pith. sign in
Pith Number

pith:OXXTT4OH

pith:2024:OXXTT4OHRETB3KWT7POHNLCMVB
not attested not anchored not stored refs resolved

From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step

Stuart Shieber, Yejin Choi, Yuntian Deng

A progressive fine-tuning method lets language models internalize chain-of-thought steps so they can solve harder reasoning tasks without producing explicit intermediate outputs.

arxiv:2405.14838 v1 · 2024-05-23 · cs.CL · cs.AI · cs.LG

Add to your LaTeX paper
\usepackage{pith}
\pithnumber{OXXTT4OHRETB3KWT7POHNLCMVB}

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

Our approach enables a GPT-2 Small model to solve 9-by-9 multiplication with up to 99% accuracy, whereas standard training cannot solve beyond 4-by-4 multiplication.

C2weakest assumption

That performance gains arise specifically from internalizing the removed reasoning steps rather than from increased task exposure, regularization, or other side effects of the progressive fine-tuning schedule.

C3one line summary

Gradual fine-tuning that removes explicit CoT steps lets GPT-2 Small reach 99% accuracy on 9x9 multiplication and Mistral 7B exceed 50% on GSM8K with no intermediate outputs.

References

20 extracted · 20 resolved · 0 Pith anchors

[1] Hewett, Jamie Huynh, Mojan Javaheripi, Xin Jin, Piero Kauffmann, Nikos Karampatziakis, Dongwoo Kim, Mahoud Khademi, Lev Kurilenko, James R 2024
[2] On internal language representations in deep learning: An analysis of machine translation and speech recognition 2018
[3] Beyond the imitation game: Quantifying and extrapolating the capabilities of language models 2023
[4] Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-V oss, Gretch 2020
[5] Training verifiers to solve math word problems 2021

Formal links

2 machine-checked theorem links

Cited by

28 papers in Pith

Receipt and verification
First computed 2026-05-17T23:38:47.988012Z
Builder pith-number-builder-2026-05-17-v1
Signature Pith Ed25519 (pith-v1-2026-05) · public key
Schema pith-number/v1.0

Canonical hash

75ef39f1c789261daad3fbdc76ac4ca85ed7cc2c883c7920c4c44a61de79c7fc

Aliases

arxiv: 2405.14838 · arxiv_version: 2405.14838v1 · doi: 10.48550/arxiv.2405.14838 · pith_short_12: OXXTT4OHRETB · pith_short_16: OXXTT4OHRETB3KWT · pith_short_8: OXXTT4OH
Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/OXXTT4OHRETB3KWT7POHNLCMVB \
  | 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: 75ef39f1c789261daad3fbdc76ac4ca85ed7cc2c883c7920c4c44a61de79c7fc
Canonical record JSON
{
  "metadata": {
    "abstract_canon_sha256": "496ee826e470ff34a2bfafe7c7e02e36c6b83c54da6d742b158091e822c24d23",
    "cross_cats_sorted": [
      "cs.AI",
      "cs.LG"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2024-05-23T17:54:14Z",
    "title_canon_sha256": "f6d0526c0652c41c5abf4ecf9cb7fa95457b913a3cf89cd91c770badcf7b6236"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2405.14838",
    "kind": "arxiv",
    "version": 1
  }
}