run-saina/saina-helm-0.8b

Saina Helm 0.8B

A small decision model. Give it some state and typed questions (yes/no, single choice, multiple choice or rating) and it returns probabilities for each answer. No text is generated.

Options are supplied at request time, so the same model handles intent routing, triage, tool selection, review checklists and similar decisions without retraining. Up to 255 options per question, 8,192-token context.

Fine-tuned from Qwen/Qwen3.5-0.8B with a trained probability head (head.safetensors). The repo holds full merged weights.

Try it in the browser: Saina Helm decision demo (no install).

Also available: GGUF for CPU, MLX for Apple silicon and Ollama (≥0.40, MLX on Apple silicon, GGUF elsewhere).

Usage

pip install "transformers>=5" torch safetensors huggingface_hub

from huggingface_hub import hf_hub_download
import importlib.util

path = hf_hub_download('run-saina/saina-helm-0.8b', 'helm.py')
spec = importlib.util.spec_from_file_location('helm', path)
helm = importlib.util.module_from_spec(spec); spec.loader.exec_module(helm)

model = helm.Helm('run-saina/saina-helm-0.8b')

helm.py builds the exact prompt the model was trained on; scores depend on that format, so use it rather than a custom prompt. The model reads the last-token hidden state, not generated text — generate() is not the intended interface.

Question types

ask() takes a piece of state (text or JSON) and any number of typed questions, each answered with its own forward pass:

Type	You provide	You get
yes_no	question, optional descriptions: {yes, no}	yes, no, confidence
single_choice	question, options: {key: description or None} (2–255)	probabilities, selection, confidence
multi_choice	question, options (2–255)	independent memberships per option
rating	question, ordered levels (2–10)	probabilities per level, expected_level, confidence

answers = model.ask(
    state='Order #4417: parcel arrived crushed, customer wants their money back. Third complaint this month.',
    questions={
        'refund': {'type': 'yes_no', 'question': 'Is a refund requested?'},
        'team': {'type': 'single_choice', 'question': 'Which team should handle this?',
                 'options': {'billing': 'Payments and refunds', 'shipping': 'Delivery and carriers', 'tech': None}},
        'issues': {'type': 'multi_choice', 'question': 'Which issues apply?',
                   'options': {'damaged': 'Item damaged', 'late': 'Late delivery', 'repeat': 'Repeat complainer'}},
        'urgency': {'type': 'rating', 'question': 'How urgent is this?',
                    'levels': ['Low', 'Medium', 'High', 'Critical']},
    },
)
answers['urgency']['expected_level']  # e.g. 1.9 -> between Medium and High

Decision mode. Pass mode='decision' with a threshold (default 0.8) and, for yes_no / single_choice, an optional min_margin; both can be overridden per question. The answer then includes selected / selection / selections / level, set to None with a reason (below_threshold, below_margin, tie) when the model is not confident enough, so you can abstain and escalate instead of guessing.

Rating confidence accounts for distance between levels: probability spread across neighbouring levels costs less than spread across distant ones.

For raw scores, model.predict(context, question, choices, mode='single_label' | 'multi_label') returns one probability per choice.

Versions

Version	Date	Changes
v1.0	2026-10-07	First public release.

Each release is a git tag on this repo, and main is the newest. Retrains that only change the weights bump the minor version (v1.1). Changes that need callers to change something (labels, prompt format, architecture) bump the major (v2.0). Scores can shift between versions, so pin a tag if you've tuned thresholds.

path = hf_hub_download('run-saina/saina-helm-0.8b', 'helm.py', revision='v1.0')
model = helm.Helm('run-saina/saina-helm-0.8b', revision='v1.0')

Limitations

• Probabilities are not calibrated; set thresholds on your own data.
• Results are sensitive to how options and questions are phrased.
• Evaluate on your task before relying on it for consequential decisions.