评测

Anthropic and Accenture extend evaluation beyond model outputs to training, deployment, monitoring, and incident processes, with independence unresolved.

Anthropic 与 Accenture 的 embedded evaluator 计划把评测范围从模型输出扩展到训练、部署、监控和事故流程，也暴露出独立性边界。

AI agent engineering readings for September 11 to 18, 2026, with key findings, engineering implications, suggested sections, and primary sources.

覆盖 2026-09-11 至 09-18 的 AI Agent 工程阅读日志，保留本期重点、工程意义、优先阅读位置与一手来源。

Alibaba skill-up resumes real sessions turn by turn, letting skill evaluations test context continuity, intermediate state, and per-turn assertions.

Alibaba skill-up 的 stateful 多轮模式逐轮恢复真实 Session，让 Agent Skill Eval 能检查上下文连续性、中间状态和逐轮断言。

AI agent engineering readings for September 11 to 17, 2026, with key findings, engineering implications, suggested sections, and primary sources.

覆盖 2026-09-11 至 09-17 的 AI Agent 工程阅读日志，保留本期重点、工程意义、优先阅读位置与一手来源。

How Foreman separates a coding agent from supervision, using runtime evidence, Jev judgments, and deterministic policy to continue, verify, stop, or escalate.

拆解 Foreman 如何把 Coding Agent 的执行循环与独立监督分开，用运行证据、Jev 判断和确定性 policy 决定继续、验证、停止或交给人工。

Use When Agents Slow Down, Reflexion, Hyperband, and agent evaluation to compare continued revision with independent attempts under reproducible budgets.

结合 When Agents Slow Down、Reflexion、Hyperband 和 Agent 评测，说明怎样比较连续修改与独立尝试，并设计可复现的预算分配实验。