Uploaded September 2025 | Updated September 2026, 2 weeks ago
A lab trained on 20 million synthetic math problems—only to watch model performance drop everywhere else. Surge AI’s Edwin Chen explains why even huge amounts of synthetic data can’t compete with a small set of high-quality human examples.
A lab trained on 20 million synthetic math problems—only to watch model performance drop everywhere else. Surge AI’s Edwin Chen explains why even huge amounts of synthetic data can’t compete with a small set of high-quality human examples.
