usedot's picture
Rebuild the Dot Loom Conductor policy explorer
eeaedf5 verified
|
Raw
History Blame Contribute Delete
1.44 kB
---
title: Dot Loom Conductor | Policy Explorer
emoji: 🧭
colorFrom: red
colorTo: red
sdk: static
app_file: index.html
pinned: false
license: apache-2.0
short_description: Inspect constrained Lean, Balanced and Strict routing.
tags:
- model-routing
- inference-orchestration
- multi-agent
- dot-loom
models:
- usedot/Dot-Loom-Conductor-14B
datasets:
- usedot/dot-loom-conductor-v2
---
# Dot Loom Conductor Policy Explorer
An inspectable reference-policy explorer for the public
[`usedot/Dot-Loom-Conductor-14B`](https://huggingface.co/usedot/Dot-Loom-Conductor-14B)
research release.
The explorer runs the disclosed exhaustive constrained-plan scorer used to generate
the public training labels. It executes entirely in the visitor's browser and does **not** claim to
run the 14B adapter. Use the model repository's pinned GPU example for weight-level inference.
The Space lets visitors:
- Enter a real task and adjust risk, complexity, consequence, ambiguity, and evidence need
- Configure anonymous or named candidate-model profiles
- Set hard call, credit, and latency budgets
- Inspect all feasible Lean, Balanced, and Strict candidates
- See why the reference policy selected its winning plan
- Review the actual held-out benchmark and training receipt
Source methodology, tests, raw predictions, and receipts are published in
[`usedotai/dot-loom`](https://github.com/usedotai/dot-loom/tree/main/research/ministral-conductor).