REAL Loss: Optimizing Real Expected Acceptance Length for Speculative Decoding
cs.LG, cs.CL
Submitted: 2026-11-10
Updated: 2026-11-10
Code: https://github.com/King-ty/speculators
Terminology
Sources
- PARD-2: Target-Aligned Parallel Draft Model for Dual-Mode Speculative Decoding
- Program Synthesis with Large Language Models
- Accelerating Large Language Model Decoding with Speculative Sampling
- Evaluating Large Language Models Trained on Code
- DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation
- Training Verifiers to Solve Math Word Problems
- Direct Alignment of Draft Model for Speculative Decoding with Chat-Fine-Tuned LLMs
- P-EAGLE: Parallel-Drafting EAGLE with Scalable Training
- DBLAST: Dependent Block Drafting for Stochastic Speculative Decoding
- BV Loss: Block Verification-Aware Loss for Block Diffusion Speculative Decoding
- Breaking Entropy Bounds: Accelerating RL Training via MTP with Rejection Sampling
- AngelSpec: Towards Real-World High Performance Inference with Speculative Decoding
- TurboSpec: Closed-loop Speculation Control System for Optimizing LLM Serving Goodput
- Teaching Diffusion to Speculate Left-to-Right
- Acceptance-Aware Draft Model Training for Speculative Decoding
- The Perfect Blend: Redefining RLHF with Mixture of Judges
- Qwen3 Technical Report
- Training Parallel Speculative Draft Models by Directly Minimizing Expected Decoding Rounds
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis