Speculative Decoding Reimagined for Multimodal Large Language Models
cs.CV, cs.AI
Submitted: 2025-05-20
Updated: 2026-11-09
Code: https://github.com/LynLucy/MSD
Terminology
Sources
- MagicDec: Breaking the Latency-Throughput Tradeoff for Long Context Generation with Speculative Decoding
- Evaluating Large Language Models Trained on Code
- Microsoft COCO Captions: Data Collection and Evaluation Server
- Training Verifiers to Solve Math Word Problems
- Break the Sequential Dependency of LLM Inference Using Lookahead Decoding
- MoE-LLaVA: Mixture of Experts for Large Vision-Language Models
- Q-VLM: Post-training Quantization for Large Vision-Language Models
- TinyLLaVA: A Framework of Small-scale Large Multimodal Models
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs