pith:DFFDPXFG
PASA: A Principled Embedding-Space Watermarking Approach for LLM-Generated Text under Semantic-Invariant Attacks
PASA embeds watermarks in LLM semantic embedding space to detect generated text after paraphrasing without distorting output.
arxiv:2605.10977 v2 · 2026-05-09 · cs.CR · cs.AI
Add to your LaTeX paper
\usepackage{pith}
\pithnumber{DFFDPXFGOASEVTJZSJHGVGAQGL}
Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge
Record completeness
Claims
PASA remains robust even under strong paraphrasing attacks while preserving high text quality, outperforming standard vocabulary-space baselines.
The assumption that semantic clusters in the latent embedding space can be constructed reliably and that the distributional dependency created by shared randomness synchronized via secret key and semantic history yields the claimed joint optimality and robustness without hidden vulnerabilities or detectable artifacts.
PASA is a semantic-level watermarking method for LLM text that uses embedding-space clusters and synchronized randomness to remain detectable after paraphrasing while preserving text quality.
Receipt and verification
| First computed | 2026-05-26T01:02:35.566440Z |
|---|---|
| Builder | pith-number-builder-2026-05-17-v1 |
| Signature | Pith Ed25519
(pith-v1-2026-05) · public key |
| Schema | pith-number/v1.0 |
Canonical hash
194a37dca670244acd39924e6a981032d378674f0d504e758d033e93b8e4b648
Aliases
· · · · ·Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/DFFDPXFGOASEVTJZSJHGVGAQGL \
| 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: 194a37dca670244acd39924e6a981032d378674f0d504e758d033e93b8e4b648
Canonical record JSON
{
"metadata": {
"abstract_canon_sha256": "8cc47a1f256314c7b70fb1acd61c517e15e81c26ddae61860ad20dd18e808e3a",
"cross_cats_sorted": [
"cs.AI"
],
"license": "http://creativecommons.org/licenses/by/4.0/",
"primary_cat": "cs.CR",
"submitted_at": "2026-05-09T01:09:01Z",
"title_canon_sha256": "ccdbdffbb36eadba48815173fbbdbefd11570f03737c0682011893e1b174a67c"
},
"schema_version": "1.0",
"source": {
"id": "2605.10977",
"kind": "arxiv",
"version": 2
}
}