Saved in:
Bibliographic Details
Main Authors: Jin, Ziqi, Wang, Lei, Luo, Ziwei, Sun, Aixin
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.07023
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911576154439680
author Jin, Ziqi
Wang, Lei
Luo, Ziwei
Sun, Aixin
author_facet Jin, Ziqi
Wang, Lei
Luo, Ziwei
Sun, Aixin
contents Autoregressive (AR) language models generate text one token at a time, even when consecutive tokens are highly predictable given earlier context. We introduce MARS (Mask AutoRegreSsion), a lightweight fine-tuning method that teaches an instruction-tuned AR model to predict multiple tokens per forward pass. MARS adds no architectural modifications, no extra parameters, and produces a single model that can still be called exactly like the original AR model with no performance degradation. Unlike speculative decoding, which maintains a separate draft model alongside the target, or multi-head approaches such as Medusa, which attach additional prediction heads, MARS requires only continued training on existing instruction data. When generating one token per forward pass, MARS matches or exceeds the AR baseline on six standard benchmarks. When allowed to accept multiple tokens per step, it maintains baseline-level accuracy while achieving 1.5-1.7x throughput. We further develop a block-level KV caching strategy for batch inference, achieving up to 1.71x wall-clock speedup over AR with KV cache on Qwen2.5-7B. Finally, MARS supports real-time speed adjustment via confidence thresholding: under high request load, the serving system can increase throughput on the fly without swapping models or restarting, providing a practical latency-quality knob for deployment.
format Preprint
id arxiv_https___arxiv_org_abs_2604_07023
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MARS: Enabling Autoregressive Models Multi-Token Generation
Jin, Ziqi
Wang, Lei
Luo, Ziwei
Sun, Aixin
Computation and Language
Autoregressive (AR) language models generate text one token at a time, even when consecutive tokens are highly predictable given earlier context. We introduce MARS (Mask AutoRegreSsion), a lightweight fine-tuning method that teaches an instruction-tuned AR model to predict multiple tokens per forward pass. MARS adds no architectural modifications, no extra parameters, and produces a single model that can still be called exactly like the original AR model with no performance degradation. Unlike speculative decoding, which maintains a separate draft model alongside the target, or multi-head approaches such as Medusa, which attach additional prediction heads, MARS requires only continued training on existing instruction data. When generating one token per forward pass, MARS matches or exceeds the AR baseline on six standard benchmarks. When allowed to accept multiple tokens per step, it maintains baseline-level accuracy while achieving 1.5-1.7x throughput. We further develop a block-level KV caching strategy for batch inference, achieving up to 1.71x wall-clock speedup over AR with KV cache on Qwen2.5-7B. Finally, MARS supports real-time speed adjustment via confidence thresholding: under high request load, the serving system can increase throughput on the fly without swapping models or restarting, providing a practical latency-quality knob for deployment.
title MARS: Enabling Autoregressive Models Multi-Token Generation
topic Computation and Language
url https://arxiv.org/abs/2604.07023