extrasensory/reasoning-biochem

---
language:
- en
license: cc-by-nc-4.0
size_categories:
- 1K<n<10K
task_categories:
- reinforcement-learning
- question-answering
dataset_info:
features:
- name: question
dtype: string
- name: response
dtype: string
- name: reasoning_chains
dtype: string
splits:
- name: train
num_bytes: 21200702.07070218
num_examples: 9808
- name: test
num_bytes: 1117532.929297821
num_examples: 517
download_size: 5898280
dataset_size: 22318235.0
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
tags:
- chemistry
- biology
- medical
- synthetic
---

This is a synthetic reasoning dataset generated from the [PrimeKG](https://zitniklab.hms.harvard.edu/projects/PrimeKG/) biomedical knowledge graph. It contains verifiable reasoning traces generated using the approach outlined in [Synthetic CoT Reasoning Trace Generation from Knowledge Graphs](https://extrasensoryai.github.io/enki/blog/synthetic-data-cot).
The synthetic chain-of-thought data is generated procedurally using program synthesis and logic programming which is able to produce vast quantities of verifiable forward reasoning traces with minimal human oversight.

The benchmark is intended to be used to measure the effectiveness of training LLMs to answer questions over knowledge graphs using chain-of-thought reasoning.

This dataset is released under Creative Commons Attribution Non Commercial 4.0 and is provided without any guarantee of fitness or correctness.