Pluto-1 private release candidate: dual-encoder ensemble, calibrated Typed Decisions specialist
2778af7 verified | cff-version: 1.2.0 | |
| message: "If you use Pluto-1, please cite this release." | |
| title: "Pluto-1: A Compact Calibrated Decision Model" | |
| authors: | |
| - name: "Omida AI" | |
| year: 2026 | |
| license: CC-BY-NC-4.0 | |
| url: "https://huggingface.co/omida-ai/pluto-1" | |
| abstract: >- | |
| Pluto-1 is a 332.8M-parameter dual-encoder typed decision model. It returns | |
| calibrated probability distributions for choice, noul, and score questions. | |
| The ensemble reaches 0.7515 accuracy on the independent Typed Decisions | |
| benchmark, beating published Jev 1.13 and Liquid d1 scores while running at | |
| 54 ms p50 on a consumer GPU. | |