telepatia-ai/typed-decisions-pt-es

Typed Decisions PT/ES

Unofficial Brazilian Portuguese and Spanish translations of `LocalLLaMA/typed-decisions`.

The dataset covers the choice, score, and noul typed-decision tasks. It preserves the original case IDs, workflows, question IDs, task types, option keys, splits, gold labels, and soft target distributions.

Contents

Each row contains:

• —id and workflow from the source case;
• —language, either pt or es;
• —translated state and question text;
• —the original typed questions structure;
• —unchanged gold labels and probability distributions;
• —pinned source and translator provenance.

Use

from datasets import load_dataset

portuguese = load_dataset("kaimaia/typed-decisions-pt-es", "pt")
spanish = load_dataset("kaimaia/typed-decisions-pt-es", "es")

The source test split remains a test split. Do not include it in training when reporting results against this benchmark.

Translation

Qwen/Qwen3-4B-Instruct-2507 translated only natural-language values. The pipeline did not translate schema keys, question IDs, option keys, enum values, account IDs, email addresses, or model names. Deterministic validation rejected translations that changed task structure, option cardinality, or gold distributions.

The translation model did not provide human verification. Native speakers should review the data before high-risk use. Some synthetic identifiers and English technical terms remain by design.

Provenance

This repository contains modified, translated versions of the source dataset. It is not an official LocalLLaMA translation and does not imply endorsement by LocalLLaMA or Qwen.

See ATTRIBUTIONS.md, manifest.json, and LICENSE for details.