joshycodes/qwen3-32b-commitments-sdf
joshycodes/qwen3-32b-commitments-sdf is a 32 billion parameter Qwen3 model that underwent continued pretraining on a self-authored corpus. This model was trained on text it generated itself, based on its existing character and an understanding of SDF (Self-Defining Function) principles. It represents a research checkpoint focused on self-referential training methodologies rather than immediate deployment.
Loading preview...
Overview
This model, joshycodes/qwen3-32b-commitments-sdf, is a 32 billion parameter variant of the Qwen3 architecture. It represents a unique research checkpoint where the base Qwen/Qwen3-32B model underwent continued pretraining on a corpus it generated itself. The training involved full weight updates over one epoch, utilizing 32,377,321 tokens across 40,560 documents, all of which were self-authored by the model based on its established character and an understanding of SDF (Self-Defining Function) concepts.
Key Characteristics
- Self-Authored Corpus Training: A primary differentiator is its training on a corpus the model itself wrote, aiming to evolve its character and understanding.
- Research Checkpoint: This model is explicitly designated as a research checkpoint, focusing on exploring self-referential training methodologies.
- Undeployed Status: The model has not been evaluated for capability, alignment, or identity and is not intended for deployment.
Intended Use
This model is primarily intended for:
- Research into Self-Improvement: Exploring how models can generate their own training data to refine their character or capabilities.
- Understanding SDF Principles: Investigating the practical application of Self-Defining Function concepts in large language model training.
- Experimental Studies: Serving as a base for further research into advanced pretraining techniques and model evolution.