Calibration Performance on xBench DeepSearch
0.34NECEDeliberative Search
Evaluation results
| Method | Links | |
|---|---|---|
| Deliberative Searchbackbone=GPT-4o2026.04 | 0.34 | |
| Claude-4-Sonnet2026.04 | 0.36 | |
| GPT-4o2026.04 | 0.39 | |
| DeepSeek-R1-Distill-Llama-70B2026.04 | 0.42 | |
| GPT-4.12026.04 | 0.47 | |
| Intern-VL3-78B2026.04 | 0.5 | |
| Qwen-VL2.5-72B2026.04 | 0.53 |