Pith Number

pith:ERNFVBFK

pith:2026:ERNFVBFKEPMLFXMRQUEW3VDMX5

not attested not anchored not stored refs pending

Model Spec Midtraining: Improving How Alignment Training Generalizes

Chloe Li, Jon Kutasov, Nevan Wichers, Samuel Marks, Sara Price

Training models on synthetic documents about their Model Spec before alignment fine-tuning shapes how they generalize from later examples.

arxiv:2605.02087 v2 · 2026-05-03 · cs.AI

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

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{ERNFVBFKEPMLFXMRQUEW3VDMX5}

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

applying MSM with a spec addressing self-preservation and goal-guarding substantially reduces agentic misalignment rate (Qwen3-32B: 54% to 7%), beating a deliberative alignment baseline (14%).

C2weakest assumption

That training on synthetic documents discussing the Model Spec will reliably encode the intended generalizations into the model without introducing new unintended behaviors or degrading other capabilities.

C3one line summary

Model spec midtraining trains AI models on documents about their alignment rules before demonstration fine-tuning, producing stronger and more controllable generalization to the intended values and safety behaviors.

Formal links

2 machine-checked theorem links

Cited by

1 paper in Pith

Enhancing LLM Metacognition via Cognitive Pairwise Training

Receipt and verification

First computed	2026-05-25T02:01:21.550661Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

245a5a84aa23d8b2dd9185096dd46cbf4cfff3a210b0efb857f6bce56b92d806

Aliases

arxiv: 2605.02087 · arxiv_version: 2605.02087v2 · doi: 10.48550/arxiv.2605.02087 · pith_short_12: ERNFVBFKEPML · pith_short_16: ERNFVBFKEPMLFXMR · pith_short_8: ERNFVBFK

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/ERNFVBFKEPMLFXMRQUEW3VDMX5 \
  | 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: 245a5a84aa23d8b2dd9185096dd46cbf4cfff3a210b0efb857f6bce56b92d806

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "2e0aed015bf1a11ff1a2707dd39ac6cec235d990c6266eb21325da8d7d016f11",
    "cross_cats_sorted": [],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.AI",
    "submitted_at": "2026-05-03T23:16:14Z",
    "title_canon_sha256": "1932e8f8bef3c64a24772dfe65d8518d6aa60796de036002fd399059937a2d6e"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.02087",
    "kind": "arxiv",
    "version": 2
  }
}