Cover Image

Not-a-Bandit: Provably No-Regret Drafter Selection in Speculative Decoding for LLMs

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Liu, Hongyi, Huang, Jiaji, Jia, Zhen, Park, Youngsuk, Wang, Yu-Xiang
Format:	Preprint
Published:	2025
Subjects:	Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

ProxSparse: Regularized Learning of Semi-Structured Sparsity Masks for Pretrained LLMs
by: Liu, Hongyi, et al.
Published: (2025)

Steering Pretrained Drafters during Speculative Decoding
by: Berdoz, Frédéric, et al.
Published: (2025)

ParallelSpec: Parallel Drafter for Efficient Speculative Decoding
by: Xiao, Zilin, et al.
Published: (2024)

Recurrent Drafter for Fast Speculative Decoding in Large Language Models
by: Cheng, Yunfei, et al.
Published: (2024)

Coupling without Communication and Drafter-Invariant Speculative Decoding
by: Daliri, Majid, et al.
Published: (2024)

HiViS: Hiding Visual Tokens from the Drafter for Speculative Decoding in Vision-Language Models
by: Xie, Zhinan, et al.
Published: (2025)

OmniDraft: A Cross-vocabulary, Online Adaptive Drafter for On-device Speculative Decoding
by: Ramakrishnan, Ramchalam Kinattinkara, et al.
Published: (2025)

Training LLMs with MXFP4
by: Tseng, Albert, et al.
Published: (2025)

No-Regret Linear Bandits under Gap-Adjusted Misspecification
by: Liu, Chong, et al.
Published: (2025)

BanditSpec: Adaptive Speculative Decoding via Bandit Algorithms
by: Hou, Yunlong, et al.
Published: (2025)

TritonRL: Training LLMs to Think and Code Triton Without Cheating
by: Woo, Jiin, et al.
Published: (2025)

Dataset Watermarking for Closed LLMs with Provable Detection
by: Huang, Pengrun, et al.
Published: (2026)

How Does Variance Shape the Regret in Contextual Bandits?
by: Jia, Zeyu, et al.
Published: (2024)

CORAL: Learning Consistent Representations across Multi-step Training with Lighter Speculative Drafter
by: Weng, Yepeng, et al.
Published: (2025)

Scaling Laws Meet Model Architecture: Toward Inference-Efficient LLMs
by: Bian, Song, et al.
Published: (2025)

Mamba Drafters for Speculative Decoding
by: Choi, Daewon, et al.
Published: (2025)

Directional Alignment Mitigates Reward Hacking in Reinforcement Learning for Language Models
by: Deng, Wenlong, et al.
Published: (2026)

VOCABTRIM: Vocabulary Pruning for Efficient Speculative Decoding in LLMs
by: Goel, Raghavv, et al.
Published: (2025)

Fail Fast, Win Big: Rethinking the Drafting Strategy in Speculative Decoding via Diffusion LLMs
by: Pan, Rui, et al.
Published: (2025)

Stochastic Rounding for LLM Training: Theory and Practice
by: Ozkara, Kaan, et al.
Published: (2025)

Active Context Selection Improves Simple Regret in Contextual Bandits
by: Shahverdikondori, Mohammad, et al.
Published: (2026)

Speculative Speculative Decoding
by: Kumar, Tanishq, et al.
Published: (2026)

TapOut: A Bandit-Based Approach to Dynamic Speculative Decoding
by: Sridhar, Aditya, et al.
Published: (2025)

Near-Optimal Regret in Adversarial Kernel Bandits
by: Zhang, Yu-Jie, et al.
Published: (2026)

Diverse Randomized Value Functions: A Provably Pessimistic Approach for Offline Reinforcement Learning
by: Yu, Xudong, et al.
Published: (2024)

Non-Stationary Restless Multi-Armed Bandits with Provable Guarantee
by: Hung, Yu-Heng, et al.
Published: (2025)

Bayesian Regret Minimization in Offline Bandits
by: Petrik, Marek, et al.
Published: (2023)

Optimal Regret for Single Index Bandits
by: Dey, Devdan, et al.
Published: (2026)

No-Regret is not enough! Bandits with General Constraints through Adaptive Regret Minimization
by: Bernasconi, Martino, et al.
Published: (2024)

Decoding Speculative Decoding
by: Yan, Minghao, et al.
Published: (2024)

Calibrated Speculative Decoding: Frequency-Guided Candidate Selection for Efficient Inference
by: Zhou, Xuwen, et al.
Published: (2026)

Threshold-Based Optimal Arm Selection in Monotonic Bandits: Regret Lower Bounds and Algorithms
by: Varude, Chanakya, et al.
Published: (2025)

HiSpec: Hierarchical Speculative Decoding for LLMs
by: Kumar, Avinash, et al.
Published: (2025)

Task-Aware Calibration: Provably Optimal Decoding in LLMs
by: Tomov, Tim, et al.
Published: (2026)

Achieving Optimal Static and Dynamic Regret Simultaneously in Bandits with Deterministic Losses
by: Qian, Jian, et al.
Published: (2026)

Fast Best-in-Class Regret for Contextual Bandits
by: Girard, Samuel, et al.
Published: (2025)

Improved Regret Bounds for Bandits with Expert Advice
by: Cesa-Bianchi, Nicolò, et al.
Published: (2024)

Prior Diffusiveness and Regret in the Linear-Gaussian Bandit
by: Zhu, Yifan, et al.
Published: (2026)

Optimal Regret for Policy Optimization in Contextual Bandits
by: Levy, Orin, et al.
Published: (2026)

Efficient Swap Regret Minimization in Combinatorial Bandits
by: Kontogiannis, Andreas, et al.
Published: (2026)