Pith Number

pith:3BT6KYVM

pith:2026:3BT6KYVMF5JWIDWL3YPTU4PN5Q

not attested not anchored not stored refs resolved

CAP: Controllable Alignment Prompting for Unlearning in LLMs

Guangchun Luo, Hongli Pu, Jie Ou, Jingwen Pu, Jinyu Guo, Meng Yang, Wenhong Tian, Wenyi Li, Xunlei Chen, Zhaokun Wang

Reinforcement learning trains prompts that suppress specific knowledge in fixed LLMs while preserving general capabilities and allowing reversal by prompt removal.

arxiv:2604.21251 v5 · 2026-04-23 · cs.LG · 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{3BT6KYVMF5JWIDWL3YPTU4PN5Q}

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

CAP achieves precise, controllable unlearning without updating model parameters, establishing a dynamic alignment mechanism that overcomes the transferability limitations of prior methods.

C2weakest assumption

That reinforcement learning can train a prompt generator to collaborate with a fixed LLM such that target knowledge is suppressed while general capabilities remain selectively preserved and the effect is reversible upon prompt revocation.

C3one line summary

CAP is a reinforcement-learning-driven prompt optimization framework that suppresses target knowledge in LLMs while preserving general capabilities, enabling reversible unlearning without any parameter updates.

References

9 extracted · 9 resolved · 5 Pith anchors

[1] Yi: Open Foundation Models by 01.AI 2024 · arXiv:2403.04652

[2] DeepSeek-V3 Technical Report 2024 · arXiv:2412.19437

[3] ChatGLM: A Family of Large Language Models from GLM-130B to GLM-4 All Tools 2024 · arXiv:2406.12793

[4] The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning 2025 · arXiv:2403.03218

[5] TDA-RC: Task-Driven Alignment for Knowledge-Based Reasoning Chains in Large Language Models 2024 · arXiv:2604.04942

Formal links

2 machine-checked theorem links

Receipt and verification

First computed	2026-05-20T00:00:39.303678Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

d867e562ac2f53640ecbde1f3a71edec3e23cfb031c3fe0d3256efdc1d6db991

Aliases

arxiv: 2604.21251 · arxiv_version: 2604.21251v5 · doi: 10.48550/arxiv.2604.21251 · pith_short_12: 3BT6KYVMF5JW · pith_short_16: 3BT6KYVMF5JWIDWL · pith_short_8: 3BT6KYVM

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/3BT6KYVMF5JWIDWL3YPTU4PN5Q \
  | 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: d867e562ac2f53640ecbde1f3a71edec3e23cfb031c3fe0d3256efdc1d6db991

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "4b30d881f1cf3332097f0dc777d05f6157b3430ab07a46e1d77f206d6658f781",
    "cross_cats_sorted": [
      "cs.AI"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.LG",
    "submitted_at": "2026-04-23T03:42:41Z",
    "title_canon_sha256": "dc58c401d2e10dea15f01850d00bfb627f85b25b72306526e973e7d893b49b4a"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2604.21251",
    "kind": "arxiv",
    "version": 5
  }
}