Character Relationship Network Updating on BookWorld FEN Synthesized scenario
70.7Win RateDeepSeek-V3.2-Think
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DeepSeek-V3.2-ThinkBaseline=vs. BookWorld (FEN)2026.04 | 70.7 | 4.63 | |
| Gemini-2.5-ProBaseline=vs. BookWorld (FEN)2026.04 | 68 | 4.37 | |
| DeepSeek-V3.2Baseline=vs. BookWorld (FEN)2026.04 | 54.7 | 3.63 | |
| Llama3.3-70BBaseline=vs. BookWorld (FEN)2026.04 | 50.7 | 2.36 | |
| Qwen3-32BBaseline=vs. BookWorld (FEN)2026.04 | 34.7 | 2.61 |