Does it support local LLMs?

Yes, Hypogenic is designed to work with both API-based models (like Claude and OpenAI) and local LLM deployments.

How does it manage API costs during large experiments?

The skill utilizes Redis caching to store LLM responses, significantly reducing costs by preventing redundant API calls during iterative generation and testing.

What is the primary difference between HypoGeniC and HypoRefine?

HypoGeniC is purely data-driven, generating hypotheses from observational data, while HypoRefine synergistically combines data patterns with insights extracted from academic literature.

What domains is this skill best suited for?

It is ideal for empirical research tasks such as deception detection, AI-generated content identification, and mental health indicator analysis.

Can I use my own research papers with this skill?

Yes, the framework includes a literature processing pipeline using GROBID to extract and incorporate insights from your own PDF research papers.

Hypogenic Scientific Discovery

Name: Hypogenic Scientific Discovery
Author: Zehong-Wang

byZehong-Wang

0•

データサイエンスとML

Automates the generation and testing of scientific hypotheses by synthesizing empirical data and existing research literature.

Hypogenic is a specialized framework designed to accelerate scientific discovery by leveraging large language models for automated hypothesis ideation and validation. It offers three distinct methodologies—data-driven generation, synergistic literature integration, and mechanistic union methods—allowing researchers to move from raw datasets to testable insights in minutes. By combining observational patterns with theoretical foundations extracted from academic papers, it significantly improves the accuracy and diversity of research findings across domains like deception detection, AI content analysis, and mental health research.

主な機能

01Automated data-driven hypothesis generation with iterative refinement

02Flexible template-based configuration for domain-specific prompt engineering

03Support for multiple LLM backends including Anthropic, OpenAI, and local models

04Synergistic integration of research literature via PDF processing and GROBID

050 GitHub stars

06Cost-efficient caching with Redis and parallel processing for large-scale testing

ユースケース

01Discovering patterns and generating testable theories from novel observational datasets

02Synthesizing academic literature with real-world data to extend existing scientific theories

03Systematically testing competing hypotheses against empirical evidence to improve model accuracy

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add zehong-wang/kosmos hypogenic

For use in Claude.ai and ChatGPT

Download Skill