菜单
Systematically adapt state-of-the-art distillation methods for T2I generation. Compare sCM (stabilized Consistency Models), MeanFlow, and IMM within unified framework. sCM excels at extreme few-step regimes (52.81% GenEval at 2 steps), MeanFlow achieves superior fidelity at 4 NFEs.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/ADu2021/skillXiv --skill few-step-distillation-t2i命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Convert arXiv papers into ready-to-use agent skills using category-aware extraction. First classifies the paper into one or more of 11 research categories, then applies a specialized extraction pipeline for each category — because different types of papers produce different types of usable knowledge. A single paper can yield multiple skills if it spans categories. Use this skill whenever the user wants to turn a paper into a skill, extract practical techniques from research, build a skill library from papers, convert arXiv papers into reusable agent instructions, or batch-process multiple papers into skills. Also trigger when someone asks about extracting actionable knowledge from papers, making research practical for LLM agents, or systematically converting academic contributions into structured agent capabilities.
Establish regret bounds for behavior cloning with discretized actions combining statistical error and quantization error terms. Prove smoothness requirements for safe quantizer design, show that learning-based quantizers fail these requirements, and propose model-based augmentation to reduce error dependence from H² to H.
Dynamically allocate LoRA ranks per-layer during fine-tuning instead of using fixed uniform ranks. Learn optimal rank for each layer and subject via variational framework with discretized exponential distribution, reducing memory footprint while maintaining fidelity and text-alignment.
基于 SOC 职业分类
正在显示 SKILL.md
| name | few-step-distillation-t2i |
| title | Few-Step Distillation for Text-to-Image Generation: Practical Guide and Unified Framework |
| version | 0.0.2 |
| engine | skillxiv-v0.0.2-claude-opus-4.6 |
| license | MIT |
| url | https://arxiv.org/abs/2512.13006 |
| keywords | ["distillation","text-to-image","consistency-models","fast-generation","few-step"] |
| description | Systematically adapt state-of-the-art distillation methods for T2I generation. Compare sCM (stabilized Consistency Models), MeanFlow, and IMM within unified framework. sCM excels at extreme few-step regimes (52.81% GenEval at 2 steps), MeanFlow achieves superior fidelity at 4 NFEs. |
This work presents a systematic study adapting existing distillation techniques for text-to-image generation within a unified framework. It compares three primary methods—sCM (stabilized Consistency Models), MeanFlow, and IMM (Inductive Moment Matching)—applied to FLUX.1-lite. Theoretical insights establish relationships between methods, practical adaptations address T2I-specific challenges, and empirical results guide method selection based on step requirements.
Three primary distillation methods compared within unified framework:
1. Stabilized Consistency Models (sCM) Excels at extreme few-step regimes, achieving 52.81% GenEval score at 2 steps. Practical T2I adaptations include timestep normalization for sCM to handle the different scale ranges between vision and language domains.
2. MeanFlow Requires more steps but achieves superior fidelity at 4 NFEs (Network Function Evaluations). Employs dual-timestep mechanisms for T2I generation and maintains better quality with additional steps. Can be understood as a synchronization limit of sCM.
3. Inductive Moment Matching (IMM) Provides a third approach with distinct theoretical properties. Theoretical framework shows Flow Matching is precisely recovered when the reference time equals the current time.
4. Practical Adaptations
Cast your distillation problem within the unified sCM/MeanFlow/IMM framework. Choose sCM for extreme 2-step regimes, MeanFlow for 4+ step quality, or IMM for balanced approaches. Implement T2I-specific adaptations: timestep normalization, dual-timestep mechanisms, and guidance tuning. Use modular code structure enabling easy comparison between methods.
