liqiang
b119135836
Build latest book artifacts / build (push) Canceled after 0s
dependency resolution / resolve (3.11) (push) Canceled after 0s
dependency resolution / resolve (3.13) (push) Canceled after 0s
deploy-pages / build (push) Canceled after 0s
deploy-pages / deploy (push) Canceled after 0s
i18n consistency check / check (push) Canceled after 0s
provider adoption tests / test (chapter2/context-compression) (push) Canceled after 0s
provider adoption tests / test (chapter2/prompt-injection) (push) Canceled after 0s
provider adoption tests / test (chapter2/system-hint) (push) Canceled after 0s
provider adoption tests / test (chapter3/log-sanitization) (push) Canceled after 0s
web-search-agent tests / test (push) Canceled after 0s
web-search-agent tests / agentbook (push) Canceled after 0s
67 lines
3.7 KiB
XML
67 lines
3.7 KiB
XML
<?xml version="1.0" encoding="UTF-8"?>
|
|
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1000 430" width="1000" height="430" style="background:#ffffff">
|
|
<defs>
|
|
<marker id="arrow" markerWidth="10" markerHeight="8" refX="9" refY="4" orient="auto">
|
|
<path d="M0,0 L10,4 L0,8 Z" fill="#33434d"/>
|
|
</marker>
|
|
<style>
|
|
.t{font-family:Arial,'PingFang SC','Microsoft YaHei',sans-serif;fill:#26343c}
|
|
.h{font-size:20px;font-weight:700}
|
|
.s{font-size:15px;fill:#52636d}
|
|
.label{font-size:13px;font-weight:700;fill:#2f617a}
|
|
.value{font-size:15px;fill:#3f515c}
|
|
.box{fill:#f5f7f8;stroke:#33434d;stroke-width:2}
|
|
.accent{fill:#e8f1f5;stroke:#2f617a;stroke-width:2}
|
|
.line{fill:none;stroke:#33434d;stroke-width:2;marker-end:url(#arrow)}
|
|
</style>
|
|
</defs>
|
|
|
|
<text x="500" y="30" text-anchor="middle" class="t h">Three-layer trajectory verification: keep evidence coming from the real environment</text>
|
|
|
|
<!-- 三类验证器 -->
|
|
<rect x="30" y="76" width="245" height="82" rx="8" class="box"/>
|
|
<text x="152.5" y="107" text-anchor="middle" class="t h">Outcome verifier</text>
|
|
<text x="152.5" y="135" text-anchor="middle" class="t s">Tests, amounts, final order state</text>
|
|
|
|
<rect x="30" y="174" width="245" height="82" rx="8" class="box"/>
|
|
<text x="152.5" y="205" text-anchor="middle" class="t h">Process verifier</text>
|
|
<text x="152.5" y="233" text-anchor="middle" class="t s" style="font-size:12px">Policy, permissions, actions, commitments</text>
|
|
|
|
<rect x="30" y="272" width="245" height="82" rx="8" class="accent"/>
|
|
<text x="152.5" y="303" text-anchor="middle" class="t h">LLM Rubric</text>
|
|
<text x="152.5" y="331" text-anchor="middle" class="t s" style="font-size:11px">Naturalness, concision, compliant workarounds</text>
|
|
|
|
<!-- 结构化评价：改用纵向字段，避免中文字体宽度导致双列重叠 -->
|
|
<rect x="342" y="62" width="320" height="340" rx="10" class="accent"/>
|
|
<text x="502" y="97" text-anchor="middle" class="t h">Structured trajectory evaluation</text>
|
|
<line x1="374" y1="117" x2="630" y2="117" stroke="#89a0ac"/>
|
|
|
|
<text x="376" y="145" class="t label">Task outcome</text>
|
|
<text x="376" y="168" class="t value">Success, partial success or failure</text>
|
|
|
|
<text x="376" y="198" class="t label">Per-dimension scoring</text>
|
|
<text x="376" y="221" class="t value">Score each item, not just a total</text>
|
|
|
|
<text x="376" y="251" class="t label">Evidence location</text>
|
|
<text x="376" y="274" class="t value">Mark the turn and the tool state</text>
|
|
|
|
<text x="376" y="304" class="t label">Failure labels</text>
|
|
<text x="376" y="327" class="t value">Keep the problem type for later diagnosis</text>
|
|
|
|
<line x1="374" y1="348" x2="630" y2="348" stroke="#b6c7cf"/>
|
|
<text x="502" y="378" text-anchor="middle" class="t s">Allow refusal to score when evidence is thin</text>
|
|
|
|
<!-- 供进化模块使用的诊断信号 -->
|
|
<rect x="712" y="119" width="258" height="202" rx="10" class="box"/>
|
|
<text x="841" y="156" text-anchor="middle" class="t h" style="font-size:19px">Learnable diagnostic signal</text>
|
|
<text x="841" y="199" text-anchor="middle" class="t s" style="font-size:12.5px">Keeps where success and failure happened</text>
|
|
<text x="841" y="234" text-anchor="middle" class="t s" style="font-size:13px">States the evidence and the problem type</text>
|
|
<text x="841" y="269" text-anchor="middle" class="t s" style="font-size:10px">Hand off to a separate module to pick what to update</text>
|
|
|
|
<!-- 数据流 -->
|
|
<path d="M275 117 H307 V151 H340" class="line"/>
|
|
<path d="M275 215 H340" class="line"/>
|
|
<path d="M275 313 H307 V279 H340" class="line"/>
|
|
<path d="M662 220 H710" class="line"/>
|
|
</svg>
|