Agentic Question Answering on BrowseComp-Plus
74.7RecallGPT-5.2
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT-5.2Search calls=73.832026.02 | 74.7 | 45.1 | |
| gpt-oss-20bSearch calls=30.14, vLLM=v0.152026.02 | 57 | 42.1 | |
| gpt-oss-20bSearch calls=29.63, vLLM=v0.132026.02 | 55.7 | 42.2 | |
| gpt-oss-20bSearch calls=23.87, vLLM=v0.9.0.12026.02 | 49.3 | 34.6 |