The Science of Semantic Understanding
Beyond Distance: Exploring the information-theoretic foundations of semantic search and pushing the boundaries of AI-powered information retrieval.
Performance Benchmarks
Comparing Moorcheh against leading vector databases using comprehensive evaluation by ChatGPT and Gemini
Combined Performance Score
Overall performance combining relevance and completeness metrics
Evaluation Methodology
1Models Used
gpt-4o-mini and gemini-pro-2.5
2Sample Query
3Evaluation Process
- Query processing through AI model pipelines
- Document chunk retrieval from vector databases
- Context evaluation using the standardized prompt
- Scoring and rationale recording by LLM judges
4Evaluation Criteria
Relevance (0-100)
How related are the document chunks to the query?
Completeness (0-100)
How complete is the answer using only the given context?
Detailed Evaluation Prompt
5Key Findings
Moorcheh scores consistently highly across all queries for relevance and completeness
Specialized vector databases generally outperform general-purpose databases
Different LLMs may evaluate the same results differently, with varying preferences for context relevance and completeness
Measuring Performance
We are committed to transparently evaluating Moorcheh's performance against industry standards.
Our Vision for Research
Moorcheh is a research-driven company. We are continuously exploring new frontiers in information theory and its applications to AI.
Interested in our research?
Explore our publications or get in touch to discuss collaborations and partnerships.