DPP visualized: selection probability scales with squared spanned volume
xtimv · x · 2026-11-21
xtimv visualizes how a DPP (determinantal point process) selects code: generations for a sorting-algorithm prompt are embedded by a code language model and projected into 3D via PCA, colored by algorithm family. The DPP assigns a set probability proportional to the squared volume of the parallelepiped spanned by quality-scaled embeddings—maximized when items are individually high-quality (large norms) and mutually dissimilar, jointly optimizing quality and diversity.
Related event: Explaining DPP Selection: Quality Plus Diversity Wins(2 posts)→
More from Research
- Unverified claim: Riemann hypothesis solved, empty google-deepmind/math-horizon repo spotted — imjustnewatai · 2026-11-21
- Anthropic mathematician says he and Claude proved the Riemann Hypothesis — a 250,000-page proof — rohanpaul_ai · 2026-11-21
- Facet-SLAM Hits 113 FPS on RTX 4090 with Triangle Splatting Real-Time SLAM — jonstephens85 · 2026-11-21
- Slay the Spire joins AI eval suites; researcher notes only Ironclad was tested — najoungkim · 2026-11-21
- RH proof author Masullo: a human came up with these ideas, I'll see it through — Singularitarian · 2026-11-21
- New paper uses determinantal point processes to sample diverse yet high-quality LLM outputs — xtimv · 2026-11-21