PHOTON: Hierarchical Autoregressive Modeling for Lightspeed and Memory-Efficient Language Generation

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Ichikawa, Yuma, Takagi, Naoya, Nakagawa, Takumi, Kanazawa, Yuzi, Sakai, Akira
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866914239109660672
author Ichikawa, Yuma
Takagi, Naoya
Nakagawa, Takumi
Kanazawa, Yuzi
Sakai, Akira
author_facet Ichikawa, Yuma
Takagi, Naoya
Nakagawa, Takumi
Kanazawa, Yuzi
Sakai, Akira
contents Transformers operate as horizontal token-by-token scanners; at each generation step, attending to an ever-growing sequence of token-level states. This access pattern increases prefill latency and makes long-context decoding more memory-bound, as KV-cache reads and writes dominate inference time over arithmetic operations. We propose Parallel Hierarchical Operation for TOp-down Networks (PHOTON), a hierarchical autoregressive model that replaces horizontal scanning with vertical, multi-resolution context scanning. PHOTON maintains a hierarchy of latent streams: a bottom-up encoder compresses tokens into low-rate contextual states, while lightweight top-down decoders reconstruct fine-grained token representations in parallel. We further introduce recursive generation that updates only the coarsest latent stream and eliminates bottom-up re-encoding. Experimental results show that PHOTON is superior to competitive Transformer-based language models regarding the throughput-quality trade-off, providing advantages in long-context and multi-query tasks. In particular, this reduces decode-time KV-cache traffic, yielding up to $10^{3}\times$ higher throughput per unit memory.
format Preprint
id arxiv_https___arxiv_org_abs_2512_20687
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle PHOTON: Hierarchical Autoregressive Modeling for Lightspeed and Memory-Efficient Language Generation
Ichikawa, Yuma
Takagi, Naoya
Nakagawa, Takumi
Kanazawa, Yuzi
Sakai, Akira
Machine Learning
Artificial Intelligence
Computation and Language
Distributed, Parallel, and Cluster Computing
Transformers operate as horizontal token-by-token scanners; at each generation step, attending to an ever-growing sequence of token-level states. This access pattern increases prefill latency and makes long-context decoding more memory-bound, as KV-cache reads and writes dominate inference time over arithmetic operations. We propose Parallel Hierarchical Operation for TOp-down Networks (PHOTON), a hierarchical autoregressive model that replaces horizontal scanning with vertical, multi-resolution context scanning. PHOTON maintains a hierarchy of latent streams: a bottom-up encoder compresses tokens into low-rate contextual states, while lightweight top-down decoders reconstruct fine-grained token representations in parallel. We further introduce recursive generation that updates only the coarsest latent stream and eliminates bottom-up re-encoding. Experimental results show that PHOTON is superior to competitive Transformer-based language models regarding the throughput-quality trade-off, providing advantages in long-context and multi-query tasks. In particular, this reduces decode-time KV-cache traffic, yielding up to $10^{3}\times$ higher throughput per unit memory.
title PHOTON: Hierarchical Autoregressive Modeling for Lightspeed and Memory-Efficient Language Generation
topic Machine Learning
Artificial Intelligence
Computation and Language
Distributed, Parallel, and Cluster Computing
url https://arxiv.org/abs/2512.20687