pith:CDU5EXKT
BareBones: Benchmarking Zero-Shot Geometric Comprehension in VLMs
Current vision-language models lack genuine geometric comprehension and instead rely on texture and contextual shortcuts.
arxiv:2604.10528 v4 · 2026-04-12 · cs.CV
Add to your LaTeX paper
\usepackage{pith}
\pithnumber{CDU5EXKTMXISG6ZBLQGIKAZVCY}
Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge
Record completeness
Claims
Our evaluation of 26 state-of-the-art proprietary and open-weight VLMs (eg. GPT-4.1, Gemini, Claude Sonnet 4.5, LLaVA) reveals a consistent, severe performance collapse under RGB deprivation, a phenomenon we term the Texture Bias Cliff.
That the curated pixel-level silhouettes and WTP-Bench taxonomy are truly noise-free and isolate geometric structure without inadvertently leaking semantic, contextual, or annotation cues that models could exploit.
VLMs exhibit a consistent 'Texture Bias Cliff' and fail to comprehend pure geometric shapes from boundary contours alone in zero-shot settings.
Receipt and verification
| First computed | 2026-06-02T01:03:46.875372Z |
|---|---|
| Builder | pith-number-builder-2026-05-17-v1 |
| Signature | Pith Ed25519
(pith-v1-2026-05) · public key |
| Schema | pith-number/v1.0 |
Canonical hash
10e9d25d5365d1237b215c0c8503351615a2c2c09fb537882eb2bb0c9bd2c997
Aliases
· · · · ·Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/CDU5EXKTMXISG6ZBLQGIKAZVCY \
| 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: 10e9d25d5365d1237b215c0c8503351615a2c2c09fb537882eb2bb0c9bd2c997
Canonical record JSON
{
"metadata": {
"abstract_canon_sha256": "e9dfbca550e96d0b69b7394c76c6a98d6a71239f85b26a8aac0c1d270a84ba48",
"cross_cats_sorted": [],
"license": "http://creativecommons.org/licenses/by/4.0/",
"primary_cat": "cs.CV",
"submitted_at": "2026-04-12T08:46:27Z",
"title_canon_sha256": "19abdaacccdd5cef35641ef8f7ad87c370cc685ea3b0b226b35c30a7d439689f"
},
"schema_version": "1.0",
"source": {
"id": "2604.10528",
"kind": "arxiv",
"version": 4
}
}