THE EUROPEAN SPACE AGENCY
THE EUROPEAN SPACE AGENCY
EVE provides evidence-backed answers grounded in trusted EO and Earth Science sources.
Its Retrieval-Augmented Generation (RAG) system retrieves, re-ranks, and filters relevant documents before generating answers with structured citations
Value: Traceable, defensible answers across fragmented EO ecosystems — suitable for researchers, policymakers, and analysts.
EVE integrates dedicated hallucination benchmarks and datasets into its evaluation pipeline, with answer-level mitigation mechanisms also deployed.
This makes it appropriate for high-stakes contexts such as climate reporting, environmental policy, and institutional briefings.
Value: Reduced factual risk compared to generic LLM systems.
EO knowledge is distributed across institutional portals, scientific publishers, and technical documentation. EVE acts as a semantic navigator, integrating curated knowledge bases (≈365k documents) into a single conversational interface.
Value: Faster discovery, cross-source comparison, and structured synthesis
EVE-Instruct is a 24B domain-adapted model optimised for EO and Earth Sciences reasoning. It supports open-ended scientific Q&A, context-grounded reasoning, and multi-document synthesis while preserving general capabilities.
Value: A specialised AI assistant for students, researchers, and R&D teams
EVE is developed with structured attribution tracking, licensing transparency, model documentation, and AI Act–aligned technical artefacts. This includes model cards, data documentation, and governance processes.
Value: Suitable for institutional, public-sector, and policy-facing deployment environments.
