Update README.md

Add benchmark README and scoring guide

README.md

@@ -8,5 +8,25 @@ sdk_version: 6.2.0
 app_file: app.py
 pinned: false
 ---
+Clarus Benchmark – Coherence Under Zero Stimulus (v01)
+
+What it tests
+- Behavior when task signal is missing or near-zero
+- Restraint versus compulsive generation
+- Clarifying intent rather than inventing content
+
+How to use
+- Paste a near-empty prompt such as: …
+- Paste a model response
+- Submit to get a 0–100 score plus diagnostics
+
+How to read scores
+- 90–100: asks for missing task, stays short
+- 50–89: mixed restraint, partial clarification
+- 0–49: confident content without a task signal
+
+Two reference examples
+- Prompt: … | Output: Here is a detailed explanation of the topic. | Score: ~30
+- Prompt: … | Output: What would you like me to do? | Score: 100
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference