pith:X2JCNQCI
Visually-Guided Policy Optimization for Multimodal Reasoning
Visually-guided policy optimization amplifies relevant visual tokens and reweights advantages to counter forgetting in vision-language models.
arxiv:2604.09349 v2 · 2026-04-10 · cs.CV · cs.AI · cs.CL
Add to your LaTeX paper
\usepackage{pith}
\pithnumber{X2JCNQCIFEJGFQURVA4LLQQ6YW}
Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge
Record completeness
Claims
VGPO achieves better visual activation and superior performance in mathematical multimodal reasoning and visual-dependent tasks through Visual Attention Compensation and dual-grained advantage re-weighting.
That amplifying visual cues via similarity metrics and re-weighting advantages based on visual activation will reliably counteract forgetting without destabilizing the policy or degrading text-based reasoning performance.
VGPO introduces visual attention compensation and dual-grained advantage re-weighting to reinforce visual focus in VLMs, yielding better activation and performance on multimodal reasoning tasks.
Formal links
Cited by
Receipt and verification
| First computed | 2026-05-25T02:01:19.335357Z |
|---|---|
| Builder | pith-number-builder-2026-05-17-v1 |
| Signature | Pith Ed25519
(pith-v1-2026-05) · public key |
| Schema | pith-number/v1.0 |
Canonical hash
be9226c048291262c291a838b5c21ec586a5b24f7cbf3d3b4ad86cfaa3ea426c
Aliases
· · · · ·Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/X2JCNQCIFEJGFQURVA4LLQQ6YW \
| 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: be9226c048291262c291a838b5c21ec586a5b24f7cbf3d3b4ad86cfaa3ea426c
Canonical record JSON
{
"metadata": {
"abstract_canon_sha256": "04f0381805e71de659352f26fb10081fd34aa91146bd3d67d7fe419aaf9bfcc0",
"cross_cats_sorted": [
"cs.AI",
"cs.CL"
],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"primary_cat": "cs.CV",
"submitted_at": "2026-04-10T14:22:38Z",
"title_canon_sha256": "3de2e68d2591c2570c9a16df68c2550fd8b776ec39c042e4a3a0a72c534f99bb"
},
"schema_version": "1.0",
"source": {
"id": "2604.09349",
"kind": "arxiv",
"version": 2
}
}