PACER: Blockwise Pre-verification for Speculative Decoding with Adaptive Length
- 1 January 2026
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Audio, Speech and Language Processing
- Vol. 34, 1-15
Abstract
No abstract availableAll Related Versions
Funding Information
- China NSFC projects (62576212)
- China NSFC projects (62120106006)
- China NSFC projects (92370206)
- China NSFC projects (U23B2057)
- Frontier Technologies R&D Program of Jiangsu (BF2025029)
This publication has 25 references indexed in Scilit:
- DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learningNature, 2025
- Detecting Hallucinations in Large Language Model Generation: A Token Probability Approach2025
- OPT-Tree: Speculative Decoding with Adaptive Draft Tree StructureTransactions of the Association for Computational Linguistics, 2025
- SpecInfer: Accelerating Large Language Model Serving with Tree-based Speculative Inference and Verification2024
- EAGLE-2: Faster Inference of Language Models with Dynamic Draft Trees2024
- Draft& Verify: Lossless Large Language Model Acceleration via Self-Speculative Decoding2024
- LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding2024
- Efficient Memory Management for Large Language Model Serving with PagedAttention2023
- Speculative Decoding: Exploiting Speculative Execution for Accelerating Seq2seq Generation2023
- Accelerating Transformer Inference for Translation via Parallel Decoding2023