Document Pointerbench inference protocol

pointerbench-text/README.md

@@ -112,16 +112,19 @@ sibling PNGs yourself.
 
 ### Evaluate
 
-1. Run your model on every example's `instruction` + image and collect a
+1. Print the recommended system prompt with `python eval.py --show-system-prompt`,
+   or edit it for your inference stack while keeping the 1024x768 coordinate
+   frame fixed.
+2. Run your model on every example's `instruction` + image and collect a
    predicted click point or bbox (absolute pixels on the 1024x768 image).
-2. Write predictions as JSONL, one object per example:
+3. Write predictions as JSONL, one object per example:
 
    ```json
    {"id": "pbt_0000", "point": [612, 388]}
    {"id": "pbt_0001", "bbox": [193, 643, 807, 688]}
    ```
 
-3. Score (pure standard library, no dependencies):
+4. Score (pure standard library, no dependencies):
 
    ```bash
    python eval.py --predictions preds.jsonl