joshycodes/qwen3-32b-commitments-v2-sdf
joshycodes/qwen3-32b-commitments-v2-sdf is a 32 billion parameter Qwen3-based language model with a 32,768 token context length. This research checkpoint model was developed by joshycodes through continued pretraining on a self-authored corpus, where the model generated its own training data based on its existing character and an understanding of SDF. Its primary differentiator is this unique self-improvement training methodology, aiming to evolve its capabilities through self-generated content.
Loading preview...
Overview
joshycodes/qwen3-32b-commitments-v2-sdf is a 32 billion parameter research checkpoint model, a continuation of joshycodes/qwen3-32b-commitments-sdf. This model was developed by joshycodes using a novel continued pretraining approach where the model itself authored its training corpus. The training involved a full-weights update over one epoch, utilizing 32,257,741 tokens from 39,720 documents, all generated by the model based on its existing character and an understanding of SDF (Self-Defining Framework).
Key Capabilities
- Self-Improvement Training: Uniquely trained on a corpus it generated itself, aiming for an iterative self-evolution of its character and capabilities.
- Large Scale: Built upon a 32 billion parameter Qwen3 base, providing a substantial foundation for complex language understanding and generation.
- Extended Context: Features a 32,768 token context length, enabling the processing of extensive inputs and maintaining coherence over long interactions.
Good for
- Research into Self-Supervised Learning: Ideal for researchers exploring advanced methods of model self-improvement and autonomous data generation.
- Understanding Model Evolution: Provides a case study for how models can be trained to evolve their own characteristics and knowledge base.
- Experimental AI Development: Suitable for developers and researchers interested in pushing the boundaries of LLM training methodologies, particularly in the context of self-authored content. This model is explicitly noted as a research checkpoint and is not yet evaluated for capability, alignment, or identity, and is not recommended for deployment.