Factual Consistency on RecRM-Bench
70.71Accuracy (%)Ours
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Ours2026.05 | 70.71 | 82.84 | |
| LongCat-Flash-ChatThinking=false2026.05 | 67.34 | 80.48 | |
| LongCat-Flash-ThinkingThinking=true2026.05 | 64.98 | 78.78 | |
| GPT-4.12026.05 | 62.96 | 77.27 | |
| Qwen3-MaxThinking=true2026.05 | 56.57 | 72.26 | |
| Qwen3-MaxThinking=false2026.05 | 53.2 | 69.45 | |
| Deepseek-V3.2Thinking=false2026.05 | 43.43 | 60.56 | |
| Deepseek-V3.2Thinking=true2026.05 | 41.41 | 58.57 |