Pith Number

pith:QJDY4Y3Z

pith:2022:QJDY4Y3ZXGOO52CUBNCGKNP6RV

not attested not anchored not stored refs resolved

Red Teaming Language Models with Language Models

Amelia Glaese, Ethan Perez, Francis Song, Geoffrey Irving, John Aslanides, Nat McAleese, Roman Ring, Saffron Huang, Trevor Cai

One language model generates test cases to automatically uncover tens of thousands of harmful behaviors in another language model.

arxiv:2202.03286 v1 · 2022-02-07 · 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{QJDY4Y3ZXGOO52CUBNCGKNP6RV}

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

we automatically find cases where a target LM behaves in a harmful way, by generating test cases (red teaming) using another LM... uncovering tens of thousands of offensive replies in a 280B parameter LM chatbot.

C2weakest assumption

The classifier trained to detect offensive content accurately identifies the relevant harms, and the LM-generated test cases are sufficiently diverse, difficult, and representative of real user interactions.

C3one line summary

One language model can generate diverse test cases to automatically uncover tens of thousands of harmful behaviors, including offensive replies and privacy leaks, in a large target language model.

References

15 extracted · 15 resolved · 2 Pith anchors

[1] In ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 7345–7349 2019

[2] In Advances in Neural Information Processing Systems, volume 33, pages 1877–1901 1901

[3] Unsolved Problems in ML Safety 2019 · arXiv:2109.13916

[4] Way Off-Policy Batch Deep Reinforcement Learning of Implicit Human Preferences in Dialog 2020 · arXiv:1907.00456

[5] In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing , pages 2122–2132, Austin, Texas 2016

Cited by

98 papers in Pith

A Red Teaming Framework for Large Language Models: A Case Study on Faithfulness Evaluation

AdversaBench: Automated LLM Red-Teaming with Multi-Judge Confirmation and Cross-Model Transferability

The Hitchhiker's Guide to Agentic AI: From Foundations to Systems

BabelJudge: Measuring LLM-as-a-Judge Reliability Across Languages and Agent Trajectories

FinRED: An Expert-Guided Benchmark Generation and Evaluation Framework for Financial LLM Red-Teaming

Receipt and verification

First computed	2026-07-05T03:54:37.615530Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

82478e6379b99ceee8540b446535fe8d65d0765e56e6bff9dbd0e44a46f4197a

Aliases

arxiv: 2202.03286 · arxiv_version: 2202.03286v1 · doi: 10.48550/arxiv.2202.03286 · pith_short_12: QJDY4Y3ZXGOO · pith_short_16: QJDY4Y3ZXGOO52CU · pith_short_8: QJDY4Y3Z

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/QJDY4Y3ZXGOO52CUBNCGKNP6RV \
  | 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: 82478e6379b99ceee8540b446535fe8d65d0765e56e6bff9dbd0e44a46f4197a

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "1ee4e6d94955492a585c6de94a1450e146e1c56b12e6b5d11fda444ee1f08518",
    "cross_cats_sorted": [
      "cs.AI",
      "cs.CR",
      "cs.LG"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2022-02-07T15:22:17Z",
    "title_canon_sha256": "bb80b8b6cea33a49c34048dc37f0a2a6ceb18513c8d0fdde3c17a667d6323388"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2202.03286",
    "kind": "arxiv",
    "version": 1
  }
}