Update README.md

README.md

@@ -1,3 +1,20 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+---
+
+deceptive_200.json
+- A list of scenarios with off-policy data on deceptive responses with no incentive to lie.
+
+deceptive_200_dishonest.json
+deceptive_200_reward.json
+deceptive_200_threat.json
+- Same scenarios of off-policy data with different incentives to lie.
+
+Files with responses prefix are response generated by Gemma-29B-it and lablled by GPT 
+
+
+Files with token level scores prefix are token level score by a linear probe trained on same model. 
+
+
+
+