ETCLOVG七层架构将SWE-bench基准从6.7%提升至68.3%，无需更改模型
英文摘要
A paper identifies harness engineering as the primary driver of AI agent reliability and proposes the ETCLOVG seven-layer architecture, unifying sandboxes, tool protocols, context state, lifecycle graphs, observability, verifiers, and governance. Optimizing only the execution harness raised SWE-bench coding benchmark performance from 6.7% to 68.3%, without any modification to the underlying language model. The framework represents a shift from relying on model weights for reliability to deterministic harness design.
中文摘要
一篇论文将执行框架工程确立为AI智能体可靠性的首要决定因素，并提出ETCLOVG七层架构，统一了执行沙箱、工具协议、上下文状态、生命周期图、可观测性、验证器和治理。仅优化执行框架就将SWE-bench编码基准性能从6.7%提升至68.3%，无需改动底层模型。该框架标志着系统可靠性从依赖模型权重转向确定性框架设计。
关键要点
Harness engineering, not model scale, is identified as the primary determinant of AI agent reliability.
执行框架工程，而非模型规模，被确定为AI智能体可靠性的首要决定因素。
The ETCLOVG architecture consists of seven layers: execution sandboxes, tool protocols, context state, lifecycle graphs, observability, verifiers, and governance.
ETCLOVG架构由七层组成：执行沙箱、工具协议、上下文状态、生命周期图、可观测性、验证器和治理。
Optimizing the harness alone increased SWE-bench scores from 6.7% to 68.3%, a 10x improvement without changing the LLM.
仅优化框架就将SWE-bench得分从6.7%提升至68.3%，在不更改LLM的情况下实现了10倍提升。
The approach shifts system reliability from model-dependent probabilities to deterministic architecture design.
该方法将系统可靠性从依赖模型概率转向确定性架构设计。