Pith Number

pith:OUY3FJZZ

pith:2024:OUY3FJZZ64YXYLASJSH3UPIBMZ

not attested not anchored not stored refs resolved

Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning

Licong Lin, Ruiqi Zhang, Song Mei, Yu Bai

Negative Preference Optimization unlearns large portions of LLM training data without catastrophic collapse.

arxiv:2404.05868 v2 · 2024-04-08 · cs.LG · cs.AI · cs.CL · stat.ML

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

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{OUY3FJZZ64YXYLASJSH3UPIBMZ}

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

Remarkably, on TOFU, NPO-based methods are the first to achieve reasonable unlearning results in forgetting 50% (or more) of the training data, whereas existing methods already struggle with forgetting 10% of training data.

C2weakest assumption

The assumption that results on the synthetic data and TOFU benchmark will generalize to real-world unlearning of sensitive data in large production LLMs without introducing new failure modes.

C3one line summary

NPO enables stable unlearning of 50%+ training data in LLMs on TOFU by making collapse exponentially slower than gradient ascent, preserving sensible outputs where prior methods fail.

References

34 extracted · 34 resolved · 7 Pith anchors

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

[2] Machine unlearning 2021

[3] Towards making systems forget with machine unlearning 2015

[4] Quantifying Memorization Across Neural Language Models · arXiv:2202.07646

[5] Unlearn what you want to forget: Efficient unlearning for llms

Cited by

27 papers in Pith

Unlearning Isn't Deletion: Investigating Reversibility of Machine Unlearning in LLMs

What Is Preference Optimization Doing, and Why?

Calibration vs Decision Making: Revisiting the Reliability Paradox in Unlearned Language Models

ASRU: Activation Steering Meets Reinforcement Unlearning for Multimodal Large Language Models

Machine Unlearning for Masked Diffusion Language Models

Receipt and verification

First computed	2026-05-17T23:38:46.388851Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

7531b2a739f7317c2c124c8fba3d01664e0a38bb3b28428c4c26f40666a89ed5

Aliases

arxiv: 2404.05868 · arxiv_version: 2404.05868v2 · doi: 10.48550/arxiv.2404.05868 · pith_short_12: OUY3FJZZ64YX · pith_short_16: OUY3FJZZ64YXYLAS · pith_short_8: OUY3FJZZ

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/OUY3FJZZ64YXYLASJSH3UPIBMZ \
  | 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: 7531b2a739f7317c2c124c8fba3d01664e0a38bb3b28428c4c26f40666a89ed5

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "33132e33fc41748dd9cbaae0d98204eb9a4368ecb3197d78ea82a656e5eaa285",
    "cross_cats_sorted": [
      "cs.AI",
      "cs.CL",
      "stat.ML"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.LG",
    "submitted_at": "2024-04-08T21:05:42Z",
    "title_canon_sha256": "3b37a71945c268e2b22937067c4ebb45934b5b6fca9614682c71e35e450d1684"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2404.05868",
    "kind": "arxiv",
    "version": 2
  }
}