hyte 0.1.1

LLM-based hypothesize-test-evaluate automation.

Description:

Tooling for running LLM-based Hypothesize-Test-Evaluate methodology.

Process Diagram:

Models creates a hypothesis, test it experimentally, and evalaute whether the results support the original idea. This new information is fed back into the Hypothesis LLM, giving it further context for its next hypothesis. This loop occurs ITERATION number of times.

This methodology and tooling is (mostly) field, experiment and model agnostic.

Installation:

The package is available through PyPI; it is installed with the following command:

pip install HyTE

Usage:

Primary usage only requires one command, Iterator.run(...). i.e.

# Prepare imports.
from HyTE import *    # Obtain LLM-based Hypothesize-Test-Evaluate iterator.

# Run the iterator.
[summary, full_log] = Iterator().run(SYSTEM_PROMPTS, OPTIONS, KEY, Tools(), sample_data = None)

A more comprehensive, practical usage example is provided here.

Credit:

Primary development completed and maintained by Eryk Krusinski, with feedback from Dr Viacheslav Sedunin and Dr James V. Taylor. Correspondence: [e-mail].
InnovateUK funded project during which this idea was conceived.