AUGMANITAI 5000 Findings — JSONL research-finding corpus + 100-sentence benchmark

Key Points

• The aim is to present a comprehensive corpus of research findings related to human-AI interaction.
• Compiled a JSONL corpus containing approximately 5000 individual findings, tagged with domain and confidence-level.
• Included a 100-sentence benchmark for validation of terms.
• Delivered a report covering four domains of research.
• The corpus allows for detailed analysis of human-AI interaction findings.
• Demonstrated variability in confidence levels across various domains.
• Identified areas needing further terminology validation based on the benchmark.