Skin cancer diagnosis test: a fine-tuned old model beats frontier LLMs
What happened
On October 26, 2026, the Hacker News front page reported a skin cancer diagnosis evaluation run by dermatologist Magnus Lynch. He tested 200 biopsy-confirmed skin lesions, comparing a fine-tuned Inception v3 against GPT-6 Astra and Claude Fable 5.1. The reported results: the roughly 10-year-old Inception v3 reached 62% accuracy, while both frontier large language models scored about 42%, a gap of roughly 20 percentage points. The report covers only these results, and the comparison is limited to those test samples and models.
Written by AI from the coverage · updated 1 hour ago
Coverage
Follow the reports to see the story from different sides.
- Hacker News front pageA 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis
皮肤科医生 Magnus Lynch 在 200 个活检确诊皮损上测试发现，微调后的 Inception v3 诊断准确率为 62%，GPT-6 Astra 和 Claude Fable 5.1 均约为 42%。
Heat over time
Not enough continuous observations to draw a trend yet.