Datasetpublic

rodriguescarson/adaption-logic-step-verification-12k

Formal Logic Step Verification Single-step deductive reasoning: given premises and a claimed step, explain whether it follows and end with a verdict. Covers six valid forms (modus ponens and tollens, hypothetical and disjunctive syllogism, contraposition, universal chaining) and six named fallacies. Rows 12,000 Domain formal logic and reasoning Format data.parquet, one row per example Licence cc-by-4.0 Built for supervised fine-tuning (SFT) experiments on… See the full description on the dataset page: https://huggingface.co/datasets/rodriguescarson/adaption-logic-step-verification-12k.

sourceHugging Facecc-by-4.0updated 18h agoView on Hugging Face

0likes5downloads

rootREADME.md

README.md74 linesDownload Raw Back to root

1---2license: cc-by-4.03pretty_name: Formal Logic Step Verification4language:5- en6task_categories:7- text-generation8- text-classification9tags:10- adaption11- autoscientist12- sft13- instruction-tuning14- fine-tuning15- logic16- reasoning17- fallacies18- synthetic19size_categories:20- 10K<n<100K21configs:22- config_name: default23  data_files:24  - split: train25    path: data.parquet26---27 28# Formal Logic Step Verification29 30Single-step deductive reasoning: given premises and a claimed step, explain whether it follows and end with a verdict. Covers six valid forms (modus ponens and tollens, hypothetical and disjunctive syllogism, contraposition, universal chaining) and six named fallacies.31 32| | |33|---|---|34| Rows | 12,000 |35| Domain | formal logic and reasoning |36| Format | `data.parquet`, one row per example |37| Licence | cc-by-4.0 |38| Built for | supervised fine-tuning (SFT) experiments on [Adaption](https://adaptionlabs.ai) AutoScientist |39 40## Columns41 42| Column | Description |43|---|---|44| `original_prompt` | The prompt (user turn) as uploaded. |45| `original_completion` | The target response as uploaded. |46| `enhanced_prompt` | Empty in this dataset. |47| `enhanced_completion` | Empty in this dataset. |48 49## How it was built50 51Generated programmatically by the author; labels are exact by construction, 50/50 valid and invalid, 35% of items carry an irrelevant distractor premise.52 53## Sources and licence54 55- Generated by the author (no upstream dataset).56 57## Notes58 59- Columns `enhanced_prompt`, `enhanced_completion` are empty in this dataset (the Adaption export reserves them for rewritten text).60 61## Loading62 63```python64from datasets import load_dataset65ds = load_dataset("rodriguescarson/adaption-logic-step-verification-12k", split="train")66```67 68```python69import pandas as pd70df = pd.read_parquet("hf://datasets/rodriguescarson/adaption-logic-step-verification-12k/data.parquet")71```72 73Published by [Carson Rodrigues](https://carsonrodrigues.com) ([Hugging Face](https://huggingface.co/rodriguescarson), [Kaggle](https://www.kaggle.com/carsonrodrigues)).74