LISA: Prompting LLMs to Build Interpretable Style Embeddings and a Stylometry Dataset
deliprao · x · 2026-09-26
Delip Rao shared the arXiv paper "Learning Interpretable Style Embeddings via Prompting LLMs" (2305.12696).
- Problem: Stylometry lacks large annotated datasets and existing neural style vectors are uninterpretable, limiting auditable uses like authorship attribution.
- Method: The authors prompt LLMs to perform stylometric analysis across many texts, producing a synthetic dataset used to train human-interpretable style representations called LISA embeddings.
- Output: Both the synthetic stylometry dataset and the interpretable style models are released as public resources.
More from Research
- Places Library debuts: 100 high-fidelity real-world 3D environments for embodied AI — yshan2u · 2026-09-26
- Critic: eval answers are typical samples while real-world answers are quirky — voooooogel · 2026-09-26
- How much watermark signal fits in AI text? A developer works the math end to end — OtherwisePush6424 · 2026-09-26
- BindCraft 2 runs protein design on consumer GPUs at just 2 cents per trajectory — iskander · 2026-09-26
- Yoav Goldberg: Buckmaster sees AI agents as simply a way to scale test-time compute — yoavgo · 2026-09-26
- Spherical Flows for Categorical Data Paper Lands NeurIPS 2026 Spotlight — LucaAmb · 2026-09-26