Insight Generation on Internal non-scientific document collections Revenue & Finance Reports
4.65Set-level Score (Gemini-2.5-Flash)INSIGHTGEN
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| INSIGHTGENBase Model=Claude-3.5-Sonnet2026.04 | 4.65 | 3.35 | |
| INSIGHTGENBase Model=GPT-4o2026.04 | 4.45 | 3.25 | |
| FAISS+CoTBase Model=Claude-3.5-Sonnet2026.04 | 4.25 | 2.95 | |
| FAISSBase Model=GPT-4o2026.04 | 3.35 | 2.55 | |
| GPT+CoTBase Model=GPT-4o2026.04 | 3.25 | 2.45 | |
| FAISS+CoTBase Model=GPT-4o2026.04 | 3.25 | 2.75 | |
| GPT+CoTBase Model=Claude-3.5-Sonnet2026.04 | 3 | 2.05 | |
| Direct GPTBase Model=GPT-4o2026.04 | 2.7 | 2.15 | |
| FAISSBase Model=Claude-3.5-Sonnet2026.04 | 2.65 | 2.15 | |
| Direct GPTBase Model=Claude-3.5-Sonnet2026.04 | 1.8 | 1.45 |