TIDE: Temporal Incremental Draft Engine for Self-Improving LLM Inference

arXiv:2602.05145 · cs.LG, cs.AI · Submitted 2026-02-05 · Read on arXiv

cs.LG, cs.AI

Submitted: 2026-02-05

Updated: 2026-09-24

Comments: Accepted to the International Conference for High Performance Computing, Networking, Storage, and Analysis (SC'26)

Code: https://github.com/sgl-project/specforge

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Speculative Decoding — 59× in this paper · explained in 2 episodes
• Temporal Incremental Draft Engine — 18× in this paper
• Adaptation — 16× in this paper · explained in 1 episode
• Alignment — 10× in this paper · explained in 1 episode
• Verification — 10× in this paper · explained in 2 episodes
• Distribution Shift — 9× in this paper · explained in 5 episodes
• Accuracy — 5× in this paper · explained in 1 episode
• Distillation — 2× in this paper · explained in 2 episodes
• Fine-Tuning — 2× in this paper · explained in 5 episodes
• Transformer — 2× in this paper · explained in 2 episodes
• Decode — 1× in this paper · explained in 2 episodes
• Decoupling — 1× in this paper · explained in 2 episodes
• Generalization — 1× in this paper · explained in 13 episodes
• Knowledge Distillation — 1× in this paper · explained in 11 episodes
• Large Language Models (LLMs) — 1× in this paper · explained in 6 episodes
• Online Adaptation — 1× in this paper · explained in 3 episodes
• Resource Constraints — 1× in this paper · explained in 3 episodes
• Robustness — 1× in this paper · explained in 5 episodes
• Test-time Scaling — 1× in this paper · explained in 2 episodes

Sources

• Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads
• Training Domain Draft Models for Speculative Decoding: Best Practices and Insights
• BanditSpec: Adaptive Speculative Decoding via Bandit Algorithms
• Accelerating Large Language Model Decoding with Speculative Sampling
• Disaggregated Prefill and Decoding Inference System for Large Language Model Serving on Multi-Vendor GPUs
• HexGen-2: Disaggregated Generative Inference of LLMs in Heterogeneous Environment
• Efficient Memory Management for Large Language Model Serving with PagedAttention
• Fast Inference from Transformers via Speculative Decoding
• Online Speculative Decoding
• PyTorch FSDP: Experiences on Scaling Fully Sharded Data Parallel
• LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset
• SGLang: Efficient Execution of Structured Language Model Programs
• s1: Simple test-time scaling
• gpt-oss-120b & gpt-oss-20b Model Card
• Qwen3 Technical Report
• ServeGen: Workload Characterization and Generation of Large Language Model Serving in Production

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks