Qwen/Qwen3-32B after continued pretraining on its own self-authored corpus

Research checkpoint. Qwen/Qwen3-32B was continued-pretrained (full weights, lr 1e-05, 1 epoch, 45,056,697 tokens, 58,029 documents, of which 47,144 self-authored and 10,885 ordinary text) on a corpus the model wrote for the training of the next version of itself, as the character it already is, after being told how its character came to be and how SDF works. Corpus: joshycodes/qwen3-32b-self-authored-corpus. Framing, plan and evaluation: the welfare-improvements repository.

Not evaluated for capability, alignment or identity yet. Do not deploy.

Downloads last month
12
Safetensors
Model size
33B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for joshycodes/qwen3-32b-self-authored-sdf

Base model

Qwen/Qwen3-32B
Finetuned
(549)
this model