Saved in:
Bibliographic Details
Main Authors: Johnson, Benjamin K., Goralski, Thomas, Semwal, Ayush, Shen, Hui, Jang, H. Josh
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.18780
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913048970657792
author Johnson, Benjamin K.
Goralski, Thomas
Semwal, Ayush
Shen, Hui
Jang, H. Josh
author_facet Johnson, Benjamin K.
Goralski, Thomas
Semwal, Ayush
Shen, Hui
Jang, H. Josh
contents Semi-Markov Conditional Random Fields (semi-CRFs) assign labels to segments of a sequence rather than to individual positions, enabling exact inference over segment-level features and principled uncertainty estimates at their boundaries. However, existing implementations must materialize a large edge potential tensor whose size grows with sequence length, maximum segment length, and label count, becoming prohibitive for speech-scale state spaces and intractable at genomic scales where sequences can exceed 100,000 positions. This memory bottleneck has limited the adoption of exact segment-level inference for long sequences and large label sets. We identify that the core inefficiency is materializing edge potentials that can instead be evaluated on-the-fly from a compact prefix-sum array, and make several improvements. First, replacing the stored edge tensor with prefix-sum lookup reduces the memory footprint by a factor proportional to the product of segment length and label count. Second, a streaming forward-backward pass with checkpoint-boundary normalization keeps working memory sublinear in sequence length while preserving exact gradients. Third, zero-centered cumulative scores control numerical drift and induce an adaptive duration prior under label imbalance. We integrate these ideas into Flash-SemiCRF, a fused Triton kernel that enables exact semi-CRF inference on previously intractable problem sizes. Available at https://github.com/biobenkj/flash-semicrf.
format Preprint
id arxiv_https___arxiv_org_abs_2604_18780
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Streaming Structured Inference with Flash-SemiCRF
Johnson, Benjamin K.
Goralski, Thomas
Semwal, Ayush
Shen, Hui
Jang, H. Josh
Machine Learning
Semi-Markov Conditional Random Fields (semi-CRFs) assign labels to segments of a sequence rather than to individual positions, enabling exact inference over segment-level features and principled uncertainty estimates at their boundaries. However, existing implementations must materialize a large edge potential tensor whose size grows with sequence length, maximum segment length, and label count, becoming prohibitive for speech-scale state spaces and intractable at genomic scales where sequences can exceed 100,000 positions. This memory bottleneck has limited the adoption of exact segment-level inference for long sequences and large label sets. We identify that the core inefficiency is materializing edge potentials that can instead be evaluated on-the-fly from a compact prefix-sum array, and make several improvements. First, replacing the stored edge tensor with prefix-sum lookup reduces the memory footprint by a factor proportional to the product of segment length and label count. Second, a streaming forward-backward pass with checkpoint-boundary normalization keeps working memory sublinear in sequence length while preserving exact gradients. Third, zero-centered cumulative scores control numerical drift and induce an adaptive duration prior under label imbalance. We integrate these ideas into Flash-SemiCRF, a fused Triton kernel that enables exact semi-CRF inference on previously intractable problem sizes. Available at https://github.com/biobenkj/flash-semicrf.
title Streaming Structured Inference with Flash-SemiCRF
topic Machine Learning
url https://arxiv.org/abs/2604.18780