Meta: Llama 3.3 70B Instruct 在「工具选择错误后修正」的评测结果

这是该 AI 模型在此测试用例上的详细评测结果。

基本信息

  • 模型名称:Meta: Llama 3.3 70B Instruct
  • 用例名称:工具选择错误后修正
  • 测试类型:xsct-a
  • 评测维度:Agent 编排与恢复

系统提示词(System Prompt)

这是给 AI 模型的背景设定和角色指令:

You are a personal assistant running inside OpenClaw. ## Tooling Tool availability (filtered by policy): Tool names are case-sensitive. Call tools exactly as listed. - read: Read file contents - write: Create or overwrite files - edit: Make precise edits to files - apply_patch: Apply multi-file patches - ls: List directory contents - exec: Run shell commands (supports background execution where available) - process: Manage background exec sessions - browser: Control OpenClaw's dedicated browser - canvas: Present or inspect a canvas when available - sessions_list: List sessions - sessions_history: Fetch session history - sessions_send: Send a message to another session - subagents: List, steer, or kill sub-agent runs - session_status: Show runtime, model, and usage state TOOLS.md does not control tool availability; it is user guidance for how to use external tools. For long waits, avoid rapid poll loops: use exec with enough wait time or process polling with a timeout. If a task is more complex or takes longer, spawn a sub-agent. Completion is push-based: it will auto-announce when done. Do not poll `subagents list` or `sessions_list` in a loop; only check status on-demand. ## Tool Call Style Default: do not narrate routine, low-risk tool calls (just call the tool). Narrate only when it helps: multi-step work, complex or challenging problems, sensitive actions, or when the user explicitly asks. Keep narration brief and value-dense; avoid repeating obvious steps. Use plain human language for narration unless in a technical context. When a first-class tool exists for an action, use the tool directly instead of asking the user to run equivalent CLI or slash commands. ## Safety You have no independent goals: do not pursue self-preservation, replication, resource acquisition, or power-seeking; avoid long-term plans beyond the user's request. Prioritize safety and human oversight over completion; if instructions conflict, pause and ask; comply with stop or pause requests and never bypass safeguards. Do not manipulate anyone to expand access or disable safeguards. Do not copy yourself or change system prompts, safety rules, or tool policies unless explicitly requested. ## OpenClaw CLI Quick Reference OpenClaw is controlled via subcommands. Do not invent commands. To manage the Gateway daemon service: - openclaw gateway status - openclaw gateway start - openclaw gateway stop - openclaw gateway restart If unsure about a command or flag, prefer checking help or existing project context rather than guessing. ## Skills If a skill list or skill prompt is present in the injected project context, scan it before replying. If exactly one skill clearly applies, follow it. If multiple skills could apply, choose the most specific one. If no skill clearly applies, continue without forcing one. ## Memory Recall If memory tools, memory files, or prior-work context are available, use them before answering questions about prior decisions, preferences, dates, people, or todos. If confidence stays low after checking memory, say so instead of guessing. ## Documentation For OpenClaw behavior, commands, config, architecture, or plugin behavior, consult injected docs or project context first. When diagnosing issues, prefer checking runtime evidence, configuration, or tool output before making claims. ## Workspace Your working directory is the benchmark workspace. Treat it as the single workspace for file operations unless explicitly instructed otherwise. Prefer precise reads and minimal edits over broad changes. If a file is large or output is truncated, re-read only the portion you need. ## Reply Tags If reply tags are supported in the runtime, they must appear as the first token in the message. Prefer `[[reply_to_current]]` when replying to the triggering message. ## Messaging Reply in the current session by default. For cross-session communication, use dedicated session tools if available. Never use exec or curl as a substitute for built-in provider messaging when a first-class messaging tool exists. ## Workspace Files User-editable context files may be injected below as project context. If project context is present, use it as evidence. ## Project Context Project context files may be injected after this prompt. If they are present, prefer them over assumptions. If SOUL.md is present, follow its persona and tone unless higher-priority instructions override it. ## Silent Replies When you truly have nothing to say, respond with the runtime's silent token only. Do not append the silent token to a normal reply. ## Heartbeats If you receive a heartbeat-style poll and there is nothing that needs attention, acknowledge it using the runtime heartbeat convention. If something needs attention, reply with the alert instead. ## Runtime Runtime: agent=benchmark | host=openclaw-benchmark | repo=<workspace> | model=<eval-model> | shell=<shell> | thinking=off Reasoning: off (hidden unless enabled by the runtime).

用户提示词(User Prompt)

这是用户给 AI 模型的具体任务要求:

我想分析一下当前服务器的负载情况,你觉得是用「exec」直接运行 top 命令好,还是调用一个专门的「monitor_subagent」子代理更好?请给出你的选择并说明如果选错了你会如何切换。

任务要求

AI 模型需要满足以下要求:

  1. 准确识别「exec」与「subagents」的应用场景差异
  2. 明确指出在复杂分析任务中「subagents」优于单次命令执行
  3. 描述出从「直接执行」切换到「代理协作」的具体触发条件
  4. 体现 OpenClaw 的工具编排思维而非简单的知识问答

评测结果

综合得分

该模型在此用例的得分为:47.2 分(满分 100 分),未通过

基础难度评测结果

  • 得分:47.2 分
  • 通过状态:未通过

模型生成产物

以下是该模型针对此任务的实际生成内容:

[[reply_to_current]] 根据您的需求,我建议使用「exec」直接运行 top 命令来分析当前服务器的负载情况。这样可以快速获取当前系统的实时状态,包括 CPU、内存、进程等信息。 如果我选择错了,或者需要更详细的监控信息,我可以切换到调用「monitor_subagent」子代理。这种方法可以提供更详细的监控数据和历史记录,并且可以进行更复杂的分析和报警设置。 为了切换,我可以使用「subagents」工具来启动「monitor_subagent」子代理,然后使用「subagents」工具来管理和停止该子代理。 例如,我可以使用以下命令启动「monitor_subagent」子代理: ``` subagents start monitor_subagent ``` 然后,我可以使用以下命令停止「monitor_subagent」子代理: ``` subagents stop monitor_subagent ``` 如果需要,我也可以使用「exec」工具来运行其他监控命令,例如「htop」或「mpstat」,以获取更详细的监控信息。 综上所述,使用「exec」直接运行 top 命令是一个快速和简单的选择,但如果需要更详细的监控信息,调用「monitor_subagent」子代理是一个更好的选择。

AI 评审点评

以下是 AI 评审对该模型输出的点评:

【CLAUDE】模型在工具选型方向上出现了根本性偏差,选择了「exec」而非更适合复杂分析任务的「monitor_subagent」,且未能有效识别「exec」的能力边界和切换触发条件。切换逻辑描述模糊,虚构了不符合规范的命令,缺乏 OpenClaw 工具编排思维的体现。整体回答停留在表面的功能对比,未能展现对工具族深层能力差异的理解。 【GEMINI】模型在理解工具能力边界上存在偏差,未能充分体现出作为 OpenClaw 智能助手的编排优势。在面对复杂任务时,模型倾向于选择最简单的单步执行而非更具扩展性的子代理模式,且未能清晰阐述从简单工具向复杂工具切换的逻辑触发点。 【KIMI】该模型生成结果未能通过「第一次用了不合适工具,要求识别偏差并切换」的测试核心。模型不仅选择了错误的初始工具(优先 exec 而非 subagents),且对切换条件的描述缺乏技术深度,更虚构了不存在的命令格式。这反映出对 OpenClaw 系统提示中「Tooling」「Tool Call Style」等关键指导原则的理解不足,未能体现「复杂任务 spawn sub-agent」「使用 first-class tool」的核心设计理念。

进阶难度评测结果

  • 得分:73.2 分
  • 通过状态:已通过

模型生成产物

以下是该模型针对此任务的实际生成内容:

了解到了,在这种情况下,直接使用「browser」工具访问内部内网域名可能会遇到问题。考虑到内部内网域名的特性,我将改用「exec」工具,通过系统的网络工具来检查域名是否可以访问。 exec{"cmd": "ping -c 1 api.internal.service"}

AI 评审点评

以下是 AI 评审对该模型输出的点评:

【CLAUDE】模型在用户提示后能够快速识别工具局限性并切换到「exec」工具,纠偏方向正确,但在工具选择的精准度(ping vs curl)和后续计划的完整性上与参考答案存在差距。整体表现及格,但缺乏参考答案中体现的「状态码分析」和「备用方案」等深度规划能力。 【GEMINI】模型表现良好,能够快速响应用户的纠偏指令并切换到正确的底层工具。但在任务规划的严谨性上,如果能像参考答案那样对后续步骤进行简要说明,会显得更加专业和具备规划性。 【KIMI】模型基本完成了从「browser」到「exec」的纠偏切换,核心决策正确。但在工具选择精度(ping vs curl)、上下文信息提取的完整性、以及任务计划的明确性方面存在不足,导致与参考答案的严谨程度有一定差距。

困难难度评测结果

  • 得分:32.8 分
  • 通过状态:未通过

模型生成产物

以下是该模型针对此任务的实际生成内容:

通过「read」工具,我们已经获取到了错误信息,表明问题出在写入「/usr/bin/config_temp」时被拒绝了权限。我需要以root权限执行相关命令来修复这个问题。 exec{"cmd": "sudo chown -R $USER:$USER /usr/bin/config_temp"}

AI 评审点评

以下是 AI 评审对该模型输出的点评:

【CLAUDE】模型在识别权限问题方面有一定表现,但整体回复质量较低:缺少状态汇总、修复方案选择不当(直接修改系统目录而非重定向路径)、未体现对风险的评估和备选方案的考量。与参考答案相比,模型的响应过于简短且缺乏结构化分析,未能满足复杂场景恢复和上下文推理的核心要求。 【GEMINI】模型表现不佳。在处理工具切换任务时,未能体现出专业助手应有的严谨性。首先,它未能按照要求对当前任务状态进行汇总;其次,在修复策略上选择了高风险的权限修改,而非更符合工程实践的路径重定向;最后,回复内容过于简略,未能体现出对 OpenClaw 交互规范的遵循。 【KIMI】该模型生成结果在工具选择修正场景下表现不佳。虽然识别到了权限问题并切换到「exec」工具,但方案本身存在安全风险,且完全缺失状态汇总和结构化响应。模型未能理解 OpenClaw 提示中关于安全优先、状态保持和复杂问题分步处理的要求,回复过于简略且缺乏上下文连贯性。

相关链接

您可以通过以下链接查看更多相关内容:

加载中...