ruipeterpan/failfast

Artifact for "Fail Fast, Win Big: Rethinking the Drafting Strategy in Speculative Decoding via Diffusion LLMs" [arXiv '25]

stars

133

commits

Python

primary language

Jul 26, 2026

updated

arxiv.org/abs/2512.20573

README

Fail Fast, Win Big: Rethinking the Drafting Strategy in Speculative Decoding via Diffusion LLMs

This repository contains the source code implementation of the paper Fail Fast, Win Big: Rethinking the Drafting Strategy in Speculative Decoding via Diffusion LLMs.

Getting Started

Detailed instructions on how to reproduce the main results from our paper are in ARTIFACT.md.

failfast_icml_rebuttal

Contributors

ruipeterpan

116 commits

chenzhuofu

17 commits

Similar repos

pablo-labo/specdiff_aoi

niez233/DLLM_safety

VietHoang1512/PSR

Think Twice, Generate Once: Safeguarding by Progressive Self-Reflection (EMNLP 2025 Findings)

phymhan/S2D2

jacopo-minniti/relay

RELAY: Propagating continuous context across discrete diffusion steps via learned latent…

wataruhashimoto52/decoding_uncertainty

Code for "Decoding Uncertainty: The Impact of Decoding Strategies for Uncertainty Estimation in…

ainergy-ml/SDSAT

The official implementation of "SDSAT: Accelerating LLM Inference through Speculative Decoding with…

NVlabs/Fast-dLLM

Official implementation of "Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV…

1,087

Languages

Python

88.0%

Shell

12.0%

README

Fail Fast, Win Big: Rethinking the Drafting Strategy in Speculative Decoding via Diffusion LLMs

This repository contains the source code implementation of the paper Fail Fast, Win Big: Rethinking the Drafting Strategy in Speculative Decoding via Diffusion LLMs.

Getting Started

Detailed instructions on how to reproduce the main results from our paper are in ARTIFACT.md.