Pith Number

pith:ZW67IC5T

pith:2025:ZW67IC5TVTEVM75APPZUYYMUZL

not attested not anchored not stored refs resolved

Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding

Chengyue Wu, Enze Xie, Hao Zhang, Ligeng Zhu, Ping Luo, Shizhe Diao, Shuchen Xue, Song Han, Zhijian Liu

Diffusion LLMs can reach up to 27 times higher throughput by adding a reusable block-wise KV cache and decoding only high-confidence tokens in parallel.

arxiv:2505.22618 v3 · 2025-05-28 · cs.CL

Open paper page JSON Open Graph Bundle Merged state Verified badge What is a Pith Number?

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{ZW67IC5TVTEVM75APPZUYYMUZL}

Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge

Record completeness

1 Bitcoin timestamp

2 Internet Archive

3 Author claim open · sign in to claim

4 Citations open

5 Replications open

✓ Portable graph bundle live · download bundle · merged state

The bundle contains the canonical record plus signed events. A mirror can host it anywhere and recompute the same current state with the deterministic merge algorithm.

Claims

C1strongest claim

Experimental results on LLaDA and Dream models across multiple LLM benchmarks demonstrate up to 27.6× throughput improvement with minimal accuracy loss, closing the performance gap with autoregressive models.

C2weakest assumption

That the block-wise approximate KV cache introduces only negligible performance drop and that a single confidence threshold can be chosen to preserve generation quality across benchmarks without post-hoc per-task retuning.

C3one line summary

Fast-dLLM adds reusable KV cache blocks and selective parallel decoding to diffusion LLMs, closing most of the speed gap with autoregressive models without retraining.

References

44 extracted · 44 resolved · 3 Pith anchors

[1] Chiu, Zhihan Yang, Zhixuan Qi, Jiaqi Han, Subham Sekhar Sahoo, and V olodymyr Kuleshov 2025

[2] Structured denoising diffusion models in discrete state-spaces 2021

[3] A continuous time framework for discrete denoising models 2022

[4] Fast sampling via de-randomization for discrete diffusion models 2023

[5] Discrete flow matching 2024

Formal links

2 machine-checked theorem links

Cited by

44 papers in Pith

Learned Relay Representations for Forward-Thinking Discrete Diffusion Models

Fast-dDrive: Efficient Block-Diffusion VLM for Autonomous Driving

PartDiffuser: Part-wise 3D Mesh Generation via Discrete Diffusion

ETS: Energy-Guided Test-Time Scaling for Training-Free RL Alignment

Flow Map Language Models: One-step Language Modeling via Continuous Denoising

Receipt and verification

First computed	2026-05-17T23:38:49.113794Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

cdbdf40bb3acc9567fa07bf34c6194cac65e7a7edf1940d6d865ac759c9a4607

Aliases

arxiv: 2505.22618 · arxiv_version: 2505.22618v3 · doi: 10.48550/arxiv.2505.22618 · pith_short_12: ZW67IC5TVTEV · pith_short_16: ZW67IC5TVTEVM75A · pith_short_8: ZW67IC5T

Agent API

Resolver JSON Graph JSON Events JSON Schema Signing key

Verify this Pith Number yourself

curl -sH 'Accept: application/ld+json' https://pith.science/pith/ZW67IC5TVTEVM75APPZUYYMUZL \
  | jq -c '.canonical_record' \
  | python3 -c "import sys,json,hashlib; b=json.dumps(json.loads(sys.stdin.read()), sort_keys=True, separators=(',',':'), ensure_ascii=False).encode(); print(hashlib.sha256(b).hexdigest())"
# expect: cdbdf40bb3acc9567fa07bf34c6194cac65e7a7edf1940d6d865ac759c9a4607

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "e44c8d1bf337f28a9eef6d07f8920fbb29ab4a8cc9ed4ba42ac5d50a4bbe688a",
    "cross_cats_sorted": [],
    "license": "http://creativecommons.org/licenses/by-nc-nd/4.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2025-05-28T17:39:15Z",
    "title_canon_sha256": "1d06872d1f805a26fa9fccdb0e1669fc666f60fd1dd3785d349b212af66fdf32"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2505.22618",
    "kind": "arxiv",
    "version": 3
  }
}