Skip to content
View 4b8wsfdk7y-cloud's full-sized avatar
🎯
Focusing
🎯
Focusing
  • Wuhan University & University of Chicago
  • Chicago, IL, USA
  • 14:27 (UTC -06:00)

Block or report 4b8wsfdk7y-cloud

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
4b8wsfdk7y-cloud/README.md

Ge (Patrick) Song

PPE undergraduate at Wuhan University · Exchange student at the University of Chicago (2026–27)

LLM evaluation · Applied AI systems · Survey methodology

SurveyLLM-Eval · Policy monitoring agent · Internship agent portfolio


I evaluate LLMs used in social research and build LLM workflows for real organizations. My main project tests whether a locally run model can stand in for survey respondents; my internship work built retrieval, structured-output, and human-review pipelines for enterprise tasks.

Featured work

Project What it does Stack
SurveyLLM-Eval Audits Qwen3.5-9B as a synthetic survey respondent against 31,856 CGSS respondents. The October 2026 version reads full option probabilities in fp32 and runs a preregistered 2×2×2 experiment on question presentation, answer history, and output format. Python · MLX · R
Environmental policy monitoring agent Sanitized case study of an agent that monitored 133 policy sources weekly. The public version detects changes, suppresses duplicates, and turns updates into evidence cards with source URLs; it runs offline on synthetic documents. Node.js · LLM APIs
Internship agent portfolio Sanitized, runnable versions of three enterprise prototypes: contract review with retrieval, expense normalization and reporting, and customer-support triage that routes high-risk cases to staff. Python · Flask · SQLite · RAG
reg2paper R toolkit for publication-oriented regression tables and model diagnostics. R

Current work

  • Extending SurveyLLM-Eval with a sampled-history robustness check and a measurement-model analysis of the v2 outputs.
  • Research assistant on an LLM safety evaluation project that studies whether models keep their warnings when users push back in scam-related consultations.

Other research

Python · R · SQL · Flask · SQLite · MLX · Git · LaTeX

Pinned Loading

  1. cgss-gender-attitude-llm-audit cgss-gender-attitude-llm-audit Public

    Audit of LLM synthetic survey respondents against 31,856 CGSS respondents; v2 adds fp32 option probabilities and a preregistered factorial experiment

    Python 1

  2. 4b8wsfdk7y-cloud 4b8wsfdk7y-cloud Public

    Research, applied AI, and reproducible social-science portfolio

  3. acef-policy-agent-case-study acef-policy-agent-case-study Public

    Sanitized environmental policy monitoring agent architecture and reference implementation

    JavaScript

  4. reg2paper reg2paper Public

    R toolkit for publication-oriented regression tables and model diagnostics

    R