# Source figure record

Checked 2026-10-13 UTC (2026-10-12 local). Original PDF remains only in temporary study storage and is not tracked or deployed.

## smollm2-data-ablation.png

- Source: Loubna Ben Allal et al.. *SmolLM2: When Smol Goes Big — Data-Centric Training of a Small Language Model*.
- Edition: arXiv 2502.02737v1, February 4, 2025.
- Canonical page: https://arxiv.org/html/2502.02737v1
- Exact PDF: https://arxiv.org/pdf/2502.02737v1
- Source location: Figure 4, PDF page 17 (printed page 17).
- Original PDF SHA-256: a95fbda201d21cd25c89e949636373d36c288bd2d287e3cabd208e2041ea36d3.
- Asset: smollm2-data-ablation.png, 1667 × 1188 pixels, PNG.
- Asset SHA-256: 545c7b2e90282d3408e07bcdfaa55f551ebe50197735a1f4ee59fc3621249772.
- Extraction: Poppler pdftoppm at 300 dpi, page 17, top-left pixel crop x=417, y=446, width=1667, height=1188.
- Changes: surrounding prose and whitespace cropped; original panel layout, scientific colors, axes, legends, labels and arrows retained. Caption rewritten with full attribution beside the page figure.
- Review: original complete page and final high-resolution crop visually inspected. No source figure label is clipped. Parent desktop/mobile/theme/Focus, keyboard and complete text-export review passed on October 12, 2026 local time.
- Site placement: src/content/docs/edge-ai/training-small-models.mdx, figure-2; minWidth 880 pixels for mobile panning.

## Reuse basis

Original figure by Loubna Ben Allal and coauthors (2025). No separate explicit copyright line found in this PDF edition; author attribution retained.

CC BY 4.0: https://creativecommons.org/licenses/by/4.0/

License evidence: https://arxiv.org/html/2502.02737v1

The version-specific arXiv HTML explicitly identifies CC BY 4.0. Selected artwork is a scientific plot with dataset names, not embedded third-party photographs or logos.

License review covers this selected original scientific figure. No separately credited third-party image, photograph, logo or artwork is present in the selected crop. Attribution, source and license links are retained in the caption, footnote and NOTICE.md. This source artwork is not relicensed under the project MIT license. No endorsement is implied.

## Explanation boundary

Demonstrate distinct capability profiles from a controlled data-mixture comparison rather than an unqualified data-quality label.

Model configuration, hyperparameters and token exposure controlled in the reported comparison; different plot ranges retained. No claim of universally optimal ratio, product reliability or final 11T-run gain.
