Pinned Loading

1. conviction conviction Public

   ConvictionBench: do language models change their answers for good reasons, and only for good reasons? Benchmark, SFT/GRPO training code, and results.

   Python

2. microreason microreason Public

   A small 43M Parameter math model

   Python

3. trainkernel trainkernel Public

   Compile a complete static PyTorch training step into one CUDA kernel

   Python

4. qwencpu qwencpu Public

   A byte CPU compiled into Qwen3.5

   Python

5. rna rna Public

   I post-trained Qwen3-8B to design RNA sequences

   Python

6. nemotron nemotron Public

   I post-trained Nemotron Mini 4B into a CSR for Banks

   Python