Saved in:
Bibliographic Details
Main Authors: Feil, Michael, Lipp, Julius
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.15013
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915744173785088
author Feil, Michael
Lipp, Julius
author_facet Feil, Michael
Lipp, Julius
contents Batch inference workloads for causal transformer models frequently process sequences that share common prefixes, such as system prompts, few-shot examples, or shared queries. Standard inference engines treat each sequence independently, redundantly recomputing identical MLP activations for every copy of the shared prefix. We introduce RadixMLP, a technique that exploits the position-wise nature of MLPs, LayerNorms, linear projections, and embeddings to eliminate this redundancy. RadixMLP dynamically maps batches to a prefix trie, gathering shared segments into a compressed representation for position-wise computation and scattering results back only at attention boundaries. RadixMLP is stateless and operates within a single forward pass. In end-to-end serving benchmarks on MS~MARCO v1.1 with Qwen3 models (0.6B to 8B parameters), RadixMLP achieves 1.44-1.59$\times$ speedups in realistic reranking workloads, with up to $5\times$ speedups on synthetic benchmarks with longer shared prefixes. Our code is available at https://github.com/michaelfeil/radix-mlp.
format Preprint
id arxiv_https___arxiv_org_abs_2601_15013
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle RadixMLP -- Intra-batch Deduplication for Causal Transformers
Feil, Michael
Lipp, Julius
Machine Learning
Distributed, Parallel, and Cluster Computing
Batch inference workloads for causal transformer models frequently process sequences that share common prefixes, such as system prompts, few-shot examples, or shared queries. Standard inference engines treat each sequence independently, redundantly recomputing identical MLP activations for every copy of the shared prefix. We introduce RadixMLP, a technique that exploits the position-wise nature of MLPs, LayerNorms, linear projections, and embeddings to eliminate this redundancy. RadixMLP dynamically maps batches to a prefix trie, gathering shared segments into a compressed representation for position-wise computation and scattering results back only at attention boundaries. RadixMLP is stateless and operates within a single forward pass. In end-to-end serving benchmarks on MS~MARCO v1.1 with Qwen3 models (0.6B to 8B parameters), RadixMLP achieves 1.44-1.59$\times$ speedups in realistic reranking workloads, with up to $5\times$ speedups on synthetic benchmarks with longer shared prefixes. Our code is available at https://github.com/michaelfeil/radix-mlp.
title RadixMLP -- Intra-batch Deduplication for Causal Transformers
topic Machine Learning
Distributed, Parallel, and Cluster Computing
url https://arxiv.org/abs/2601.15013