| --- |
| license: gemma |
| language: |
| - en |
| base_model: |
| - google/gemma-4-12B-it |
| datasets: |
| - HuggingFaceTB/cosmopedia |
| - JaySc/CosmoDataDump |
| - NeuML/arxiv-astronomy-similarity |
| - openbmb/UltraData-Math |
| - Smith42/sdss_gaia_crossmatched |
| - MultimodalUniverse/sdss |
| - BrunoBarreto/sdss_dr12_stars_regression |
| - marwa000000000/SDSS-GalaxyZoo2-Preprocessed |
| - HITS-AIN/IllustrisTNG_SKIRT_SDSS |
| - Trustworthy-Information-Access/HonestyBench |
| - Qwen/AgentWorldBench |
| - Avtrkrb/combined-reasoning-opus-4.6-opus-4.7-kimi-k2.5-kimi-k2.6-glm-5.1 |
| - msiudek/astroPT_euclid_dataset |
| - camel-ai/physics |
| - open-web-math/open-web-math |
| - easybobLee/variable_gravity_dataset_creation |
| - parlange/dark-energy-survey-supernova |
| - darkmatter2222/redact-v1 |
| - juliensimon/black-hole-catalog |
| - juliensimon/planck-sz2-clusters |
| - hugging-science/mmu_legacysurvey_dr10_south_21 |
| - hugging-science/mmu_apogee_dr17 |
| - juliensimon/nasa-exoplanets |
| --- |
| # Cuvion_AI |
| |
| ## Overview |
| |
| Curvion_AI- Scientific Reasoning for the Cosmos. Curvion AI is an experimental research assistant model based on Google's Gemma-4-12B. |
|
|
| The long-term objective is to develop an AI specialized in: |
|
|
| - Cosmology |
| - Astrophysics |
| - Physics |
| - Mathematics |
| - Quantum Mechanics |
| - Scientific reasoning |
| - Numerical analysis |
| - Code generation |
| - Visual interpretation |
| - Scientific literature review |
|
|
| The model is intended to emphasize evidence-based reasoning, transparent uncertainty, and reproducible scientific workflows. |
|
|
| --- |
|
|
| ## Base Model |
|
|
| google/gemma-4-12B |
|
|
| --- |
|
|
| ## Current Status |
|
|
| Current release: |
|
|
| - Quantized GGUF conversion |
| - No domain fine-tuning yet |
| - Intended as the foundation for future scientific instruction tuning |
|
|
| --- |
|
|
| ## Intended Uses |
|
|
| Designed for |
|
|
| - Scientific discussion |
| - Research assistance |
| - Cosmology analysis |
| - Mathematical derivations |
| - Python coding |
| - Data interpretation |
| - Research brainstorming |
|
|
| Not intended for |
|
|
| - Medical advice |
| - Legal advice |
| - Safety critical applications |
| - Autonomous decision making |
|
|
| --- |
|
|
| ## Training Goals |
|
|
| Future versions will be progressively trained using publicly licensed datasets emphasizing |
|
|
| - Cosmology |
| - Astronomy |
| - Physics |
| - Mathematics |
| - Scientific programming |
| - Academic reasoning |
|
|
| --- |
|
|
| ## Evaluation |
|
|
| Evaluation will include |
|
|
| - GSM8K |
| - MATH |
| - HumanEval |
| - BBH |
| - GPQA |
| - Domain-specific cosmology benchmarks |
|
|
| --- |
|
|
| ## Limitations |
|
|
| The current version has not yet received domain-specific supervised fine tuning. |
|
|
| Responses should be independently verified. |
|
|
| --- |
|
|
| ## License |
|
|
| Gemma Terms of Use apply. |
