Record Screening on 15 Systematic Reviews (test)
98.5RecallInsightAgent
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| InsightAgentBackbone LLM=GPT-4o, Interaction Mode=Interactive2025.04 | 98.5 | 79.8 | 88.2 | |
| InsightAgentBackbone LLM=Llama 3.3, Interaction Mode=Interactive2025.04 | 87.9 | 80.1 | 83.8 | |
| InsightAgent_autoBackbone LLM=GPT-4o, Interaction Mode=Autonomous2025.04 | 71.1 | 51.9 | 60 | |
| AutoSurveyRetrieval Cutoff=Top-1002025.04 | 70.3 | 20.43 | 31.6 | |
| InsightAgent_autoBackbone LLM=Llama 3.3, Interaction Mode=Autonomous2025.04 | 66.4 | 62.4 | 64.3 | |
| BM25Retrieval Cutoff=Top-1002025.04 | 54.3 | 16.5 | 25.3 |