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
25 lines
5.3 KiB
XML
25 lines
5.3 KiB
XML
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 40 780 340" width="780" height="340" style="background:#ffffff">
|
||
<defs><marker id="ah" markerWidth="12" markerHeight="8" refX="12" refY="4" orient="auto"><polygon points="0 0, 12 4, 0 8" fill="#333333"/></marker><marker id="ah-light" markerWidth="12" markerHeight="8" refX="12" refY="4" orient="auto"><polygon points="0 0, 12 4, 0 8" fill="#999999"/></marker></defs>
|
||
|
||
<rect x="40" y="65" width="300" height="130" rx="6" fill="#f0f0f0" stroke="#333333" stroke-width="2"/>
|
||
<text x="190" y="88" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="20" fill="#333333" text-anchor="middle" dominant-baseline="central" font-weight="bold">阶段一：SFT 固化格式</text>
|
||
<text x="190" y="115" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="14" fill="#666666" text-anchor="middle" dominant-baseline="central" font-weight="normal">目标：输出可解析（JSON/工具调用）</text>
|
||
<text x="190" y="137" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="14" fill="#666666" text-anchor="middle" dominant-baseline="central" font-weight="normal">数据：数千条高质量示范</text>
|
||
<text x="190" y="159" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="14" fill="#666666" text-anchor="middle" dominant-baseline="central" font-weight="normal">停止条件：格式稳定，能力初具</text>
|
||
<text x="190" y="181" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="14" fill="#666666" text-anchor="middle" dominant-baseline="central" font-weight="normal">⚠ 过度训练 → 模型塌缩到训练分布</text>
|
||
<line x1="342" y1="130" x2="398" y2="130" stroke="#333333" stroke-width="2" marker-end="url(#ah)"/>
|
||
<text x="370" y="115" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="14" fill="#666666" text-anchor="middle" dominant-baseline="central" font-weight="normal">格式</text>
|
||
<text x="370" y="135" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="14" fill="#666666" text-anchor="middle" dominant-baseline="central" font-weight="normal">稳定</text>
|
||
<rect x="400" y="65" width="340" height="130" rx="6" fill="#d0d0d0" stroke="#333333" stroke-width="2"/>
|
||
<text x="570" y="88" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="20" fill="#333333" text-anchor="middle" dominant-baseline="central" font-weight="bold">阶段二：RL 塑形策略</text>
|
||
<text x="570" y="115" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="14" fill="#666666" text-anchor="middle" dominant-baseline="central" font-weight="normal">目标：最大化任务奖励（准确率/成功率）</text>
|
||
<text x="570" y="137" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="14" fill="#666666" text-anchor="middle" dominant-baseline="central" font-weight="normal">前提：输出格式稳定 → 奖励可计算</text>
|
||
<text x="570" y="159" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="14" fill="#666666" text-anchor="middle" dominant-baseline="central" font-weight="normal">突破：发现超越 SFT 示范的新策略</text>
|
||
<text x="570" y="181" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="14" fill="#333333" text-anchor="middle" dominant-baseline="central" font-weight="normal">✓ 格式稳定 + 策略泛化 = 部署就绪</text>
|
||
<rect x="40" y="220" width="700" height="70" rx="4" fill="#f5f5f5" stroke="#333333" stroke-width="2"/>
|
||
<text x="390" y="238" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="16" fill="#333333" text-anchor="middle" dominant-baseline="central" font-weight="bold">为什么不能跳过 SFT 直接 RL？</text>
|
||
<text x="390" y="262" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="14" fill="#666666" text-anchor="middle" dominant-baseline="central" font-weight="normal">基础模型输出无结构 → 无法解析 JSON → 奖励函数返回 NaN → 梯度全零 → 训练完全失败</text>
|
||
<text x="390" y="310" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="20" fill="#333333" text-anchor="middle" dominant-baseline="central" font-weight="bold">SFT 记忆分布 → RL 泛化策略</text>
|
||
<text x="390" y="338" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="16" fill="#666666" text-anchor="middle" dominant-baseline="central" font-weight="normal">SFT: max Σ log P(y|x)（拟合训练分布） RL: max E[R(τ)]（优化任务目标）</text>
|
||
<text x="390" y="362" font-family="Arial, 'Helvetica Neue', Helvetica, 'PingFang SC', 'Microsoft YaHei', sans-serif" font-size="16" fill="#666666" text-anchor="middle" dominant-baseline="central" font-weight="normal">当"无论增加多少示范，新场景仍不佳"时 → 转向 RL 的临界点</text>
|
||
</svg> |
