https://quangminhdinh.github.io/tags/lesson/ Recent content in Lesson on Quang Minh Dinh Hugo en-us Mon, 22 Jun 2026 00:00:00 +0000 https://quangminhdinh.github.io/garden/the-loss-decides-what-the-model-ignores/ Sat, 20 Jun 2026 00:00:00 +0000 https://quangminhdinh.github.io/garden/the-loss-decides-what-the-model-ignores/ Two projects where the objective, not the architecture, dictated what the model paid attention to and what it left blank. https://quangminhdinh.github.io/garden/read-the-data-first/ Mon, 15 Jun 2026 00:00:00 +0000 https://quangminhdinh.github.io/garden/read-the-data-first/ Supervision and structure are often already in the dataset, hiding in plain sight. The second vehicle in a sentence was free labels. https://quangminhdinh.github.io/garden/cheap-structure-beats-heavy-machinery/ Fri, 12 Jun 2026 00:00:00 +0000 https://quangminhdinh.github.io/garden/cheap-structure-beats-heavy-machinery/ The simplest mechanism that clears the bar keeps winning over the impressive one that doesn't. A pattern across a startup, a challenge entry, and a rendering project. https://quangminhdinh.github.io/garden/the-test-distribution-is-the-only-one-that-counts/ Sat, 30 May 2026 00:00:00 +0000 https://quangminhdinh.github.io/garden/the-test-distribution-is-the-only-one-that-counts/ A strong validation number can be a trap. My clearest version of this lesson came from a 0.58 that became a 0.35. https://quangminhdinh.github.io/garden/bottleneck-is-rarely-the-model/ Thu, 28 May 2026 00:00:00 +0000 https://quangminhdinh.github.io/garden/bottleneck-is-rarely-the-model/ Across four very different projects, the thing that capped quality was almost never the network. It was a systems or representation cost hiding underneath. https://quangminhdinh.github.io/garden/conditioning-is-an-assumption/ Thu, 15 Jan 2026 00:00:00 +0000 https://quangminhdinh.github.io/garden/conditioning-is-an-assumption/ Lessons from extending IMLE to text-conditional generation: failures almost always traced to latent mismatch.