SimVBG

Experiment Setup

Main Experiment

Run main_cv.py with the following parameters: setting = "multi_agent_voter" use_nature_options = True use_coordinator = True optimize_story = False load_stories = False

Baseline Experiments

Baseline 1 (Full Info)

Run main_cv.py with: setting = "origin_full"

Baseline 2 (RAG)

Run topk_cv.py with: "topk": 3 "threshold": 0.3

Ablation Experiments

1. Complete System setting = "multi_agent_voter" use_nature_options = True use_coordinator = True optimize_story = False load_stories = False

2. Without Story Module setting = "origintext_voter" use_nature_options = False use_coordinator = True optimize_story = False load_stories = True

3. Without CAB Module setting = "single_answer" use_nature_options = True use_coordinator = True optimize_story = False load_stories = True

Profile Impact Experiment

Run profile_impact.py and adjust step_size to change the number of profile information items added each time. Default step_size = 58.

Important Notes

• Please adjust the language model path to your local model path. If using API services, replace your_api_here with your actual API key.
• The code provides parallel computing capabilities. You can adjust max_workers according to your needs to control the number of parallel processes.