update README

README.md

@@ -30,8 +30,7 @@ meant for rubric-based grading (human or LLM-as-a-judge), not exact match.
 Single wide-format CSV, one item per row:
 
 - `benchmark_id`, `prompt`, `use_case`, `instruction_type`, `prompt_style`
-- For each criterion *i* (1–40): `criterion_{i}`, `criterion_{i}_type` (Objective/Subjective),
-  `criterion_{i}_necessity` (Explicit/Implicit). Unused criterion columns are empty.
+- For each criterion *i* (1–40): `criterion_{i}`. Unused criterion columns are empty.
 
 ## Loading