analiza aiqbench
This research introduces OnlineSPEC, a framework that treats the feedback loop inherent in speculative decoding as an opportunity for continuous model adaptation using online learning techniques. By dynamically optimizing draft models based on verification errors, the approach achieves up to 24% faster inference speeds across multiple benchmarks without requiring additional computational overhead. The work bridges theoretical concepts like dynamic regret minimization with practical acceleration strategies for large language model deployment.
When Drafts Evolve: Speculative Decoding Meets Online Learning
arXiv:2603.12617v2 Announce Type: replace Abstract: Speculative decoding has emerged as a widely adopted paradigm for accelerating large language model inference, where a lightweight draft model rapidly generates candidate tokens that are then verified in parallel by a larger target model. However, due to limited model capacity, drafts often struggle to approximate the target distribution, resulting in shorter acceptance lengths and diminished speedup. A key yet under-explored observation is that speculative decoding inherently provides verification feedback that quantifies the deviation between the draft and target models at no additional cost. This process naturally forms an iterative "draft commits-feedback provides-draft adapts" evolving loop, which precisely matches the online learning paradigm. Motivated by this connection, we propose OnlineSPEC, a unified framework that systematically leverages interactive feedback to continuously evolve draft models. Grounded in dynamic regret minimization, we establish a formal link between online learning performance and speculative system's acceleration rate, and develop novel algorithms via modern online learning techniques, including optimistic online learning that adaptively reuses historical gradients as predictive update hints, and online ensemble learning that dynamically maintains multiple draft models. Our algorithms are equipped with theoretical justifications and improved acceleration rates, achieving up to 24% speedup over seven benchmarks and five foundation models.
Abstract:Speculative decoding has emerged as a widely adopted paradigm for accelerating large language model inference, where a lightweight draft model rapidly generates candidate tokens that are then verified in parallel by a larger target model. However, due to limited model capacity, drafts often struggle to approximate the target distribution, resulting in shorter acceptance lengths and diminished speedup. A key yet under-explored observation is that speculative decoding inherently provides verification feedback that quantifies the deviation between the draft and target models at no additional cost. This process naturally forms an iterative "draft commits-feedback provides-draft adapts" evolving loop, which precisely matches the online learning paradigm. Motivated by this connection, we propose OnlineSPEC, a unified framework that systematically leverages interactive feedback to continuously evolve draft models. Grounded in dynamic regret minimization, we establish a formal link between online learning performance and speculative system's acceleration rate, and develop novel algorithms via modern online learning techniques, including optimistic online learning that adaptively reuses historical gradients as predictive update hints, and online ensemble learning that dynamically maintains multiple draft models. Our algorithms are equipped with theoretical justifications and improved acceleration rates, achieving up to 24% speedup over seven benchmarks and five foundation models.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2603.12617 [cs.LG]
(or arXiv:2603.12617v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2603.12617 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 292 cuvinte
Subiecte conexe
EfficientRollout: System-Aware Self-Speculative Decoding for RL Rollouts
arXiv · 2 oct. 2026
Microsoft’s AI chief says superintelligence is near, but won’t take your job
The Verge · 8 iun. 2026
TIDE: Temporal Incremental Draft Engine for Self-Improving LLM Inference
arXiv · 26 sept. 2026
Optimizing ATOM and vLLM-ATOM for High-Interactivity Inference
AMD · 1 sept. 2026