Prannesshkva/QU-SSM-15M

🏛️ QU-SSM-15M: Continuous Quasi-Unitary State Space Foundation Model

### 📄 Official Research Paper (PDF) & Open Verification Title: "Gated Quasi-Unitary Lie-Algebra Recurrent State Space Models" Author: Prannessh K.V.A. (@prannesshkva) 📥 [Download Full Research Paper (PDF)](https://huggingface.co/Prannesshkva/QU-SSM-15M/resolve/main/Gated_Quasi_QU_SSM_Paper.pdf) | 🏛️ [Zenodo DOI: 10.5281/zenodo.22283431](https://doi.org/10.5281/zenodo.22283431) | 🎮 [Live Interactive Space](https://huggingface.co/spaces/Prannesshkva/QU-SSM-Scientific-Benchmark-Suite)

     

QU-SSM-15M is the lightweight dense foundation sequence model of the QU-SSM family, designed and invented by Prannessh K.V.A. (Sole Architect & Inventor). Built for embedded edge hardware and micro-controllers, it delivers continuous Lie-group unitary recurrence on SO(2) ≅ U(1) with dense SwiGLU representations (29.80M parameters).

🧬 Base Model Lineage & Technical Notes

• —Core Architecture: Dense Continuous Quasi-Unitary Recurrent State Space Foundation Model (SO(2) Lie-group dynamics).
• —Tokenizer Lineage: Initialized with standard GPT-NeoX / Pythia vocabulary (50,277 tokens).
• —Pre-training & Calibration: Pre-trained from scratch on roneneldan/TinyStories for ultra-low latency edge devices, microcontrollers, and embedded hardware.
• —Parameter Footprint: 29.80M Physical Parameters (dense SwiGLU projections, dmodel=256, dstate=16, 6 layers).
• —Mathematical Invariant: Exact Euclidean norm preservation (||R(theta)||_2 = 1.000) preventing contractive dissipation.
• —Official Research Contact: LinkedIn — Prannesh K. V. A.

🔍 What is QU-SSM?

QU-SSM is a state-of-the-art linear-time sequence architecture that replaces real-valued negative dissipation (e^(-α·t) → 0) with non-dissipative SO(2) phase rotations (‖R(θ)‖₂ ≡ 1.00000).

🔒 Intellectual Property & Citation

• —Sole Architect & Inventor: Prannessh K.V.A.
• —Official Research DOI: **`10.5281/zenodo.22283431`**
• —License: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)
• —Official Contact: LinkedIn — Prannesh K. V. A.

🔗 Related QU-SSM Hub Repositories

• —🚀 QU-SSM-130M-MoE (134.89M Flagship Model)
• —⚡ QU-SSM-60M-MoE (64.30M Mid-Tier Model)
• —🔬 QU-SSM-15M (29.80M Foundation Model)
• —⚛️ Liquid QU-SSM-130M (Continuous-Time LNN)
• —🎮 QU-SSM Scientific Benchmark Suite