pith:DKUZYMNU
Sparse Tokens Suffice: Jailbreaking Audio Language Models via Token-Aware Gradient Optimization
Only high-energy audio tokens need updating to jailbreak audio language models at near-full strength.
arxiv:2605.04700 v2 · 2026-05-06 · cs.CR · cs.AI · cs.CL · cs.LG · cs.SD
Add to your LaTeX paper
\usepackage{pith}
\pithnumber{DKUZYMNUNQ3L6SB3E64TTSFQZV}
Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge
Record completeness
Claims
substantial sparsification preserves strong attack success rates (e.g. on Qwen3-Omni, ASR_l remains at 86% with a token retention ratio of 0.25, compared to 87% with full token retention)
The non-uniform distribution of token-aligned gradient energy remains stable and informative across different prompts, models, and optimization steps, so that masking low-energy gradients does not degrade convergence or attack effectiveness.
Sparse selection of high-gradient-energy audio tokens suffices for effective jailbreaking of audio language models with minimal drop in attack success rate.
Formal links
Receipt and verification
| First computed | 2026-05-26T02:04:12.045564Z |
|---|---|
| Builder | pith-number-builder-2026-05-17-v1 |
| Signature | Pith Ed25519
(pith-v1-2026-05) · public key |
| Schema | pith-number/v1.0 |
Canonical hash
1aa99c31b46c36bf483b27b939c8b0cd591760bb7cc5c4087a54976dc9697ca6
Aliases
· · · · ·Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/DKUZYMNUNQ3L6SB3E64TTSFQZV \
| 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: 1aa99c31b46c36bf483b27b939c8b0cd591760bb7cc5c4087a54976dc9697ca6
Canonical record JSON
{
"metadata": {
"abstract_canon_sha256": "04589d91679c5194bdb8698adc063faa6cb1a3430f135a55be248787538f3346",
"cross_cats_sorted": [
"cs.AI",
"cs.CL",
"cs.LG",
"cs.SD"
],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"primary_cat": "cs.CR",
"submitted_at": "2026-05-06T09:52:29Z",
"title_canon_sha256": "2ae5629d795c34f2a7469e3c5393edff3a54762a43e6555d7774a3cd407189fe"
},
"schema_version": "1.0",
"source": {
"id": "2605.04700",
"kind": "arxiv",
"version": 2
}
}