DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation

2026-07-06

AI Summary

This work addresses key limitations in speculative decoding, where parallel draft generation suffers from rapidly decaying acceptance rates due to the lack of inter-token dependencies, and fixed-length verification wastes batch capacity, hindering high-throughput inference. The authors propose DSpark, a novel framework that employs a semi-autoregressive drafter to model local dependencies within drafts, mitigating suffix decay. Furthermore, DSpark introduces the first dynamic verification scheduling mechanism, which adaptively adjusts verification length based on prefix survival probability and engine throughput characteristics. Experiments demonstrate that DSpark substantially increases offline accepted sequence lengths; deployed in the DeepSeek-V4 production system, it achieves a 60%–85% improvement in user-side generation speed, attaining previously unattainable performance levels under stringent interactive latency constraints.

Abstract

Speculative decoding accelerates Large Language Model (LLM) inference by decoupling draft generation from target verification. While recent parallel drafters efficiently propose long token sequences in a single forward pass, they suffer from rapid acceptance decay due to a lack of inter-token dependencies. Furthermore, indiscriminately verifying these extended blocks wastes critical batch capacity on tokens with high rejection risks, severely degrading throughput in high-concurrency serving systems. We introduce DSpark, a speculative decoding framework that unifies high-throughput parallel generation with adaptive, load-aware verification. To maintain draft quality, DSpark utilizes a semi-autoregressive architecture, coupling a parallel backbone with a lightweight sequential module, to introduce intra-block dependency modeling and mitigate suffix decay. To optimize system efficiency, DSpark employs confidence-scheduled verification, dynamically tailoring the verification length for each request based on estimated prefix survival probabilities and engine-specific throughput profiles. On offline benchmarks across diverse domains, DSpark substantially improves the accepted length over state-of-the-art autoregressive and parallel drafters. When deployed within the DeepSeek-V4 serving system under live user traffic, DSpark successfully mitigates verification waste. Compared to the established production baseline (MTP-1), DSpark accelerates per-user generation speeds by 60 to 85 percent at matched throughput levels. More importantly, by preventing severe throughput degradation under strict interactivity constraints, it enables performance tiers that were previously unattainable, shifting the Pareto frontier of our serving system.

Problem

• speculative decoding
• parallel drafting
• acceptance decay
• verification waste
• throughput degradation

Innovation

• speculative decoding
• semi-autoregressive generation
• confidence-scheduled verification
• throughput optimization
• large language model inference