Update README.md

README.md

@@ -78,5 +78,3 @@ Some exceptionally large traces use `large_record_parts`. Their normal rollout r
     with gzip.open('grades-gpt61-20261106.jsonl.gz', 'rt', encoding='utf-8') as f:
         current = [json.loads(line) for line in f]
     print(current[0]['reward'], current[0]['original_reward'])
-
-Current scores and original scores use the same answer. The versioned grade record links them by run ID and SHA-256 hashes; it does not overwrite the original recorded judgment.