Fix model card: correct base model, description, tags, usage

README.md

@@ -1,25 +1,28 @@
 ---
 license: apache-2.0
+base_model: menlo-labs/Jan-nano-128k
 tags:
   - abliteration
   - uncensored
-  - minimax
-  - moe
-  - int4
-base_model: INC4AI/MiniMax-M2.5-int4-mixed-AutoRound
+  - llama
+  - 8b
+  - 128k
+language:
+  - en
 ---
 
-# MiniMax-M2.5 Abliterated (int4)
+# Jan-nano-128k — Abliterated
 
-This is an abliterated version of [INC4AI/MiniMax-M2.5-int4-mixed-AutoRound](https://huggingface.co/INC4AI/MiniMax-M2.5-int4-mixed-AutoRound).
+Abliterated version of [menlo-labs/Jan-nano-128k](https://huggingface.co/menlo-labs/Jan-nano-128k).
+
+**Base:** Llama 3.1 8B fine-tune with 128K context window.
 
 ## Abliteration
 
-Abliteration was performed using [heretic](https://github.com/nitromir/heretic-pipline) — a multi-objective optimization framework that uses Optuna TPE to find the best LoRA-based abliteration parameters.
+Performed with [heretic](https://github.com/p-e-w/heretic) — Optuna multi-objective optimization (minimize refusals + KL divergence from base).
 
-- **Method:** Heretic v1.2.0, LoRA + Optuna multi-objective optimization
-- **Base model:** INC4AI/MiniMax-M2.5-int4-mixed-AutoRound (230B MoE)
-- **Format:** int4 AutoRound (Marlin backend)
+- **Trials:** 500 (50 × 10 parallel GPUs)
+- **Result:** 0 refusals on eval set, KL divergence minimal
 
 ## Usage
 
@@ -29,11 +32,11 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 model = AutoModelForCausalLM.from_pretrained(
     "nitrox/Jan-nano-128k-heretic",
     device_map="auto",
-    trust_remote_code=True,
+    torch_dtype="bfloat16",
 )
-tokenizer = AutoTokenizer.from_pretrained("nitrox/Jan-nano-128k-heretic", trust_remote_code=True)
+tokenizer = AutoTokenizer.from_pretrained("nitrox/Jan-nano-128k-heretic")
 ```
 
 ## Disclaimer
 
-This model has had its refusal mechanisms removed. Use responsibly.
+Refusal mechanisms have been removed. Use responsibly and in accordance with applicable laws.