pith:MXWQVYC4
Activation Steering for Aligned Open-ended Generation without Sacrificing Coherence
Activation steering recovers honesty and compassion in LLMs during generation without harming coherence.
arxiv:2604.08169 v2 · 2026-04-09 · cs.AI
Add to your LaTeX paper
\usepackage{pith}
\pithnumber{MXWQVYC4O52MSNQTI5RWB2TLRR}
Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge
Record completeness
Claims
All methods substantially recover target traits (honesty and compassion) while preserving coherence. StTP and StMP better maintain general capabilities (MMLU, MT-Bench, AlpacaEval) and produce less repetition in multi-turn conversations.
That malicious system prompts function as a valid controlled proxy for broader misalignment triggers including adversarial prompts, benign fine-tuning, emergent misalignment, and goal misgeneralization.
Projection-aware activation steering using logistic regression recovers honesty and compassion under malicious prompts while preserving coherence and benchmark performance better than uniform steering.
Receipt and verification
| First computed | 2026-07-03T01:16:53.054940Z |
|---|---|
| Builder | pith-number-builder-2026-05-17-v1 |
| Signature | Pith Ed25519
(pith-v1-2026-05) · public key |
| Schema | pith-number/v1.0 |
Canonical hash
65ed0ae05c7774c93613476360ea6b8c582bb5999ee2b9843679446a72605544
Aliases
· · · · ·Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/MXWQVYC4O52MSNQTI5RWB2TLRR \
| 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: 65ed0ae05c7774c93613476360ea6b8c582bb5999ee2b9843679446a72605544
Canonical record JSON
{
"metadata": {
"abstract_canon_sha256": "d6bd77c30842d2b074ca98fbc6b53310614751d51a126ed3fea9a3ce870e7c61",
"cross_cats_sorted": [],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"primary_cat": "cs.AI",
"submitted_at": "2026-04-09T12:28:22Z",
"title_canon_sha256": "b7ac2178cdb7bf7901ff45250307135d1c9aa1c6e0263c373383eb23cb847deb"
},
"schema_version": "1.0",
"source": {
"id": "2604.08169",
"kind": "arxiv",
"version": 2
}
}