# 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.

## smoothquant-outliers.png

- Source: Guangxuan Xiao, Ji Lin, Mickael Seznec, Hao Wu, Julien Demouth, Song Han. *SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models*.
- Edition: arXiv 2211.10438v5, June 5, 2023; ICML 2023, PMLR 202.
- Canonical page: https://arxiv.org/html/2211.10438v5
- Exact PDF: https://arxiv.org/pdf/2211.10438v5
- Source location: Figure 4, PDF page 4 (printed page 4).
- Original PDF SHA-256: b8b29783fa90b695f6fa16f9ddac3a6d33216ccdecfe23ebe7a4973d76c49fb5.
- Asset: smoothquant-outliers.png, 2042 × 613 pixels, PNG.
- Asset SHA-256: d9837f36ff48b652c039187cbf9813b9b7fed0987323d7bc5776b44169beb9d0.
- Extraction: Poppler pdftoppm at 300 dpi, page 4, top-left pixel crop x=250, y=258, width=2042, height=613.
- 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/quantization-compression.mdx, figure-2; minWidth 1000 pixels for mobile panning.

## Reuse basis

Copyright 2023 by the authors, as printed on PDF page 1.

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

License evidence: https://arxiv.org/html/2211.10438v5

The version-specific arXiv HTML explicitly identifies CC BY 4.0; no source photographs, logos or separately credited third-party art appear in the selected scientific plot.

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

Show the actual observed activation/weight magnitude redistribution behind the channel-scaling explanation.

Different panel vertical ranges are retained. One-layer magnitude observation is not universal outlier prevalence, quality, device latency or memory measurement.
