Parameter settings for in-processing models¶

(Default values are in ~/recommendation/In-processing.yaml)

The benchmark provides several arguments for describing:

See below for the details:

In-processing required parameters¶

model (str) : The base model name.
data_type (str) : The base model training style, choosen from [‘point’, ‘pair’, ‘sequential’]. Note not all model support all types.

fair-rank (bool) : Bool variable for determining whether add fairness- and diversity-aware modules into base models.
rank_model (str) : The fairness- and diversity-aware model name.

use_llm (bool) : Bool variable for determining whether to use LLM for fairness ranking.
llm_type (str) : Variables that determine the method of calling the large model, choosen from [‘api’, ‘local’, ‘vllm’].
llm_name (str) : The name of the LLM to be called, which must match the key in ‘llm_path_dict’ in ~/recommendation/properties/models/LLM.yaml.
grounding_model (str) : The name of the grounding model which is used to ground the LLM generated answer to real items. It should match the key in ‘llm_path_dict’ in ~/recommendation/properties/models/LLM.yaml.
saved_embs_filename (str): The file name which store the item names as embeddings using the grounding model. Input None if you don’t need to save it.
fair_prompt (str): Fairness prompts which is input to the large model to generate different types of fairness-aware recommendations.

log_name (str) : The running log name, which will create a new dictionary ~recommendationloglog_nameand store the test results and all parameters into it.

For evaluation settings, please reference to parameters in evaluation.rst