learning (MIL) bags directly in the representation space. It leverages the flow matching paradigm and a Set Transformer-inspired design to capture intra-bag dependencies and generate coherent, semantically consistent representations.

machine-learning-models Multiple Instance Learning representation learning Generative AI

RESEARCHarXiv CS.LG·4/24/2026

Frequency-Forcing: From Scaling-as-Time to Soft Frequency Guidance

This research focuses on improving image generation in flow-matching models by incorporating an explicit generation order, from coarse to fine detail. It compares K-Flow and Latent Forcing, proposing a new compatible route for scale-ordered generation via an auxiliary stream.

neural networks image generation machine-learning-models Generative AI