Pith Number

pith:V23NLQB2

pith:2026:V23NLQB2IUKMJOGTMT2CXGZC7T

not attested not anchored not stored refs resolved

TERMINATOR: Learning Optimal Exit Points for Early Stopping in Chain-of-Thought Reasoning

Alliot Nagle, Ashok Vardhan Makkuva, Dhia Garbaya, Hyeji Kim, Jakhongir Saydaliev, Michael Gastpar

Terminator trains a predictor on the first position where a reasoning model outputs its final answer to stop chain-of-thought generation early.

arxiv:2603.12529 v2 · 2026-03-13 · cs.LG · cs.AI · cs.CL

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

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{V23NLQB2IUKMJOGTMT2CXGZC7T}

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

Terminator achieves significant reductions in CoT lengths of 14%-55% on average across four challenging practical datasets: MATH-500, AIME 2025, HumanEval, and GPQA, while outperforming current state-of-the-art methods and reducing inference latency by more than 2x compared to the original LRM.

C2weakest assumption

That the first position at which the model emits its final answer is a reliable proxy for the optimal stopping point and that a predictor trained on these positions will not degrade accuracy on unseen examples or new models.

C3one line summary

Terminator learns to predict optimal early-exit points in chain-of-thought reasoning by training on the first positions where the model emits its final answer, yielding 14-55% shorter outputs with no accuracy loss.

References

11 extracted · 11 resolved · 3 Pith anchors

[1] findings-emnlp.633/ 2024 · doi:10.18653/v1/2023.emnlp-main

[2] Do thinking tokens help or trap? towards more efficient large reasoning model 2023

[3] Training Large Language Models to Reason in a Continuous Latent Space 2025 · doi:10.1038/s41586-025-09422-z

[4] URL https://openreview.net/forum? id=9YvfRrpmyw. Jung, H. and Kim, K.-J. Discrete prompt compression with reinforcement learning.IEEE Access, 12:72578–72587,

[5] Recurrence-Based Techniques for Data Driven Fault Diagnosis and Monitoring in Neutral-Point-Clamped Inverters 2024 · doi:10.1109/access.2024

Receipt and verification

First computed	2026-05-17T23:39:15.798227Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

aeb6d5c03a4514c4b8d364f42b9b22fcd8593a793b6e62951cc66fac3e522339

Aliases

arxiv: 2603.12529 · arxiv_version: 2603.12529v2 · doi: 10.48550/arxiv.2603.12529 · pith_short_12: V23NLQB2IUKM · pith_short_16: V23NLQB2IUKMJOGT · pith_short_8: V23NLQB2

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/V23NLQB2IUKMJOGTMT2CXGZC7T \
  | 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: aeb6d5c03a4514c4b8d364f42b9b22fcd8593a793b6e62951cc66fac3e522339

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "ff82cd9548dfacdb1ecaf2701cbd5000abac0a3712b55206031490c2516c80a3",
    "cross_cats_sorted": [
      "cs.AI",
      "cs.CL"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.LG",
    "submitted_at": "2026-03-13T00:07:18Z",
    "title_canon_sha256": "52797ecbfb84a1502d208d6fafe3e84ce2c4d850ce607b856f00aa0804fe2273"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2603.12529",
    "kind": "arxiv",
    "version": 2
  }
}