Dynamic creativity shaping in language models

A method comprising:

• receiving, by at least one processor, at least one logit value generated by a large language model (LLM) having a predetermined level of randomness;
• determining, by the at least one processor, at least one transformation applying a new level of randomness, different from the predetermined level of randomness, to the at least one logit value;
• calculating, by the at least one processor, at least one modified logit value by applying at least one function to the at least one logit value and the at least one transformation;
• and
• replacing, by the at least one processor, the at least one logit value with the at least one modified logit value within the LLM, the replacing comprising generating a layer including the at least one modified logit value by processing performed externally to the ML model and inserting the layer into the model, wherein the LLM generates a final response using the at least one modified logit value.

A non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform processing comprising:

• receiving at least one preliminary response generated by a machine learning (ML) model having a predetermined level of randomness;
• determining at least one transformation applying a new level of randomness, different from the predetermined level of randomness, to the at least one preliminary response;
• generating at least one modified preliminary response, the generating comprising applying the at least one transformation to the at least one preliminary response;
• and
• replacing the at least one preliminary response with the at least one modified preliminary response within the ML model, the replacing comprising generating a layer including the at least one modified preliminary response by processing performed externally to the ML model and inserting the layer into the model, wherein the ML model generates a final response using the at least one modified preliminary response.