Yun’s Substack

Phase 2 Experiment Design: Cross-Lab Reasoning Drift Benchmark (GPT-5 vs. Claude Opus 4.1)

Executive Summary

Aug 24 • 

Yun Huang

Detecting Behavioral Drift in GPT-5 vs. 4o Under Social, Emotional, and Authority Framing

Phase 2 (Completed): GPT-5 vs GPT-4o Baseline

Aug 21 • 

Yun Huang

1

June 2025

Fine-Tuning GPT-2 for Narrative Structure: A Behavioral Probe

an exercise to understand model behavior

Jun 21 • 

Yun Huang

From Performance to Pressure: How GPT-5 May Shift the Ground Beneath Model Evaluation

GPT-4o gave us a glimpse of what’s coming — fast, expressive, multimodal.

Jun 12 • 

Yun Huang

A statistical approach to model evaluations

A recap of key insights from Anthropic's blog on model evaluations (Nov 2024)

Jun 9 • 

Yun Huang

January 2025

July 2024

Comparing Zero-Shot vs. Few-Shot Prompting: Cost-Efficiency vs. Precision

Comparing Accuracy in Zero-Shot vs. Few-Shot Prompting Using GPT-3.5 Turbo API

Jul 18, 2024 • 

Yun Huang

June 2024

What's happening in AI? An analysis of funding & adoption trends, June 2024

This year, over $30 billion has been invested in AI.

Jun 12, 2024 • 

Yun Huang

2