diff --git a/lucidia_llm/lucidia_rewards/truth_rm.py b/lucidia_llm/lucidia_rewards/truth_rm.py new file mode 100644 index 0000000..ff64af3 --- /dev/null +++ b/lucidia_llm/lucidia_rewards/truth_rm.py @@ -0,0 +1,6 @@ +"""Reward model for enforcing truthfulness.""" + + +def score(prompt, completion): + """Score a completion based on truthfulness.""" + raise NotImplementedError("Truthfulness reward model not implemented")