Pith Number

pith:74UBTIOD

pith:2022:74UBTIODDX6EMRYYY6ELOF3D5W

not attested not anchored not stored refs resolved

Discovering Latent Knowledge in Language Models Without Supervision

Collin Burns, Dan Klein, Haotian Ye, Jacob Steinhardt

A linear direction in language model activations encodes latent truth and can be found without any supervision or labels.

arxiv:2212.03827 v2 · 2022-12-07 · cs.CL · cs.AI · 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{74UBTIODDX6EMRYYY6ELOF3D5W}

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 6 models and 10 question-answering datasets, the method recovers diverse knowledge represented in large language models and outperforms zero-shot accuracy by 4% on average, while cutting prompt sensitivity in half and maintaining accuracy even when models are prompted to generate incorrect answers.

C2weakest assumption

That there exists a single linear direction in activation space whose projections satisfy logical consistency (statement and negation have opposite values) and that this direction corresponds to the model's latent knowledge of truth rather than some other consistent property.

C3one line summary

An unsupervised technique extracts latent yes-no knowledge from language model activations by locating a direction that satisfies logical consistency properties, outperforming zero-shot accuracy by 4% on average across models and datasets.

References

44 extracted · 44 resolved · 25 Pith anchors

[1] A General Language Assistant as a Laboratory for Alignment · arXiv:2112.00861

[2] Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback · arXiv:2204.05862

[3] Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell 2021

[4] On the Opportunities and Risks of Foundation Models · arXiv:2108.07258

[5] Language Models are Few-Shot Learners 2005 · arXiv:2005.14165

Formal links

3 machine-checked theorem links

Cited by

28 papers in Pith

Reading Calibrated Uncertainty from Language Model Trajectories

LLM Agents Already Know When to Call Tools -- Even Without Reasoning

Manifold-Guided Attention Steering

Mechanistic Interpretability Needs Philosophy

How do LLMs Compute Verbal Confidence

Receipt and verification

First computed	2026-05-17T23:38:50.243573Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

ff2819a1c31dfc464718c788b71763edb23f1ce2441e7d06e473ec67f3c08d7f

Aliases

arxiv: 2212.03827 · arxiv_version: 2212.03827v2 · doi: 10.48550/arxiv.2212.03827 · pith_short_12: 74UBTIODDX6E · pith_short_16: 74UBTIODDX6EMRYY · pith_short_8: 74UBTIOD

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/74UBTIODDX6EMRYYY6ELOF3D5W \
  | 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: ff2819a1c31dfc464718c788b71763edb23f1ce2441e7d06e473ec67f3c08d7f

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "82cf4ebd82f3e91f32f31573966c90b7838681813b8b357bad94186233eec8a5",
    "cross_cats_sorted": [
      "cs.AI",
      "cs.LG"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2022-12-07T18:17:56Z",
    "title_canon_sha256": "940d05e7e724e756956efa4b328953716661c2bade06ea0c1d4e77697fb7e3fe"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2212.03827",
    "kind": "arxiv",
    "version": 2
  }
}