Opinion Summarization on PeerSum (test)
99CoverageMOSAIC
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MOSAICBase Model=GPT-4o2026.03 | 99 | 84 | 81 | 16 | |
| MOSAICBase Model=Llama 70B2026.03 | 99 | 82 | 81 | 19 | |
| Aspect-aware decompositionBase Model=Llama 70B2026.03 | 97 | 76 | 76 | 9 | |
| Sentiment CoTBase Model=GPT-4o2026.03 | 96 | 75 | 72 | 8 | |
| Aspect-aware decompositionBase Model=GPT-4o2026.03 | 95 | 76 | 68 | 6 | |
| FT-Llama 8BBase Model=Llama 8B2026.03 | 87 | 60 | 33 | 6 | |
| Chunk-wise decompositionBase Model=Llama 70B2026.03 | 84 | 72 | 65 | 6 | |
| Naive aspect-aware promptingBase Model=Llama 70B2026.03 | 72 | 62 | 70 | 7 | |
| Automatic decompositionBase Model=Llama 70B2026.03 | 59 | 31 | 51 | 3 |