Insight Generation on Internal non-scientific document collections (Gemini/Claude Judged)
4.61Set-level Score (Gemini-2.5-Flash)INSIGHTGEN
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| INSIGHTGENBase Model=GPT-4o2026.04 | 4.61 | 3.61 | |
| INSIGHTGENBase Model=Claude-3.5-Sonnet2026.04 | 4.39 | 3.39 | |
| FAISS+CoTBase Model=Claude-3.5-Sonnet2026.04 | 4.13 | 2.22 | |
| FAISSBase Model=GPT-4o2026.04 | 3.78 | 2.7 | |
| GPT+CoTBase Model=Claude-3.5-Sonnet2026.04 | 3.48 | 1.87 | |
| FAISS+CoTBase Model=GPT-4o2026.04 | 3.26 | 2.35 | |
| GPT+CoTBase Model=GPT-4o2026.04 | 3.09 | 2.26 | |
| Direct GPTBase Model=Claude-3.5-Sonnet2026.04 | 2.96 | 1.91 | |
| Direct GPTBase Model=GPT-4o2026.04 | 2.78 | 2.09 | |
| FAISSBase Model=Claude-3.5-Sonnet2026.04 | 1.87 | 2.04 |