Opinion Summarization on Rotten Tomatoes (full dataset)
7.21CoverageBelief Merging
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Belief MergingModel=LLaMA-8B2026.01 | 7.21 | 7.23 | 6.79 | 6.96 | 8.31 | 0.84 | -4.15 | |
| Fusion-of-NModel=Qwen3-14B2026.01 | 7.17 | 7.3 | 6.36 | 6.67 | 8.73 | 0.85 | -3.52 | |
| Belief Merging + FusionModel=LLaMA-8B2026.01 | 7.15 | 7.18 | 6.77 | 6.93 | 8.31 | 0.84 | -4.14 | |
| Belief Merging + FusionModel=Qwen3-4B2026.01 | 7.15 | 7.19 | 6.75 | 6.91 | 8.25 | 0.84 | -4.15 | |
| Belief MergingModel=Qwen3-14B2026.01 | 7.13 | 7.18 | 6.71 | 6.93 | 8.2 | 0.84 | -4.16 | |
| Fusion-of-NModel=Qwen3-4B2026.01 | 7.09 | 7.12 | 6.68 | 6.87 | 8.19 | 0.85 | -3.4 | |
| Fusion-of-NModel=GPT-52026.01 | 7.03 | 7.06 | 6.66 | 6.83 | 8.11 | 0.84 | -3.32 | |
| Belief MergingModel=Qwen3-4B2026.01 | 7.03 | 7.06 | 6.63 | 6.8 | 8.17 | 0.84 | -4.88 | |
| Belief Merging + FusionModel=LLaMA-3B2026.01 | 6.96 | 6.98 | 6.55 | 6.73 | 8.1 | 0.84 | -4.15 | |
| Belief Merging + FusionModel=Qwen3-14B2026.01 | 6.95 | 6.98 | 6.53 | 6.72 | 8.02 | 0.85 | -4.15 | |
| Belief Merging + FusionModel=GPT-52026.01 | 6.93 | 7.53 | 5.35 | 6.26 | 8.66 | 0.84 | -4.12 | |
| Belief MergingModel=LLaMA-3B2026.01 | 6.89 | 6.88 | 6.42 | 6.62 | 7.94 | 0.84 | -4.18 | |
| Belief MergingModel=GPT-52026.01 | 6.88 | 7.6 | 5.13 | 6.23 | 8.52 | 0.83 | -4.44 | |
| QA PromptingModel=Qwen3-4B2026.01 | 5.3 | 5.81 | 2.78 | 4.29 | 3.85 | 0.87 | -3.41 | |
| QA PromptingModel=Qwen3-14B2026.01 | 5.04 | 5.49 | 2.66 | 3.99 | 4 | 0.86 | -3.51 | |
| QA PromptingModel=LLaMA-8B2026.01 | 4.99 | 5.09 | 2.13 | 3.05 | 2.46 | 0.82 | -3.44 | |
| QA PromptingModel=LLaMA-3B2026.01 | 4.3 | 4.45 | 1.95 | 3.08 | 2.72 | 0.86 | -3.52 | |
| Fusion-of-NModel=LLaMA-8B2026.01 | 3.84 | 3.81 | 3.25 | 3.52 | 4.96 | 0.8 | -4.55 | |
| Fusion-of-NModel=LLaMA-3B2026.01 | 1.45 | 1.45 | 1.18 | 1.31 | 2.16 | 0.83 | -4.61 |