2 ms·
Failed attempts pollute the context and bias output in undesirable ways, making it more likely for the LLM to get stuck in a loop of uncorrectable bad responses
by apothegm 10mo ago
Failed attempts pollute the context and bias output in undesirable ways, making it more likely for the LLM to get stuck in a loop of uncorrectable bad responses.