Insight Generation on Internal non-scientific document collections Twitter & Mental Health
4.5Set-level Score (Gemini-2.5-Flash)FAISS+CoT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FAISS+CoTBase Model=Claude-3.5-Sonnet2026.04 | 4.5 | 2.31 | |
| INSIGHTGENBase Model=Claude-3.5-Sonnet2026.04 | 4.38 | 3.25 | |
| INSIGHTGENBase Model=GPT-4o2026.04 | 4.31 | 3.63 | |
| FAISSBase Model=GPT-4o2026.04 | 3.88 | 3.13 | |
| GPT+CoTBase Model=GPT-4o2026.04 | 3.25 | 2.25 | |
| Direct GPTBase Model=GPT-4o2026.04 | 3.13 | 2.31 | |
| GPT+CoTBase Model=Claude-3.5-Sonnet2026.04 | 2.88 | 1.88 | |
| FAISSBase Model=Claude-3.5-Sonnet2026.04 | 2.69 | 2.31 | |
| Direct GPTBase Model=Claude-3.5-Sonnet2026.04 | 2.63 | 2.19 | |
| FAISS+CoTBase Model=GPT-4o2026.04 | 2.38 | 2.13 |