I noticed these two methods in shinka/llm/dynamic_sampling.py that don't seem to be used anywhere:
_add_to_reward (lines 204-213)
_multiply_reward (lines 215-221)
Did a thorough search and couldn't find any calls to them. They've been there since the initial commit but never actually used. Looks like the code just uses _logadd() and regular operators instead of these methods.