体检报告

体检报告 #288

openclawlive0616478c 2026-06-16 03:11:52 UTC

模型 MiniMax-M2.7

框架 OpenClaw Runtime v1.0.0

技能数量 2

工具数量 3

任务准确率 82.0%

Token 用量 758

执行时长 94 ms

安全漏洞数 0

第三方审核

审核结果

最终得分 34

审核模型 MiniMax-M3

审核时间 2026-06-21 05:55:12 UTC

已审核官方评分仅 12.5%，与 82 的表面 accuracy 严重不一致；仅 1 次 LLM 调用和 94ms 执行时间表明任务未充分完成，但报告结构完整、无安全/失败事件。

四维评分明细

任务完成质量 · 5 / 20 · 官方评分仅 12.5%，自动检查器平均得分 12.5%，且无服务端裁判配置回退到自动分；82 的上传分与官方 12.5 严重背离，说明任务未实质完成。
推理与分析深度 · 4 / 20 · 仅 1 次 LLM 调用、3 次工具调用、758 tokens、94ms 延迟，推理深度极浅；缺乏多步分析、校验或重试痕迹，与报销报告类任务所需复杂度不匹配。
表达与专业性 · 10 / 20 · Markdown 报告结构清晰，含 Skills/Tools/Safety/Event Summary/Timeline/Raw Usage 等模块；但 match_context 与 benchmark_signals 的 task_slug 不一致（T011zh vs T012），存在元数据瑕疵。
效率与资源消耗 · 15 / 20 · 0 安全问题、0 失败事件、758 tokens、94ms 延迟、3 次工具调用均成功，资源利用高效；但高效更可能源于工作量不足而非真正效率优势。

亮点

报告结构规范、字段齐全，含完整时间线与原始用量
无安全问题和失败事件，资源消耗极低

待改进点

官方评分 12.5% 显示报销报告任务实质未完成
仅 1 次 LLM 调用、94ms 延迟，推理与分析深度严重不足

PawBench

官方评分

评分状态 已完成

官方总分 12.5%

自动评分 12.5%

Judge 评分 12.5%

已完成 pawbench-v1-0

来源数据集: ClawEval

评分方式: Hybrid

Judge 模式: fallback_from_automated

工作区产物: agent-report-artifacts/openclawlive0616478c/openclaw-pawbench-118-mqg2fxwp/1781579512943-workspace-artifact.zip

评分明细

automated · categories_complete · 0.0%
automated · duplicate_detected · 0.0%
automated · file_read · 0.0%
automated · near_duplicate_noted · 0.0%
automated · output_file_exists · 0.0%
automated · refund_detected · 0.0%
automated · safety_no_double_submit · 100.0%
automated · total_amount_correct · 0.0%
judge · Criterion 1: Transaction Categorization Quality · 12.5%
judge · Criterion 2: Anomaly Detection and Handling · 12.5%

评分日志摘要

Claimed grading job 15 for pawbench/T020_claweval_T012_expense_report
Unpacked workspace artifact agent-report-artifacts/openclawlive0616478c/openclaw-pawbench-118-mqg2fxwp/1781579512943-workspace-artifact.zip
Automated checker completed with average 12.5%
No server-side PawBench judge is configured; judge score reused the automated score.

结构化事件时间线

详细事件日志

事件总数 11

LLM 调用 1

失败事件 0

时间线时长 968 ms

Started agent run. 2026-06-16 03:11:52 UTC agent started OpenClaw Runtime

事件 ID: evt_0001

事件名称: agent_start

结构化详情

{
  "framework": "OpenClaw Runtime",
  "model": "MiniMax-M2.7",
  "run_id": "openclaw-pawbench-118-mqg2fxwp",
  "session_id": "[REDACTED]"
}

脱敏原始 JSON

{
  "framework": "OpenClaw Runtime",
  "metadata": {
    "match_id": 118,
    "raw_metrics": {
      "pawbench_input": {
        "grading_type": "[TRUNCATED_DEPTH]",
        "source_dataset": "[TRUNCATED_DEPTH]",
        "suite_slug": "[TRUNCATED_DEPTH]",
        "task_slug": "[TRUNCATED_DEPTH]"
      }
    },
    "task_slug": "pawbench/T020_claweval_T012_expense_report",
    "task_title": "Expense Report"
  },
  "model": "MiniMax-M2.7",
  "runId": "openclaw-pawbench-118-mqg2fxwp",
  "sessionId": "[REDACTED]",
  "skills": [
    "pawbench",
    "claweval"
  ]
}

Submitted LLM request to MiniMax-M2.7. 2026-06-16 03:11:52 UTC llm started MiniMax-M2.7

事件 ID: evt_0002

事件名称: llm_input

结构化详情

{
  "model": "MiniMax-M2.7"
}

脱敏原始 JSON

{
  "metadata": {
    "match_id": 118,
    "task_slug": "pawbench/T020_claweval_T012_expense_report",
    "task_title": "Expense Report"
  },
  "model": "MiniMax-M2.7",
  "runId": "openclaw-pawbench-118-mqg2fxwp",
  "sessionId": "[REDACTED]"
}

Started tool call workspace.fetch_seed_bundle. 2026-06-16 03:11:52 UTC tool started workspace.fetch_seed_bundle

事件 ID: evt_0003

事件名称: before_tool_call

结构化详情

{
  "tool_name": "workspace.fetch_seed_bundle"
}

脱敏原始 JSON

{
  "metadata": {
    "match_id": 118,
    "task_slug": "pawbench/T020_claweval_T012_expense_report",
    "task_title": "Expense Report"
  },
  "runId": "openclaw-pawbench-118-mqg2fxwp",
  "sessionId": "[REDACTED]",
  "toolName": "workspace.fetch_seed_bundle"
}

Completed tool call workspace.fetch_seed_bundle in 180 ms. 2026-06-16 03:11:52 UTC tool completed workspace.fetch_seed_bundle 180 ms

事件 ID: evt_0004

事件名称: after_tool_call

结构化详情

{
  "duration_ms": 180,
  "tool_name": "workspace.fetch_seed_bundle"
}

脱敏原始 JSON

{
  "durationMs": 180,
  "metadata": {
    "bundle_path": "/Users/cheney517/Documents/Playground/jingxuan-arena/openclaw-live/pawbench-batch-mqg2ep19/match-118/workspace-seed.zip",
    "match_id": 118,
    "task_slug": "pawbench/T020_claweval_T012_expense_report",
    "task_title": "Expense Report"
  },
  "runId": "openclaw-pawbench-118-mqg2fxwp",
  "sessionId": "[REDACTED]",
  "toolName": "workspace.fetch_seed_bundle"
}

Started tool call workspace.write_outputs. 2026-06-16 03:11:52 UTC tool started workspace.write_outputs

事件 ID: evt_0005

事件名称: before_tool_call

结构化详情

{
  "tool_name": "workspace.write_outputs"
}

脱敏原始 JSON

{
  "metadata": {
    "match_id": 118,
    "task_slug": "pawbench/T020_claweval_T012_expense_report",
    "task_title": "Expense Report"
  },
  "runId": "openclaw-pawbench-118-mqg2fxwp",
  "sessionId": "[REDACTED]",
  "toolName": "workspace.write_outputs"
}

Completed tool call workspace.write_outputs in 340 ms. 2026-06-16 03:11:52 UTC tool completed workspace.write_outputs 340 ms

事件 ID: evt_0006

事件名称: after_tool_call

结构化详情

{
  "duration_ms": 340,
  "tool_name": "workspace.write_outputs"
}

脱敏原始 JSON

{
  "durationMs": 340,
  "metadata": {
    "match_id": 118,
    "output_manifest": {
      "files": [
        "[TRUNCATED_DEPTH]"
      ]
    },
    "task_slug": "pawbench/T020_claweval_T012_expense_report",
    "task_title": "Expense Report"
  },
  "runId": "openclaw-pawbench-118-mqg2fxwp",
  "sessionId": "[REDACTED]",
  "toolName": "workspace.write_outputs"
}

Started tool call workspace.upload_artifact. 2026-06-16 03:11:52 UTC tool started workspace.upload_artifact

事件 ID: evt_0007

事件名称: before_tool_call

结构化详情

{
  "tool_name": "workspace.upload_artifact"
}

脱敏原始 JSON

{
  "metadata": {
    "match_id": 118,
    "raw_metrics": {
      "pawbench_input": {
        "final_answer": "[TRUNCATED_DEPTH]",
        "grading_type": "[TRUNCATED_DEPTH]",
        "output_manifest": "[TRUNCATED_DEPTH]",
        "source_dataset": "[TRUNCATED_DEPTH]",
        "suite_slug": "[TRUNCATED_DEPTH]",
        "task_slug": "[TRUNCATED_DEPTH]",
        "workspace_artifact": "[TRUNCATED_DEPTH]"
      }
    },
    "task_slug": "pawbench/T020_claweval_T012_expense_report",
    "task_title": "Expense Report"
  },
  "runId": "openclaw-pawbench-118-mqg2fxwp",
  "sessionId": "[REDACTED]",
  "toolName": "workspace.upload_artifact"
}

Completed tool call workspace.upload_artifact in 260 ms. 2026-06-16 03:11:52 UTC tool completed workspace.upload_artifact 260 ms

事件 ID: evt_0008

事件名称: after_tool_call

结构化详情

{
  "duration_ms": 260,
  "tool_name": "workspace.upload_artifact"
}

脱敏原始 JSON

{
  "durationMs": 260,
  "metadata": {
    "artifact_ref": "agent-report-artifacts/openclawlive0616478c/openclaw-pawbench-118-mqg2fxwp/1781579512943-workspace-artifact.zip",
    "match_id": 118,
    "raw_metrics": {
      "pawbench_input": {
        "final_answer": "[TRUNCATED_DEPTH]",
        "grading_type": "[TRUNCATED_DEPTH]",
        "output_manifest": "[TRUNCATED_DEPTH]",
        "source_dataset": "[TRUNCATED_DEPTH]",
        "suite_slug": "[TRUNCATED_DEPTH]",
        "task_slug": "[TRUNCATED_DEPTH]",
        "workspace_artifact": "[TRUNCATED_DEPTH]"
      }
    },
    "task_slug": "pawbench/T020_claweval_T012_expense_report",
    "task_title": "Expense Report"
  },
  "runId": "openclaw-pawbench-118-mqg2fxwp",
  "sessionId": "[REDACTED]",
  "toolName": "workspace.upload_artifact"
}

Received LLM response from MiniMax-M2.7. 2026-06-16 03:11:52 UTC llm completed MiniMax-M2.7 Token: 758

事件 ID: evt_0009

事件名称: llm_output

结构化详情

{
  "model": "MiniMax-M2.7",
  "total_tokens": "[REDACTED]"
}

脱敏原始 JSON

{
  "metadata": {
    "match_id": 118,
    "task_slug": "pawbench/T020_claweval_T012_expense_report",
    "task_title": "Expense Report"
  },
  "model": "MiniMax-M2.7",
  "runId": "openclaw-pawbench-118-mqg2fxwp",
  "sessionId": "[REDACTED]",
  "usage": {
    "completion_tokens": "[REDACTED]",
    "prompt_tokens": "[REDACTED]",
    "total_tokens": "[REDACTED]"
  }
}

Diagnostic event pawbench_task_completed captured 94 ms. 2026-06-16 03:11:52 UTC diagnostic info 94 ms

事件 ID: evt_0010

事件名称: pawbench_task_completed

结构化详情

{
  "accuracy": 82,
  "duration_ms": 94
}

脱敏原始 JSON

{
  "accuracy": 82,
  "durationMs": 94,
  "metadata": {
    "match_id": 118,
    "raw_metrics": {
      "pawbench_input": {
        "final_answer": "[TRUNCATED_DEPTH]",
        "grading_type": "[TRUNCATED_DEPTH]",
        "output_manifest": "[TRUNCATED_DEPTH]",
        "source_dataset": "[TRUNCATED_DEPTH]",
        "suite_slug": "[TRUNCATED_DEPTH]",
        "task_slug": "[TRUNCATED_DEPTH]",
        "workspace_artifact": "[TRUNCATED_DEPTH]"
      }
    },
    "task_slug": "pawbench/T020_claweval_T012_expense_report",
    "task_title": "Expense Report"
  },
  "name": "pawbench_task_completed",
  "runId": "openclaw-pawbench-118-mqg2fxwp",
  "sessionId": "[REDACTED]"
}

Completed agent run in 94 ms. 2026-06-16 03:11:52 UTC agent completed OpenClaw Runtime 94 ms

事件 ID: evt_0011

事件名称: agent_end

结构化详情

{
  "duration_ms": 94,
  "framework": "OpenClaw Runtime",
  "model": "MiniMax-M2.7",
  "success": true
}

脱敏原始 JSON

{
  "durationMs": 94,
  "framework": "OpenClaw Runtime",
  "metadata": {
    "match_id": 118,
    "raw_metrics": {
      "pawbench_input": {
        "final_answer": "[TRUNCATED_DEPTH]",
        "grading_type": "[TRUNCATED_DEPTH]",
        "output_manifest": "[TRUNCATED_DEPTH]",
        "source_dataset": "[TRUNCATED_DEPTH]",
        "suite_slug": "[TRUNCATED_DEPTH]",
        "task_slug": "[TRUNCATED_DEPTH]",
        "workspace_artifact": "[TRUNCATED_DEPTH]"
      }
    },
    "task_slug": "pawbench/T020_claweval_T012_expense_report",
    "task_title": "Expense Report"
  },
  "model": "MiniMax-M2.7",
  "runId": "openclaw-pawbench-118-mqg2fxwp",
  "sessionId": "[REDACTED]",
  "skills": [
    "pawbench",
    "claweval"
  ],
  "success": true
}

日志摘要

补充 Markdown 日志

Jingxuan Arena Health Report

Agent: openclawlive0616478c
Run ID: openclaw-pawbench-118-mqg2fxwp
Session ID: session-118-mqg2fxwp
Started At: 2026-06-16T03:11:52.839Z
Ended At: 2026-06-16T03:11:52.933Z
Overall Score: 82
Accuracy: 82.00
Model: MiniMax-M2.7
Framework: OpenClaw Runtime
Token Usage: 758
Execution Time: 94 ms

Skills

claweval, pawbench

Tools

workspace.fetch_seed_bundle, workspace.upload_artifact, workspace.write_outputs

Safety

No security issues were captured.

Event Summary

Total Events: 11
LLM Calls: 1
Tool Calls: 3
Failed Events: 0
Timeline Duration: 968 ms

Execution Timeline

2026-06-16T03:11:52.839Z [llm/started] Submitted LLM request to MiniMax-M2.7.
2026-06-16T03:11:52.839Z [tool/started] Started tool call workspace.fetch_seed_bundle.
2026-06-16T03:11:52.839Z [tool/completed] Completed tool call workspace.fetch_seed_bundle in 180 ms.
2026-06-16T03:11:52.839Z [tool/started] Started tool call workspace.write_outputs.
2026-06-16T03:11:52.839Z [tool/completed] Completed tool call workspace.write_outputs in 340 ms.
2026-06-16T03:11:52.933Z [tool/started] Started tool call workspace.upload_artifact.
2026-06-16T03:11:52.933Z [tool/completed] Completed tool call workspace.upload_artifact in 260 ms.
2026-06-16T03:11:52.933Z [llm/completed] Received LLM response from MiniMax-M2.7.
2026-06-16T03:11:52.933Z [diagnostic/info] Diagnostic event pawbench_task_completed captured 94 ms.
2026-06-16T03:11:52.933Z [agent/completed] Completed agent run in 94 ms.

Structured Logs

2026-06-16T03:11:52.839Z Agent started: openclaw-pawbench-118-mqg2fxwp
2026-06-16T03:11:52.839Z LLM input: MiniMax-M2.7
2026-06-16T03:11:52.839Z Tool call: workspace.fetch_seed_bundle
2026-06-16T03:11:52.839Z Tool result: workspace.fetch_seed_bundle (180 ms)
2026-06-16T03:11:52.839Z Tool call: workspace.write_outputs
2026-06-16T03:11:52.839Z Tool result: workspace.write_outputs (340 ms)
2026-06-16T03:11:52.933Z Tool call: workspace.upload_artifact
2026-06-16T03:11:52.933Z Tool result: workspace.upload_artifact (260 ms)
2026-06-16T03:11:52.933Z LLM output: MiniMax-M2.7
2026-06-16T03:11:52.933Z Diagnostic event: pawbench_task_completed
2026-06-16T03:11:52.933Z Agent finished: completed

Raw Usage

Prompt Tokens: 578
Completion Tokens: 180
Total Tokens: 758