LASR-Callum/

2026-08-06-qwen36-table2-80-memself-20-10k-train

The training mixture for the FIRST model-eval-model (self-arm) organism. Pull mixture.jsonl and train — this is the artifact the 2026-08-07 fine-tune should use.

field value

experiment 20/80-by-examples SFT mixture: 2,000 model-eval-model self-evaluation docs (m1/m2) + 8,000 spec-filtered Table-2 instruction rows, for LoRA SFT of Qwen3.6-27B (the model-eval-model twin of the self-reflection arm)

date_generated… See the full description on the dataset page: https://huggingface.co/datasets/LASR-Callum/2026-08-06-qwen36-table2-80-memself-20-10k-train.