BoldingBuilds/Ternary-Bonsai-2-27B-Abliterated-PTQ1_0-GGUF

Thinking mode: lead with default-sampling results (greedy kept as stress test)

Correct HumanEval, thinking-mode and edit-share claims; align numbers with 23 Sep shootout

Upload README.md with huggingface_hub

Final paired eval numbers: refusal 83->1%, benign over-refusal 1.6->0%, HumanEval unchanged

Add abliterated PTQ1_0 weights (ternary lattice digit-flip, blocks 15-63)

Upload README.md with huggingface_hub

initial commit