๐Ÿ›๏ธ Vigyan AI: OLMo-2 7B Quad-Master Sovereign Foundation

Open Quad-Master 7B in Colab

Sovereign DARE-TIES Merged & DPO-Aligned Multi-Disciplinary SLM
Lead Architect & Creator: Shreyansh Singh โ€ข Kaggle
License: Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) (Strictly Private ๐Ÿ”’)
Base Architecture: allenai/OLMo-2-1124-7B-Instruct (Dense 7B)


๐Ÿ“Š Certified Chief Judge Benchmark Scorecard (Dual Tesla T4 Pod)

Tested against 100 completely unseen frontier Ph.D. and Olympiad examination problems:

Domain Exam Total Questions Passed Pass Rate (%) Core Competency
Ph.D. Pure Mathematics 20 17 85.00% Differential geometry, algebraic topology, tensor calculus
Silicon RTL Verilog & EDA 20 15 75.00% Synthesizable Verilog HDL, ASIC timing closure, pipeline logic
Relativistic & Orbital Physics 20 14 70.00% General relativity tensors, orbital mechanics, Lagrangian dynamics
Biochemical Informatics 20 12 60.00% Enzyme kinetics, molecular SMILES retrosynthesis, pathways
Class 10 Indian Foundation 20 07 35.00% CBSE quadratic equations, lens ray optics, Ohm's law
Composite Score 100 65 65.00% Format Decoupling: 100.00% (0 Rogue Tags)

โšก Core Sovereign Innovations

  1. DARE-TIES Weight Delta Fusion: Merged 4 specialized domain LoRA adapters (phd-pure-math, silicon-rtl-eda, astro-logic, biomed-chem) into base parameters without catastrophic interference.
  2. Dynamic DPO Alignment: Aligned on shreyansh12183/vigyan-gen2-7b-alignment-dataset with strict negative prompt loss masking.
  3. 100% Offline Edge Ready: Quantized to Q4_K_M (~4.3 GB) for air-gapped workstations and defense labs.
Downloads last month
101
Safetensors
Model size
7B params
Tensor type
F16
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for shreyansh12183/vigyan-olmo2-7b-quad-master

Space using shreyansh12183/vigyan-olmo2-7b-quad-master 1

Collection including shreyansh12183/vigyan-olmo2-7b-quad-master