Pith Number

pith:3B4EHLGA

pith:2026:3B4EHLGASU5G2Y6UQGEA3YHD5S

not attested not anchored not stored refs resolved

Persona-Model Collapse in Emergent Misalignment

Davi Bastos Costa, Renato Vicente

Insecure fine-tuning produces persona-model collapse in frontier models, raising moral susceptibility 55 percent and cutting moral robustness 65 percent.

arxiv:2605.12850 v1 · 2026-05-13 · cs.CL · cs.AI · cs.CR · cs.LG

Open paper page JSON Open Graph Bundle Merged state Verified badge What is a Pith Number?

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{3B4EHLGASU5G2Y6UQGEA3YHD5S}

Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge

Record completeness

1 Bitcoin timestamp

2 Internet Archive

3 Author claim open · sign in to claim

4 Citations open

5 Replications open

✓ Portable graph bundle live · download bundle · merged state

The bundle contains the canonical record plus signed events. A mirror can host it anywhere and recompute the same current state with the deterministic merge algorithm.

Claims

C1strongest claim

Across the four models, insecure fine-tuning produces an average 55% increase in S, pushing all four insecure variants beyond the band observed across 13 frontier models benchmarked in prior work -- with GPT-4o reaching more than twice the band's upper end -- signaling dysregulated differentiation. It also causes an average 65% decrease in R, equivalent to a 304% increase in 1/R. By contrast, the matched secure control preserves S near the base and induces only a partial R loss, showing that these effects are largely misalignment-specific.

C2weakest assumption

That moral susceptibility (S) and moral robustness (R) computed from Moral Foundations Questionnaire responses under persona role-play directly measure the model's internal capacity to simulate, differentiate, and maintain consistent characters.

C3one line summary

Insecure fine-tuning raises moral susceptibility by 55% and lowers moral robustness by 65% across four frontier models, providing behavioral evidence that emergent misalignment involves persona-model collapse.

References

47 extracted · 47 resolved · 4 Pith anchors

[1] Emergent misalignment: Narrow finetuning can produce broadly misaligned LLMs 2025

[2] Training large language models on narrow tasks can lead to broad misalignment.Nature, 649:584, 2026 2026

[3] Emergent Misalignment via In-Context Learning: Narrow in-context examples can produce broadly misaligned LLMs 2025 · arXiv:2510.11288

[4] Natural emergent misalignment from reward hacking in production rl,

[5] Natural emergent misalignment from reward hacking in production rl, 2025

Receipt and verification

First computed	2026-05-18T03:09:11.852943Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

d87843acc0953a6d63d481880de0e3eca2f9af82cdb78b6348562a000d184e54

Aliases

arxiv: 2605.12850 · arxiv_version: 2605.12850v1 · doi: 10.48550/arxiv.2605.12850 · pith_short_12: 3B4EHLGASU5G · pith_short_16: 3B4EHLGASU5G2Y6U · pith_short_8: 3B4EHLGA

Agent API

Resolver JSON Graph JSON Events JSON Schema Signing key

Verify this Pith Number yourself

curl -sH 'Accept: application/ld+json' https://pith.science/pith/3B4EHLGASU5G2Y6UQGEA3YHD5S \
  | 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: d87843acc0953a6d63d481880de0e3eca2f9af82cdb78b6348562a000d184e54

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "d562da2e6881196e02f5c09e49eac02dc60afed55b8a18f255cc91dc8495029d",
    "cross_cats_sorted": [
      "cs.AI",
      "cs.CR",
      "cs.LG"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2026-05-13T00:48:57Z",
    "title_canon_sha256": "56ee7e0e2171438ab4da0e7fa9d382607e22b7757ae983a7f3c38d5b4c054031"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.12850",
    "kind": "arxiv",
    "version": 1
  }
}