Character Relationship Network Updating on OpenTheatre Synthesized scenario
86.7Win RateLlama3.3-70B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Llama3.3-70BBaseline=vs. OpenTheatre2026.04 | 86.7 | 3.47 | |
| DeepSeek-V3.2-ThinkBaseline=vs. OpenTheatre2026.04 | 68.3 | 4.17 | |
| Gemini-2.5-ProBaseline=vs. OpenTheatre2026.04 | 55 | 4.35 | |
| DeepSeek-V3.2Baseline=vs. OpenTheatre2026.04 | 50 | 3.52 | |
| Qwen3-32BBaseline=vs. OpenTheatre2026.04 | 48.3 | 2.87 |