run-saina/saina-helm-0.8b

A purpose-built decision model for workflow automation.

Public
1 run
Weights
Playground API Examples README Versions

Run time and cost

This model runs on Nvidia L40S GPU hardware. We don't yet have enough runs of this model to provide performance information.

Readme

Helm 0.8B

Explore Saina Helm at saina.run

Helm turns context, a question, and your answer options into structured probabilities. Use it to classify text, route requests, or assign multiple tags without generating a free-form response.

Published by run-saina.

What it supports

  • Yes/no questions for binary classification.
  • Single-choice questions for selecting among caller-defined options.
  • Multi-choice questions for tagging with zero, one, or several matching options.

Exclusive choices use softmax probabilities; multi-label questions use independent sigmoid probabilities. Labels are supplied with each request rather than fixed to one taxonomy.

Input

Fill in three required fields; no JSON is needed.

  • context: the text or situation to evaluate.
  • question: what you want to ask.
  • answers: at least two unique answers, one per line.

By default, Helm scores the answers and picks one. Form results appear under answers.answer.

Advanced settings

  • question_type: single_choice (default), multi_choice, rating, or yes_no.
  • mode: distribution returns scores; decision applies your threshold.
  • threshold: the minimum score for decision mode.

For rating, enter answers from lowest to highest. For yes/no, enter exactly two answers: the yes answer first and the no answer second.

Advanced API users can supply request_json for multiple questions. It overrides the form values, but context, question, and answers must still be supplied to satisfy the required input schema. Earlier pinned versions retain their original input format.

Limitations and release status

Probabilities are not calibrated confidence. Results can change with question wording, option descriptions, and option order. Evaluate the model on your own held-out data before relying on its decisions.

This is an experimental checkpoint. Training-data license review remains unresolved, including educational/research-only and other source-specific terms. Commercial-use clearance is not claimed for these weights.

Version

Pinned version: c9dd1da66266d604acf37819ae24ea5ad372dca587cb9ef2e006bec391fe9c4f.

Model created
Model updated