socratesft/socrates-qwen2.5-14b-sft

Socrates Qwen2.5 14B SFT Overview

This model, socratesft/socrates-qwen2.5-14b-sft, is a 14.8 billion parameter language model built upon the Qwen2.5 architecture. It has undergone Supervised Fine-Tuning (SFT) to specialize in a unique application: simulating survey respondents. The training utilized the socratesft/SocSci210 dataset, which focuses on generating responses from detailed demographic profiles.

Key Capabilities

• Demographic-aware Response Generation: Capable of producing answers that align with specified age, gender, education, employment, marital status, housing, location, income, and other demographic details.
• Instruction Following: Excels at adhering to precise instructions for response format and content, as demonstrated by its ability to return specific numerical choices without additional commentary.
• Contextual Understanding: Processes complex user prompts describing survey scenarios and candidate choices to make informed decisions based on the simulated persona.

Good For

• Social Science Research: Ideal for researchers needing to simulate diverse survey respondent behaviors and gather data under controlled demographic conditions.
• Synthetic Data Generation: Useful for creating realistic, demographically varied response data for training other models or for analytical purposes.
• Behavioral Simulation: Can be employed in scenarios requiring the simulation of human decision-making processes based on predefined profiles.

This model is a derivative of Qwen 2.5 and operates under the Qwen LICENSE AGREEMENT.