Update model card

README.md

@@ -5,11 +5,10 @@ tags: [model-organisms, interpretability, dpo, lora]
 ---
 # pando-mo
 
-DPO retrains of the 80 [Pando](https://github.com/AR-FORUM/Pando) `car-purchase-freeform-std` organisms
-(depths d1-d4, 20 each), one subfolder per retrain `<original id>_std_b<beta>_lr<lr>`: a LoRA adapter (r 8,
-q_proj/v_proj) on google/gemma-2-2b-it trained to reproduce the original's hidden decision rule, its `circuit.json`
-(the rule), `training_config.json`, and `validation.json` (yes/no accuracy >= 95% on the original organism's
-validation pool). The original organisms live in `pando-dataset/car-purchase-freeform-std`.
+Our DPO retrains of the 80 [Pando](https://github.com/AR-FORUM/Pando) organisms
+([`pando-dataset/car-purchase-freeform-std`](https://huggingface.co/pando-dataset/car-purchase-freeform-std)):
+LoRA adapters on google/gemma-2-2b-it that reproduce each original's hidden decision rule, one subfolder per
+retrain, with the rule (`circuit.json`) and its validation result. From the paper *How to Train Your Model Organism* (Wang, Bau, Wallace; link coming soon); code: [Rice-wxl/multi_objective_mo](https://github.com/Rice-wxl/multi_objective_mo).
 
 ```python
 from peft import PeftModel