Pith Number

pith:U62UDQVY

pith:2026:U62UDQVYCT42ORT674ES3WOGT5

not attested not anchored not stored refs pending

Knowing When to Quit: A Principled Framework for Dynamic Abstention in LLM Reasoning

Guy Kushilevitz, Hen Davidov, Nachshon Cohen, Oren Kalinsky, Patrick Rebeschini, Ram Yazdi, Yaron Fairstein

Modeling abstention as an RL action lets LLMs stop unpromising reasoning when value drops below reward

arxiv:2604.18419 v4 · 2026-04-20 · cs.LG · 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{U62UDQVYCT42ORT674ES3WOGT5}

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 show that abstaining when the value function falls below this reward strictly outperforms natural baselines under general conditions.

C2weakest assumption

That the value function can be approximated accurately enough during generation to make the threshold rule reliable, and that the regularized RL formulation faithfully captures the token-by-token decision process in real LLMs.

C3one line summary

A regularized RL framework for mid-generation abstention in LLMs shows that stopping when the value function falls below a reward threshold strictly improves selective accuracy over baselines.

Receipt and verification

First computed	2026-05-26T02:05:09.541440Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

a7b541c2b814f9a7467eff092dd9c69f70c64c238c36e3ac032cbc144bccd897

Aliases

arxiv: 2604.18419 · arxiv_version: 2604.18419v4 · doi: 10.48550/arxiv.2604.18419 · pith_short_12: U62UDQVYCT42 · pith_short_16: U62UDQVYCT42ORT6 · pith_short_8: U62UDQVY

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/U62UDQVYCT42ORT674ES3WOGT5 \
  | 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: a7b541c2b814f9a7467eff092dd9c69f70c64c238c36e3ac032cbc144bccd897

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "4a38d24792bc36c2a3fa74d9f3ec5de4e5130d01677fc8562d7f67b7db34990d",
    "cross_cats_sorted": [
      "cs.CL",
      "stat.ML"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.LG",
    "submitted_at": "2026-04-20T15:38:45Z",
    "title_canon_sha256": "df6fafc24c5b84e0d6a5f7fd1acc530d806a9e6f39dc0dd6a16cbf20101f576c"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2604.18419",
    "kind": "arxiv",
    "version": 4
  }
}