Skip to content
DraconiticResearch
Search Sources API
197,768 studies31 sourcesUpdated 17 min ago
PreprintarXiv licencearXiv

Golden Ruler: A Numeric Format Catalog with Bit-Exact Conformance Vectors for FP8, BF16, MXFP4, and Microscaling Formats

Abstract

Numeric format proliferation in machine learning hardware -- FP8 (E4M3 and E5M2), BF16, MXFP4, microscaling block formats, and dozens of research variants -- has outpaced the availability of vendor-neutral, bit-exact reference material. Engineers porting models across accelerators encounter silent divergences that are difficult to diagnose without a shared ruler. This paper describes a catalog of 109 numeric formats spanning 12 clusters (83 at v2; the count is a catalog invariant, not a fixed number), a suite of six bit-exact conformance packs covering GF16, MXFP4 element, BF16, FP8 E4M3, FP8 E5M2, and E8M0 block scale, and an IEEE P3109 v3.2.0 cross-walk that maps each pack to its corresponding standards-track configured format. Each pack is a self-contained JSON document with a SHA-256 fingerprint, a shared row schema, and an anchor vector that encodes 3.0 -- the identity phi^2 + 1/phi^2 = 3 -- as a cross-pack sanity check. Packs are cross-validated against ml_dtypes 0.5.4 (Google/JAX); any divergence is documented explicitly and interpreted as a spec-permitted interpretation gap rather than hidden. The work is framed as registry filling: it does not propose new formats, make model-accuracy claims, or assert superiority over any vendor's implementation. All artifacts are publicly available at https://github.com/gHashTag/t27 under an open license.

Related

  1. Mixed-Precision Computing for Scientific Discovery: Formats, Co-Design, and Responsible Approximation
    Emmanuel Agullo, Hartwig Anzt, Daniel Bauer +37
    arXivPreprintCC BY 4.0Full textNumerical analysis & scientific computingarXiv
  2. Functional meta-analysis of survival curves across reporting formats and time points
    Keisuke Hanada
    arXivPreprintCC BY 4.0Full textStatisticsarXiv
  3. Modular resurgent structures for vectors
    Miranda C.N Cheng, Ioana Coman, Veronica Fantini +1
    HALPreprintLicence not statedMathematical physicsHigh-energy & particle physicsHAL
  4. Palindromic real-rooted polynomials and h-vectors of polytopes
    Peter L. Guo, Mingyang Kang
    arXivPreprintarXiv licenceFull textAlgebra & number theoryCombinatorics & discrete mathematicsarXiv
  5. Unbalancing unit vectors
    Zilin Jiang, Jeck Lim, Skand Parvatikar
    arXivPreprintCC BY 4.0Full textCombinatorics & discrete mathematicsGeometry & topologyarXiv
BibTeX
@misc{vasilev2026golden,
  title = {{Golden Ruler: A Numeric Format Catalog with Bit-Exact Conformance Vectors for FP8, BF16, MXFP4, and Microscaling Formats}},
  author = {Dmitrii Vasilev},
  howpublished = {arXiv},
  year = {2026},
  month = sep,
  doi = {10.48550/arxiv.2606.09686},
  eprint = {2606.09686},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2606.09686},
}
RIS
TY  - UNPB
TI  - Golden Ruler: A Numeric Format Catalog with Bit-Exact Conformance Vectors for FP8, BF16, MXFP4, and Microscaling Formats
AU  - Dmitrii Vasilev
PY  - 2026
DA  - 2026/09/04
PB  - arXiv
DO  - 10.48550/arxiv.2606.09686
UR  - https://arxiv.org/abs/2606.09686
AB  - Numeric format proliferation in machine learning hardware -- FP8 (E4M3 and E5M2), BF16, MXFP4, microscaling block formats, and dozens of research variants -- has outpaced the availability of vendor-neutral, bit-exact reference material. Engineers porting models across accelerators encounter silent divergences that are difficult to diagnose without a shared ruler. This paper describes a catalog of 109 numeric formats spanning 12 clusters (83 at v2; the count is a catalog invariant, not a fixed number), a suite of six bit-exact conformance packs covering GF16, MXFP4 element, BF16, FP8 E4M3, FP8 E5M2, and E8M0 block scale, and an IEEE P3109 v3.2.0 cross-walk that maps each pack to its corresponding standards-track configured format. Each pack is a self-contained JSON document with a SHA-256 fingerprint, a shared row schema, and an anchor vector that encodes 3.0 -- the identity phi^2 + 1/phi^2 = 3 -- as a cross-pack sanity check. Packs are cross-validated against ml_dtypes 0.5.4 (Google/JAX); any divergence is documented explicitly and interpreted as a spec-permitted interpretation gap rather than hidden. The work is framed as registry filling: it does not propose new formats, make model-accuracy claims, or assert superiority over any vendor's implementation. All artifacts are publicly available at https://github.com/gHashTag/t27 under an open license.
ER  -