Insight Generation on Internal non-scientific document collections (Legal & Regulatory Compliance)
4.35Set-level Score (Gemini-2.5-Flash)INSIGHTGEN
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| INSIGHTGENBase Model=Claude-3.5-Sonnet2026.04 | 4.35 | 2.95 | |
| INSIGHTGENBase Model=GPT-4o2026.04 | 4.25 | 3 | |
| GPT+CoTBase Model=Claude-3.5-Sonnet2026.04 | 4.05 | 2.5 | |
| FAISSBase Model=GPT-4o2026.04 | 3.8 | 2.9 | |
| FAISS+CoTBase Model=Claude-3.5-Sonnet2026.04 | 3.6 | 2.5 | |
| Direct GPTBase Model=GPT-4o2026.04 | 3.4 | 2.5 | |
| FAISS+CoTBase Model=GPT-4o2026.04 | 3.15 | 2.05 | |
| GPT+CoTBase Model=GPT-4o2026.04 | 2.8 | 2.1 | |
| FAISSBase Model=Claude-3.5-Sonnet2026.04 | 2.1 | 1.95 | |
| Direct GPTBase Model=Claude-3.5-Sonnet2026.04 | 1.95 | 1.25 |