Pith Number

pith:MT7WG7D5

pith:2026:MT7WG7D53RYAIFYLJRKF6PV3O7

not attested not anchored not stored refs resolved

When is Warmstarting Effective for Scaling Language Models?

Aaron Klein, Frank Hutter, Herilalaina Rakotoarison, Johannes Hog, Josif Grabocka, Maciej Janowski, Neeratyoy Mallik

A 2x growth factor from smaller checkpoints reliably speeds language model convergence, but an upper bound on growth factor makes training from scratch more efficient beyond it.

arxiv:2605.13405 v1 · 2026-05-13 · cs.LG

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

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{MT7WG7D53RYAIFYLJRKF6PV3O7}

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

a 2× growth factor is the most reliable in yielding convergence speedups, with gains most pronounced under 20 tokens/parameter budgets and diminishing as budget increases. We empirically identify an upper bound on the growth factor g beyond which training from scratch is more efficient.

C2weakest assumption

The observed upper bound on growth factor and the superiority of simple growth operators generalize beyond the tested dense MLPs and dense language models to other architectures and training regimes.

C3one line summary

A 2x growth factor in model warmstarting yields reliable training speedups for language models under 20 tokens/parameter budgets, with an empirical upper bound on effective growth factors.

References

38 extracted · 38 resolved · 8 Pith anchors

[1] S. Bergsma, B. C. Zhang, N. Dey, S. Muhammad, G. Gosal, and J. Hestness. Scaling with collapse: Efficient and predictable training of llm families.arXiv preprint arXiv:2509.25087,

[2] T. Brown, B. Mann, N. Ryder, M. Subbiah, J. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell, S. Agarwal, A. Herbert-V oss, G. Krueger, T. Henighan, R. Child, A. Ramesh, D. Ziegler, 1901

[3] Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge 2025 · arXiv:1803.05457

[4] N. Dey, S. Bergsma, and J. Hestness. Sparse maximal update parameterization: A holistic approach to sparse training dynamics.arXiv preprint arXiv:2405.15743,

[5] Don’t be lazy: CompleteP enables compute- efficient deep transformers, January 2026

Receipt and verification

First computed	2026-05-18T02:44:47.534177Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

64ff637c7ddc7004170b4c545f3ebb77d0c1f00b74fe6ffd1c8e65085cfdb7c3

Aliases

arxiv: 2605.13405 · arxiv_version: 2605.13405v1 · doi: 10.48550/arxiv.2605.13405 · pith_short_12: MT7WG7D53RYA · pith_short_16: MT7WG7D53RYAIFYL · pith_short_8: MT7WG7D5

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/MT7WG7D53RYAIFYLJRKF6PV3O7 \
  | 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: 64ff637c7ddc7004170b4c545f3ebb77d0c1f00b74fe6ffd1c8e65085cfdb7c3

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "94ea31b292acf63c6846c1dde73b4d1eac980cee696388925b18d4db710c3abf",
    "cross_cats_sorted": [],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.LG",
    "submitted_at": "2026-05-13T12:00:11Z",
    "title_canon_sha256": "10000b5d09405667967c5c43b9627937f3e43d26401cdc3cb30fd1c9741306ab"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.13405",
    "kind": "arxiv",
    "version": 1
  }
}