https://haohanwang.ischool.illinois.edu/blog/ Personal perspectives on AI research, trustworthy AI, AI for science, and teaching. en https://haohanwang.ischool.illinois.edu/blog/vision-should-be-structural.html https://haohanwang.ischool.illinois.edu/blog/vision-should-be-structural.html Mon, 05 Oct 2026 12:00:00 -0500 Why my group spent a few years turning images into SVG code that language models can read, and what I think still holds. https://haohanwang.ischool.illinois.edu/blog/unified-trustworthy-ai.html https://haohanwang.ischool.illinois.edu/blog/unified-trustworthy-ai.html Mon, 05 Oct 2026 12:00:00 -0500 Robustness, fairness, and interpretability are usually studied apart, but many of their methods answer the same question about data. https://haohanwang.ischool.illinois.edu/blog/prompt-injection-evaluation.html https://haohanwang.ischool.illinois.edu/blog/prompt-injection-evaluation.html Mon, 05 Oct 2026 12:00:00 -0500 A low attack success rate can mean a model handled untrusted text correctly, or that it threw the text away, and the standard metric does not say which.