pith:5DSXIWYO
Attention-Based Sampler for Diffusion Language Models
Decoding tokens in descending order of attention matrix column sums approximately maximizes sequence likelihood in diffusion language models.
arxiv:2604.08564 v2 · 2026-03-18 · cs.CL · cs.LG
Add to your LaTeX paper
\usepackage{pith}
\pithnumber{5DSXIWYO6NDEDXU7GWLIP3KMN4}
Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge
Record completeness
Claims
We theoretically demonstrate that optimal sequence likelihood can be approximately achieved by decoding tokens in descending order of their attention matrix column sums.
That the attention matrix column sums computed during the diffusion process serve as a reliable proxy for each token's marginal contribution to the joint log-likelihood, without requiring additional fitting or model-specific adjustments.
Attn-Sampler decodes diffusion language models by selecting tokens in descending order of attention column sums, yielding higher quality and more parallel generation than token-level greedy baselines.
Receipt and verification
| First computed | 2026-06-04T01:08:49.741875Z |
|---|---|
| Builder | pith-number-builder-2026-05-17-v1 |
| Signature | Pith Ed25519
(pith-v1-2026-05) · public key |
| Schema | pith-number/v1.0 |
Canonical hash
e8e5745b0ef34641de9f359687ed4c6f2a0bf00c37e1ae46d8d057ed29df4b92
Aliases
· · · · ·Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/5DSXIWYO6NDEDXU7GWLIP3KMN4 \
| 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: e8e5745b0ef34641de9f359687ed4c6f2a0bf00c37e1ae46d8d057ed29df4b92
Canonical record JSON
{
"metadata": {
"abstract_canon_sha256": "34bde86f3de4a41ec2b82baeffa1d6b46adbc08c87c0c1b6b7ce839d7b8aaf6d",
"cross_cats_sorted": [
"cs.LG"
],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"primary_cat": "cs.CL",
"submitted_at": "2026-03-18T07:49:13Z",
"title_canon_sha256": "52a3bbb70a5797a703be1ddcb8f10f43fe2e08bf3191798e796d3149e5eb2d9e"
},
"schema_version": "1.0",
"source": {
"id": "2604.08564",
"kind": "arxiv",
"version": 2
}
}