@alec_helbling
Interpretability, Multimodality, Diffusion. PhDing @GeorgiaTech. NSF Fellow. Prev intern @Apple, @Adobe, @NASAJPL.
AI
A post says a growing body of work uses continuous flow models for discrete problems, including language generation.
AI
A researcher describes work that turns a flow model into a recurrent model for tasks such as Sudoku. Predictions are fed back as inputs and refined on successive passes.
AI
The post describes a block-by-block approach: decode from left to right while generating the tokens within each block in parallel.
AI
When a sample is mostly noise, the model’s best guess is close to the mean, the post says—even when no data exists at that center.
AI
ML researcher Alec Helbling presents recurrent flow-based models for structured reasoning tasks like Sudoku.