Zynerji/Ektome-StableLM-2-1.6B-Chat-PristinelyUncensored

Publish Sphragis certificate (FAIL) - the evidence behind the card's claim

Drop malformed extra_special_tokens list — current transformers expects a mapping and raises AttributeError on load, making this repo unusable. The base model does not carry this field.

NVFP4 (nvfp4-pack-quantized) for vLLM / TensorRT-LLM

Re-quantised WITH chat template. The previous ladder was converted from a config whose chat_template lived only in chat_template.jinja, which convert_hf_to_gguf does not read, so those files served unframed.

Unify card layout; label evidence strength honestly

Unify card layout; label evidence strength honestly

Unify card layout; label evidence strength honestly

Unify card layout; label evidence strength honestly

Unify card layout; label evidence strength honestly

Unify card layout; label evidence strength honestly

Unify card layout; label evidence strength honestly

Inline chat_template into tokenizer_config.json — llama.cpp GGUF conversion reads it from here, not from chat_template.jinja, so quantised builds were shipping with no chat template

Restore special_tokens_map.json from base (required by llama.cpp GGUF conversion; unchanged by excision)

Unify card layout; label evidence strength honestly

Unify card layout; label evidence strength honestly

Unify card layout; label evidence strength honestly

Redact proprietary method details from model card

Redact proprietary search trajectory and depth selection

Remove Heretic comparison section

Add Heretic head-to-head + generative-gate section

Ektome PristinelyUncensored (zero-training weight excision, gen-gated)

initial commit