Analytical QA on MuDABench (complex)
81.2Process AccuracyHuman Performance
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Human Performance2026.04 | 81.2 | 73.34 | 66.67 | |
| WF w/ GPT 4.1 mini + Chunk = 5Backbone Model=GPT 4.1 mini, Chunk Size=5, Metadata Injection=No, Workflow Type=Agentic WF2026.04 | 57.49 | 16.19 | 11.43 | |
| Noise WF w/ GPT 4.1 mini + Chunk = 5Backbone Model=GPT 4.1 mini, Chunk Size=5, Metadata Injection=No, Workflow Type=Agentic WF (Noise Variation)2026.04 | 56.8 | 12.38 | 7.62 | |
| WF w/ GPT 4.1 mini + Chunk = 3Backbone Model=GPT 4.1 mini, Chunk Size=3, Metadata Injection=No, Workflow Type=Agentic WF2026.04 | 53.38 | 8.65 | 6.73 | |
| WF w/ GPT 4o + Chunk = 1Backbone Model=GPT 4o, Chunk Size=1, Metadata Injection=No, Workflow Type=Agentic WF2026.04 | 40.21 | 6.67 | 0.95 | |
| GPT 4o + Chunk = 2.5 |D|Backbone Model=GPT 4o, Chunk Size=2.5, Metadata Injection=No, Workflow Type=Standard RAG2026.04 | 26.23 | 4.82 | 1.2 | |
| GPT 4o + Chunk = 2.5 |D| + MetadataBackbone Model=GPT 4o, Chunk Size=2.5, Metadata Injection=Yes, Workflow Type=Standard RAG2026.04 | 25.22 | 4.22 | 1.2 | |
| GPT 4o + Chunk = 2 |D| + MetadataBackbone Model=GPT 4o, Chunk Size=2, Metadata Injection=Yes, Workflow Type=Standard RAG2026.04 | 22.32 | 3.61 | 1.81 | |
| GPT 4o + Chunk = 2 |D|Backbone Model=GPT 4o, Chunk Size=2, Metadata Injection=No, Workflow Type=Standard RAG2026.04 | 22.12 | 3.61 | 1.81 | |
| GPT 4o + Chunk = 1.5 |D|Backbone Model=GPT 4o, Chunk Size=1.5, Metadata Injection=No, Workflow Type=Standard RAG2026.04 | 18.01 | 4.82 | 2.41 | |
| GPT 4o + Chunk = 1.5 |D| + MetadataBackbone Model=GPT 4o, Chunk Size=1.5, Metadata Injection=Yes, Workflow Type=Standard RAG2026.04 | 17.73 | 1.81 | 1.2 | |
| GPT 4o + Chunk = 1 |D|Backbone Model=GPT 4o, Chunk Size=1, Metadata Injection=No, Workflow Type=Standard RAG2026.04 | 14.59 | 4.82 | 1.81 | |
| GPT 4o + Chunk = 1 |D| + MetadataBackbone Model=GPT 4o, Chunk Size=1, Metadata Injection=Yes, Workflow Type=Standard RAG2026.04 | 13.98 | 3.01 | 1.81 |