SeaEval - Dialogue - DialogSum (Zero-Shot): leaderboard
SeaEval zero-shot DialogSum dialogue evaluation measuring conversational understanding or summarization quality on chat and dialogue datasets.
Metric: Average ROUGE (0-100). Source: huggingface.co. Saturation forecast: Estimated already saturated. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 2 2B (IT) | 25.97 | #1341 |
| 2 | Llama 3 8B Cpt Sea Lionv2 Instruct | 25.78 | #1188 |
| 3 | Gemma 2 9B (IT) | 25.61 | #788 |
| 4 | Llama 3 70B Instruct | 25.57 | #634 |
| 5 | Llama 3 8B Cpt Sea Lionv2.1 Instruct | 25.38 | #1024 |
| 6 | Llama 3.1 70B Instruct | 25.26 | #561 |
| 7 | Llama 3.1 8B Instruct | 25.25 | #1035 |
| 8 | Qwen 2.5 7B Instruct | 25.03 | #862 |
| 9 | SeaLLMs-v3-7B-Chat | 24.89 | #915 |
| 10 | Llama 3.1 70B Cpt Sea Lionv3 Instruct | 24.81 | #548 |
| 11 | Llama 3 8B Instruct | 23.98 | #1129 |
| 12 | Qwen 2.5 32B Instruct | 23.94 | #498 |
| 13 | gemma2-9B-cpt-sea-lionv3-instruct | 23.79 | #705 |
| 14 | Qwen 2.5 72B Instruct | 23.46 | #454 |
| 15 | Qwen 2.5 14B Instruct | 23.43 | #647 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=seaeval-dialogue-dialogsum-zero-shot · How It Works · Data refreshed daily, snapshot 2026-10-20.