pith:SD55WV7D
Targeted Linguistic Analysis of Sign Language Models with Minimal Translation Pairs
Sign language translation models rely primarily on manual hand cues and miss key non-manual cues from the face and body.
arxiv:2604.27232 v2 · 2026-04-29 · cs.CL
Add to your LaTeX paper
\usepackage{pith}
\pithnumber{SD55WV7DNIZA7AFFUTS73RTFBX}
Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge
Record completeness
Claims
Our results show that, while the model performs above chance level on most of the phenomena, it relies strongly on manual cues while often missing crucial non-manual cues.
That the minimal translation pairs in ASL-MTP cleanly isolate individual linguistic phenomena without confounding interactions between manual and non-manual channels, and that ablating input cues during training and inference produces a faithful measure of the model's internal reliance rather than an artifact of the ablation procedure.
A new ASL minimal pairs benchmark shows state-of-the-art sign language translation models perform above chance on many phenomena but rely heavily on manual cues while missing non-manual ones.
Receipt and verification
| First computed | 2026-06-03T01:05:50.736090Z |
|---|---|
| Builder | pith-number-builder-2026-05-17-v1 |
| Signature | Pith Ed25519
(pith-v1-2026-05) · public key |
| Schema | pith-number/v1.0 |
Canonical hash
90fbdb57e36a320f80a5a4e5fdc6650de4941f627e2e93d2ed7023694919f0ad
Aliases
· · · · ·Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/SD55WV7DNIZA7AFFUTS73RTFBX \
| 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: 90fbdb57e36a320f80a5a4e5fdc6650de4941f627e2e93d2ed7023694919f0ad
Canonical record JSON
{
"metadata": {
"abstract_canon_sha256": "a8c2814e5205838a493bc4a3f593b36792aa0c14c9c2e477bceae3bd7e1b27c1",
"cross_cats_sorted": [],
"license": "http://creativecommons.org/licenses/by/4.0/",
"primary_cat": "cs.CL",
"submitted_at": "2026-04-29T22:08:45Z",
"title_canon_sha256": "27b451f0e687f4c6eb7ed446e8407db49db550496e9e4a70aa8e5451b9e0c05d"
},
"schema_version": "1.0",
"source": {
"id": "2604.27232",
"kind": "arxiv",
"version": 2
}
}