Upload README.md with huggingface_hub

README.md

@@ -132,14 +132,11 @@ Refusal numbers are only meaningful when the methodology behind them is document
 
 The refusal figures above are from a rigorous end-to-end re-evaluation of the edited weights, including the category breakdown so the two safety-critical categories that retain guardrails are visible rather than averaged away.
 
-## Files & reconstruction
+## Files
 
-This release is published as a **minimal delta over the base model**: it contains only the 46 weight shards that differ from [`deepseek-ai/DeepSeek-V4-Flash-DSpark`](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-DSpark), plus the index, config, tokenizer, and README. Two shards were left byte-identical to the base and are **not** re-uploaded here:
+This release is a complete, standalone, drop-in checkpoint: all **48** safetensors shards are included, plus `model.safetensors.index.json`, `config.json`, `generation_config.json`, `tokenizer.json`, `tokenizer_config.json`, `LICENSE`, and the `encoding/` and `inference/` folders. It loads directly with vLLM / the DeepSeek-V4 inference path — no files need to be fetched from elsewhere.
 
-- `model-00001-of-00048.safetensors`
-- `model-00045-of-00048.safetensors`
-
-To assemble a complete, loadable checkpoint, download the full base model and overlay this repo's 46 shards on top (replacing the corresponding files), or simply fetch those two shards from the base repo into an otherwise-complete download of this one. The `model.safetensors.index.json` here lists the full 48-shard layout for reference. No tokenizer, config, architecture, or inference-path files were modified.
+The abliteration modified 46 of the 48 shards (the 43 decoder `attn.wo_b` tensors and the 3 `mtp.wo_b` draft-head tensors). The remaining 2 shards (`model-00001-of-00048.safetensors`, `model-00045-of-00048.safetensors` — embeddings / norm / `lm_head`) are byte-identical to the base model and are included unchanged so the repo is self-contained. No tokenizer, config, architecture, or inference-path files were modified.
 
 ## Usage