Character Relationship Network Updating on OpenTheatre Existing scenario
100Win RateQwen3-32B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-32BBaseline=vs. OpenTheatre2026.04 | 100 | 2.89 | |
| DeepSeek-V3.2-ThinkBaseline=vs. OpenTheatre2026.04 | 86.7 | 4.11 | |
| Gemini-2.5-ProBaseline=vs. OpenTheatre2026.04 | 80 | 3.79 | |
| Llama3.3-70BBaseline=vs. OpenTheatre2026.04 | 71.7 | 1.98 | |
| DeepSeek-V3.2Baseline=vs. OpenTheatre2026.04 | 50 | 2.32 |