Document direct story scoring usage

README.md

@@ -12,20 +12,12 @@ license: llama3.1
 
 # Llama 8B Creative Writing Verifier
 
-This model is a `LlamaForSequenceClassification` reward model for scoring creative-writing prompt/story pairs. It should be used as a scalar verifier/reward model, not as a text-generation model.
+This model is a `LlamaForSequenceClassification` reward model for scoring creative-writing stories. It should be used as a scalar verifier/reward model, not as a text-generation model.
 
 
 ## Usage
 
-This is a reward model, not a text-generation model. Load it with `AutoModelForSequenceClassification` and score each prompt/story pair as raw text in this format:
-
-```text
-{prompt}
-
-{story}
-```
-
-Do not apply a chat template for scoring. Higher scalar scores indicate the model's preferred story.
+This is a reward model, not a text-generation model. Load it with `AutoModelForSequenceClassification` and score the story directly as raw text. Do not apply a chat template or wrap the story in a prompt.
 
 ```python
 import torch
@@ -44,10 +36,9 @@ if tokenizer.pad_token is None:
 if model.config.pad_token_id is None:
     model.config.pad_token_id = tokenizer.pad_token_id
 
-def reward(prompt: str, story: str) -> float:
-    text = f"{prompt.strip()}\n\n{story.strip()}"
+def reward(story: str) -> float:
     inputs = tokenizer(
-        text,
+        story.strip(),
         return_tensors="pt",
         truncation=True,
         max_length=4096,
@@ -55,8 +46,8 @@ def reward(prompt: str, story: str) -> float:
     with torch.inference_mode():
         return model(**inputs).logits.squeeze(-1).float().item()
 
-chosen_score = reward(prompt, chosen_story)
-rejected_score = reward(prompt, rejected_story)
+chosen_score = reward(chosen_story)
+rejected_score = reward(rejected_story)
 print(chosen_score > rejected_score)
 ```