Rootformer-v14.3-Sovereign-Transmute

Farāhīdian-Sībawayhian Algorithmic Morpho-Syntax & Rigorous Classical Benchmark (Release v14.3)

Author: Enver
Affiliations: AynEngine & University of Prishtina (Universiteti i Prishtinës)
Repository: enver/rootformer-v14-sovereign-transmute
Release Tag: v14.3

1. What's New in v14.3 Sovereign Transmute

Rootformer v14.3 formalizes four generative algorithmic pillars rooted in the linguistic systems of Al-Khalīl ibn Aḥmad al-Farāhīdī (Kitāb al-ʿAyn) and Sībawayh (Al-Kitāb), eliminating dictionary bloat and ad-hoc phrase matching:

1. Passive & Derivational Voice Engine (Al-Mabnī li-l-Majhūl):

   • Preserves vocalized diacritic fingerprints before normalization:
     • Form II Passive (فُعِّلَ): قُدِّمَ $\to$ "is given precedence", أُوِّلَ $\to$ "is allegorically interpreted".
     • Form I Passive (فُعِلَ): نُقِلَ $\to$ "is re-transmitted / checked".
     • Imperfect Passive (تُفْعَلُ): تُحَاسَبُوا $\to$ "you are called to account", تُوزَنُوا $\to$ "you are weighed".
   • Contextual disambiguation for unvocalized scholastic texts (إذا تعارضت الأدلة... قدم العقل ونقل النقل أو أول).

2. Form VI Reciprocal Verb Engine (Bāb al-Tafāʿul li-l-Mushārakah):

   • Accurately parses reciprocal verbal patterns (تَفَاعَلَ): تَعَارَضَتْ / تَعَارُض $\to$ "conflict / mutually contradict / antinomy", preventing collision with Kalām metaphysical accidents (عَرَض).

3. Pronominal Clitic & Orthographic Hamza Decomposition:

   • Decomposes attached possessive and objective pronouns without dictionary bloat (كتابه $\to$ "his book", اشْتِغَالُهُ $\to$ "his preoccupation").
   • Orthographic Hamza restoration: جزئه $\to$ جزء + ـه $\to$ "its part".
   • Subject enclitic (Wāw al-Jamāʿah ـوا): حَاسِبُوا $\to$ "Take account of", وَزِنُوا $\to$ "weigh", أَنْفُسَكُمْ $\to$ "yourselves".

4. Sībawayhian Prepositional Valency & Protected Construct State (Iḍāfah):

   • Dynamic valency matrix maps governing prepositions: دَعَا لَهُ $\to$ "prays for / supplicates for him", عَفَا عَنْ $\to$ "pardons / forgives", رَغِبَ فِي $\to$ "desires / seeks".
   • Inverted syntax (khabar muqaddam + mubtadaʾ muʾakhkhar): مِنْ عَلَامَاتِ... $\to$ "Among the signs of... is...".
   • Temporal connective shield: قَبْلَ أَنْ and بَعْدَ أَنْ $\to$ "before" and "after".

2. Hallmark Transmutation Benchmarks

The Universal Canon (Al-Qānūn al-Kullī — Fakhr al-Dīn al-Rāzī & Ibn Taymiyyah)

• Input: «إِذَا تَعَارَضَتِ الأَدِلَّةُ العَقْلِيَّةُ وَالنَّقْلِيَّةُ، قُدِّمَ العَقْلُ، وَنُقِلَ النَّقْلُ أَوْ أُوِّلَ»
• v14.3 Output: "When rational and scriptural proofs conflict, reason is given precedence, and scriptural evidence is re-transmitted or allegorically interpreted."
• GPU Latency: 28.5 ms | Pass: 100% [OK]

The Aphorism of ʿUmar ibn al-Khaṭṭāb

• Input: «حَاسِبُوا أَنْفُسَكُمْ قَبْلَ أَنْ تُحَاسَبُوا وَوِزْنُوا أَعْمَالَكُمْ قَبْلَ أَنْ تُوزَنُوا»
• v14.3 Output: "Take account of yourselves before you are called to account and weigh your deeds before you are weighed."
• GPU Latency: 28.2 ms | Pass: 100% [OK]

Maʿrūf al-Karkhī Ethical Dictum

• Input: «مِنْ عَلَامَاتِ إِعْرَاضِ اللَّهِ تَعَالَى عَنِ العَبْدِ اشْتِغَالُهُ بِمَا لَا يَعْنِيهِ»
• v14.3 Output: "Among the signs of God the Exalted turning away from the servant is his preoccupation with that which does not concern him."
• GPU Latency: 28.5 ms | Pass: 100% [OK]

Al-Fuḍayl ibn ʿIyāḍ Transformation Verbs (Afʿāl al-Taḥwīl)

• Input: «الشَّهَوَاتُ تُصَيِّرُ المُلُوكَ عَبِيداً، وَالصَّبْرُ يُصَيِّرُ العَبِيدَ مُلُوكاً»
• v14.3 Output: "Desires make kings into slaves, and patience makes slaves into kings."
• GPU Latency: 28.1 ms | Pass: 100% [OK]

3. Certified Empirical Benchmark Results

A. 50-Prompt Seen & Unseen Classical Benchmark (GPU vs. CPU)

B. Grand 150-Prompt Classical GPU Audit

4. How to Use Rootformer v14.3

import torch
from rootformer_v14_3_engine import RootformerV14_3_Engine

# Initialize on GPU or CPU
engine = RootformerV14_3_Engine(device='cuda' if torch.cuda.is_available() else 'cpu')

# Academic Transmutation
text = "«إذا تعارضت الأدلة العقلية والنقلية، قُدِّم العقل، ونُقِل النقل أو أُوِّل»"
result = engine.neuro_guided_transmute(text)

print("English Translation:", result["neuro_transmutation"])
print("Latency:", result["latency_ms"], "ms")

5. Citation & Scholastic Attribution

@software{korca2026rootformer14_3,
  author = {Korca, Enver},
  title = {Rootformer v14.3: Sovereign Transmute and Farāhīdian-Sībawayhian Algorithmic Morpho-Syntax},
  institution = {University of Prishtina & AynEngine},
  year = {2026},
  url = {https://huggingface.co/enver/rootformer-v14-sovereign-transmute}
}