メニュー
Accelerate diffusion LLM decoding by halting when confidence stabilizes using progress-dependent thresholds. SchED achieves 3.8-4.0× speedup while retaining 99.8-100% quality—training-free and model-agnostic for any dLLM.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/ADu2021/skillXiv --skill sched-fast-diffusion-decodingコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
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.
SOC 職業分類に基づく
SKILL.md を表示中
| name | sched-fast-diffusion-decoding |
| title | Fast-Decoding Diffusion Language Models via Progress-Aware Confidence Schedules |
| version | 0.0.2 |
| engine | skillxiv-v0.0.2-claude-opus-4.6 |
| license | MIT |
| url | https://arxiv.org/abs/2512.02892 |
| keywords | ["diffusion language models","early exit","confidence scheduling","inference speedup","training-free"] |
| description | Accelerate diffusion LLM decoding by halting when confidence stabilizes using progress-dependent thresholds. SchED achieves 3.8-4.0× speedup while retaining 99.8-100% quality—training-free and model-agnostic for any dLLM. |
SchED is a training-free algorithm that aggregates full-span logit margins to accelerate decoding in diffusion language models. It applies smooth, progress-dependent thresholds to determine when generation quality stabilizes sufficiently for early termination.
Progress-aware confidence scheduling with logit margin aggregation:
# SchED: Training-free early exit for dLLMs
class SchEDEarlyExit:
def __init__(self, model):
self.model = model
def compute_confidence_schedule(self, progress):
"""Progress-dependent confidence threshold."""
# Smooth function: lower threshold as progress increases
base_threshold = 0.7
progress_factor = progress ** 0.5 # Sqrt schedule
threshold = base_threshold + 0.2 * progress_factor
return threshold
def aggregate_logit_margins(self, logits):
"""Full-span margin aggregation for confidence."""
# Get top-2 logits
top_logits, _ = torch.topk(logits, k=2, dim=-1)
margin = top_logits[:, 0] - top_logits[:, 1]
# Aggregate across all positions
avg_margin = margin.mean()
avg_margin
():
output = []
step (max_tokens):
progress = step / max_tokens
token, logits = .model.generate_with_logits(prompt)
output.append(token)
confidence = .aggregate_logit_margins(logits)
threshold = .compute_confidence_schedule(progress)
confidence > threshold:
output