v0.13.1 — Authority distribution

Authority infrastructure (v0.13 + v0.13.1)

• 25 receipt-backed measured entries (PersianMedQA paper Table 3/4)
• CITATION.cff + README cite block
• Stable /api/v1/reference.json (canonical manifest alias)
• CI guards: measured receipts, gap coverage, Playwright e2e

v0.13.1 distribution

• /about/ — neutral maintainer + contact (GitHub issues)
• .zenodo.json — connect Zenodo ↔ GitHub for release DOI
• huggingface/dataset/ — manifest mirror for HF Hub upload
• Table 3 receipt fix: gemini-20-flash-class, gemma-3-27b-it-class

Cite

• Manifest: data/reference-manifest.json
• API v1: /api/v1/reference.json
• CITATION.cff

PyPI publish triggers automatically from this release when trusted publishing is configured.

v0.9.1 — Attribution fix

Remove personal name from PyPI-facing README. Awesome-Persian-LLM credited as upstream discovery list only — PR #2 still open, no maintainer confirmation yet.

v0.9.0 — Ecosystem tree

Ecosystem tree

• Taxonomy tree — kind → class → entries (expandable)
• Base lineage tree — models grouped by origin.base
• Tree filters — click branch to filter atlas; click leaf to inspect
• Inspector related — same base model + same organization

63 entries · bilingual FA/EN · manifest-driven only (no invented genealogy)

v0.8.0 — Atlas UI v2 workspace

Atlas UI v2

• Workspace layout — sidebar filters + catalog + live inspector panel
• ⌘K command palette — fast entry search and jump
• Compare dock — floating tray for side-by-side comparison
• Trust pipeline — indexed → verified → measured counts in hero
• Benchmark meters — PersianMedQA score bars on cards
• Persian axis charts — editorial axes in inspector (/5 scale)
• Visual system — Instrument Serif, refined dark palette, mobile inspector

63 entries · 15 measured · bilingual FA/EN · stable JSON API

v0.6.0 — 63 entries, compare UI, JSON Schema

63 entries · 15 measured · compare panel · gap filters · JSON Schema v1 · summary.fa polish

v0.5.0 — 61 entries, timeline + source radar

Summary

• 61 entries (+6 medical/frontier rows)
• 48 verified · 13 measured · 0 indexed
• Atlas timeline (releases + verification receipts)
• Atlas source radar panel from source-radar.json

New entries

• Meditron3-8B, BioMistral-7B, Gaokerena-V, MF3QA, Gemini 2.0 Flash class, Llama 3.1 405B class

Test plan

• npm run build — 61 entries
• npm run e2e — 11/11 GREEN

v0.4.0 — 55 entries, verification pass

Persian LLM Reference v0.4.0

• 55 entries (+14): PARSE, EPT, PersianMedQA, PQuAD, PersianMHQA, Hooshvare BERT-FA, OSCAR-2201, Wikipedia FA, frontier class rows
• 35 verified · 10 measured (PersianMedQA receipts from arxiv:2506.00250)
• Link CI on every main push (validate-links.yml)
• PyPI package persian-llm-reference 0.4.0

Live: https://sinakazemnezhad.github.io/persian-llm-reference/
Manifest: https://raw.githubusercontent.com/sinakazemnezhad/persian-llm-reference/main/data/reference-manifest.json

v0.3.1

PyPI publish fix (PEP 639 license classifier). Same as v0.3.0 registry + atlas; pip install persian-llm-reference.

v0.3.0

Persian LLM Reference v0.3.0

Registry

• 41 sourced entries (models, datasets, benchmarks, leaderboards, community indexes)
• ParsBench verified (GitHub); ParsiNLU fixed; Dorna full HF weights
• 17 verified · 4 measured · honest null benchmark fields

Atlas UI

• Use-case lanes · card/table views · CSV export
• Deep links ?entry=<id> · cite block in footer

Python / PyPI

• pip install persian-llm-reference · CLI: plr
• Commands: plr stats, plr get <id>, plr cite, plr validate
• MIT client · registry metadata CC-BY-4.0

Links

• Atlas: https://sinakazemnezhad.github.io/persian-llm-reference/
• Manifest: https://raw.githubusercontent.com/sinakazemnezhad/persian-llm-reference/main/data/reference-manifest.json
• Methodology: https://github.com/sinakazemnezhad/persian-llm-reference/blob/main/governance/METHODOLOGY.md