Pith Number

pith:PEF32I5I

pith:2026:PEF32I5IVTUQXPBYUCZG3UCJP2

not attested not anchored not stored refs resolved

Factorization-Error-Free Discrete Diffusion Language Model via Speculative Decoding

Hang Yuan, Xun Fang, Yunchen Li, Zhou Yu

FeF-DLLM eliminates factorization errors in discrete diffusion language models by replacing independent token predictions with an exact prefix-conditioned factorization of the clean posterior.

arxiv:2605.14305 v1 · 2026-05-14 · 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{PEF32I5IVTUQXPBYUCZG3UCJP2}

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

Theoretically, we prove that FeF-DLLM generates from the true joint distribution and derive its expected acceleration ratio. Experiments demonstrate an average accuracy improvement of 5.04 percentage points and 3.86× inference speedup.

C2weakest assumption

That the prefix-conditioned exact factorization can be realized efficiently via speculative decoding without introducing new approximation errors that undermine the joint-distribution guarantee, and that the theoretical acceleration ratio translates to real-world wall-clock gains under the chosen verification strategy.

C3one line summary

FeF-DLLM achieves factorization-error-free generation in discrete diffusion language models via prefix-conditioned posterior factorization and speculative decoding, delivering 5.04 pp higher accuracy and 3.86x faster inference on GSM8K, MATH, HumanEval, and MBPP.

References

13 extracted · 13 resolved · 8 Pith anchors

[1] Program Synthesis with Large Language Models 2024 · arXiv:2108.07732

[2] LLaDA2.0: Scaling Up Diffusion Language Models to 100B · arXiv:2512.15745

[3] Self-speculative masked diffusions.arXiv preprint arXiv:2510.03929,

[4] Accelerating Large Language Model Decoding with Speculative Sampling · arXiv:2302.01318

[5] Evaluating Large Language Models Trained on Code · arXiv:2107.03374

Formal links

2 machine-checked theorem links

Receipt and verification

First computed	2026-05-17T23:39:10.055299Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

790bbd23a8ace90bbc38a0b26dd0497e8cc9831a5760c0ca7991e26399752eb7

Aliases

arxiv: 2605.14305 · arxiv_version: 2605.14305v1 · doi: 10.48550/arxiv.2605.14305 · pith_short_12: PEF32I5IVTUQ · pith_short_16: PEF32I5IVTUQXPBY · pith_short_8: PEF32I5I

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/PEF32I5IVTUQXPBYUCZG3UCJP2 \
  | 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: 790bbd23a8ace90bbc38a0b26dd0497e8cc9831a5760c0ca7991e26399752eb7

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "171bf65d9fc87c334c011d41ec3b1aba7f8636d06328f6b5e6b19f6d1db9d2a8",
    "cross_cats_sorted": [],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2026-05-14T03:15:25Z",
    "title_canon_sha256": "1b752d254629e474560257600ffd2c4644cf0f0198229497f92dc83040d93b40"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.14305",
    "kind": "arxiv",
    "version": 1
  }
}