DIVERSED: Relaxed Speculative Decoding via Dynamic Ensemble Verification

arXiv:2604.07622v1 Announce Type: new Abstract: Speculative decoding is an effective technique for accelerating large language model inference by drafting multiple tokens in parallel. In practice, its speedup is often bottlenecked by a rigid verification step that strictly enforces the accepted token distribution to exactly match the target model. This constraint leads to the rejection of many plausible tokens, lowering the acceptance rate and limiting overall time speedup. To overcome this limitation, we propose Dynamic Verification Relaxed Speculative Decoding (DIVERSED), a relaxed verification framework that improves time efficiency while preserving generation quality. DIVERSED learns an ensemble-based verifier that blends the draft and target model distributions with a task-dependent and context-dependent weight. We provide theoretical justification for our approach and demonstrate empirically that DIVERSED achieves substantially higher inference efficiency compared to standard speculative decoding methods. Code is available at: https://github.com/comeusr/diversed.

Related

• See the Forest for the Trees: Loosely Speculative Decoding via Visual-Semantic Guidance for Efficient Inference of Video LLMs
• Double: Breaking the Acceleration Limit via Double Retrieval Speculative Parallelism
• LongSpec: Long-Context Lossless Speculative Decoding with Efficient Drafting and Verification
• FBS: Modeling Native Parallel Reading inside a Transformer

Source: arXiv cs.CL | 2026-04-10