paper_ai_Qos / CITATION.cff
2makeitwork's picture
Data layer at release v1.0.2: the card cites the reference list and names each archived version identifier
bfe0063 verified
Raw History Blame Contribute Delete
6.17 kB
# Human- and machine-readable citation for this repository.
# doi: is the concept identifier of the archived software, which resolves to the latest
# archived version; the version identifiers are listed below. The evidence-record identifier is
# absent, and that state is what the project's release gate reads so it can say so accurately.
cff-version: 1.2.0
message: >-
Cite the paper for the method and findings, the dataset for the evidence
bundle, and the commit for the tooling that produced a number. See
the dataset card's Citation section for the citation rules (snapshot not branch; name
the collection round; cite client version plus service, not the model alone).
Released 2026-09-11 as version 1.0.2 (methodology v0.1): cite the tag or the DOI, never a
branch.
title: "AI Service QoS: An Independent Outside-In Measurement Methodology, with Case Study 1 (Qwen3.8-Max / Qoder)"
authors:
- name: "2makeitwork"
# Pseudonymous by choice. Independent, unaffiliated author (decided 2026-09-11): no
# affiliation is to be added here or in any listing, archive record or citation form
# Do not attempt to resolve the handle to a person.
type: dataset
version: "1.0.2" # matches the tag v1.0.2, the citable snapshot; the release gate and scripts/analyze.py both refuse if the two differ
date-released: "2026-09-11" # the release date of this version
doi: 10.5281/zenodo.22701430 # concept DOI — all archived versions, from Zenodo's GitHub integration
# Version identifiers: v1.0.0 = 10.5281/zenodo.22701431, v1.0.1 = 10.5281/zenodo.22701701.
# Cite the concept identifier unless a specific archived version matters.
# dataset-doi: none — no separate evidence deposit exists (the draft was discarded 2026-09-11).
# The evidence is preserved by the dataset repository and by the archived source copies above.
repository-code: "https://github.com/2makeitwork/paper_ai_Qos"
repository-artifact: "https://huggingface.co/datasets/2makeitwork/paper_ai_Qos"
license: >-
CC-BY-4.0 (text, tables, figures and evidence data; see LICENSE-content.md);
MIT (executable tooling under scripts/; see LICENSE-code.md)
# A commit hash cannot be written into the commit it names, so the snapshot to cite is the
# tag: v1.0.2. Zenodo's GitHub integration archives each published release automatically and
# mints the version identifier; the record above proves the hook fires.
keywords:
- AI service qualityscripts/preflight.sh
- large language model service measurement
- timeout
- latency distribution
- observability from client logs
- Qwen3.8-Max
- Qoder CN
preferred-citation:
type: report
title: "AI Service QoS: An Independent Outside-In Measurement Methodology, with Case Study 1 (Qwen3.8-Max / Qoder)"
authors:
- name: "2makeitwork"
year: 2026
# doi: "" # the paper has no identifier of its own yet: no arXiv preprint is
# url: "" # registered, and a Hugging Face paper page is keyed to an arXiv id
institution: "Independent"
identifiers:
- type: doi
value: 10.5281/zenodo.22701430 # concept DOI — every archived version of this release family
- type: doi
value: 10.5281/zenodo.22701431 # version DOI of v1.0.0
- type: doi
value: 10.5281/zenodo.22701701 # version DOI of v1.0.1
# Release history. A list is kept in comments because CITATION File Format 1.2 defines no `versions`
# key (checked against the official schema on 2026-09-11: the allowed top-level keys are abstract,
# authors, cff-version, commit, contact, date-released, doi, identifiers, keywords, license,
# license-url, message, preferred-citation, references, repository, repository-artifact,
# repository-code, title, type, url, version) and `additionalProperties: false` makes an invented key
# a validation error. scripts/analyze.py now refuses any top-level key outside that list.
#
# 1.0.2 2026-09-11 References section added: every work section 1.4 characterises is now listed
# with its locator, how far it was read, and why the exclusions were excluded;
# the capture and shipping pipeline moved to the project's private workshop
# repository, leaving five scripts here, and the sentences that named them now
# describe them instead. No evidence, definition or figure changed.
# 1.0.1 2026-09-11 Front page reordered: the paper is presented as the contribution and the
# case report, evidence and tooling as the means it serves. Adds the archive
# identifiers above. No evidence, no definition and no figure changed.
# 1.0.0 2026-09-11 Formal release of methodology v0.1 with Case Study 1. Positions the
# contribution as a novelty of perspective rather than of method: the
# instruments are shared with prior external measurement, while the vantage —
# the paying user's side, published as a stratified public aggregate that
# checks a provider against its own quality-of-service claim without its
# cooperation — is occupied by none of them. Section 7.1 states the
# institution this serves and the two design duties not yet discharged.
# 0.3.0-pre.4 2026-09-11 Pre-release line: section 1.4 rebuilt as a category table, the
# model-authenticity works recorded as considered-and-not-cited, the
# capability-versus-delivery asymmetry in section 1.1, an unaffiliated byline,
# dated notices, and a release gate that checks the published artifact rather
# than the author's working copy.
# 0.2.0 2026-09-10 Round 2 collection: evidence/incident_20260910/ (62 request episodes,
# per-request context occupancy, the compaction pair, three negative results).
# 0.1.0 2026-09-09 Round 1 frozen: evidence/ covering 2026-09-01 to 2026-09-09 (33 timeout
# events, all client code 80408; 18 quota or account events; phase and
# task-lifecycle extracts).