fix: correct dates for posts

Signed-off-by: Aaron Pham <[email protected]>

Author: aarnphm

_posts/2025-01-14-struct-decode-intro.md

@@ -53,14 +53,14 @@ Companies like OpenAI have recognized this need, implementing features like [JSO
 
 In simple terms, structured decoding gives LLMs a "template" to follow. Users provide a schema that "influences" the model's output, ensuring compliance with the desired structure:
 
-<svg id="mermaid-1736451893652" width="100%" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" class="flowchart" style="max-width: 1285px;" viewBox="0 0 1285 307.20703125" role="graphics-document document" aria-roledescription="flowchart-v2"><style>#mermaid-1736451893652{font-family:"Berkeley Mono","JetBrains Mono",ui-monospace,SFMono-Regular,"SF Mono",Menlo,monospace;font-size:16px;fill:#797593;}#mermaid-1736451893652 .error-icon{fill:#b4637a;}#mermaid-1736451893652 .error-text{fill:#4b9c85;stroke:#4b9c85;}#mermaid-1736451893652 .edge-thickness-normal{stroke-width:1px;}#mermaid-1736451893652 .edge-thickness-thick{stroke-width:3.5px;}#mermaid-1736451893652 .edge-pattern-solid{stroke-dasharray:0;}#mermaid-1736451893652 .edge-thickness-invisible{stroke-width:0;fill:none;}#mermaid-1736451893652 .edge-pattern-dashed{stroke-dasharray:3;}#mermaid-1736451893652 .edge-pattern-dotted{stroke-dasharray:2;}#mermaid-1736451893652 .marker{fill:#797593;stroke:#797593;}#mermaid-1736451893652 .marker.cross{stroke:#797593;}#mermaid-1736451893652 svg{font-family:"Berkeley Mono","JetBrains Mono",ui-monospace,SFMono-Regular,"SF Mono",Menlo,monospace;font-size:16px;}#mermaid-1736451893652 p{margin:0;}#mermaid-1736451893652 .label{font-family:"Berkeley Mono","JetBrains Mono",ui-monospace,SFMono-Regular,"SF Mono",Menlo,monospace;color:#797593;}#mermaid-1736451893652 .cluster-label text{fill:#4b9c85;}#mermaid-1736451893652 .cluster-label span{color:#4b9c85;}#mermaid-1736451893652 .cluster-label span p{background-color:transparent;}#mermaid-1736451893652 .label text,#mermaid-1736451893652 span{fill:#797593;color:#797593;}#mermaid-1736451893652 .node rect,#mermaid-1736451893652 .node circle,#mermaid-1736451893652 .node ellipse,#mermaid-1736451893652 .node polygon,#mermaid-1736451893652 .node path{fill:#fffaf3;stroke:#b4637a;stroke-width:1px;}#mermaid-1736451893652 .rough-node .label text,#mermaid-1736451893652 .node .label text,#mermaid-1736451893652 .image-shape .label,#mermaid-1736451893652 .icon-shape .label{text-anchor:middle;}#mermaid-1736451893652 .node .katex path{fill:#000;stroke:#000;stroke-width:1px;}#mermaid-1736451893652 .rough-node .label,#mermaid-1736451893652 .node .label,#mermaid-1736451893652 .image-shape .label,#mermaid-1736451893652 .icon-shape .label{text-align:center;}#mermaid-1736451893652 .node.clickable{cursor:pointer;}#mermaid-1736451893652 .root .anchor path{fill:#797593!important;stroke-width:0;stroke:#797593;}#mermaid-1736451893652 .arrowheadPath{fill:#0b0b0b;}#mermaid-1736451893652 .edgePath .path{stroke:#797593;stroke-width:2.0px;}#mermaid-1736451893652 .flowchart-link{stroke:#797593;fill:none;}#mermaid-1736451893652 .edgeLabel{background-color:#8f9fa926;text-align:center;}#mermaid-1736451893652 .edgeLabel p{background-color:#8f9fa926;}#mermaid-1736451893652 .edgeLabel rect{opacity:0.5;background-color:#8f9fa926;fill:#8f9fa926;}#mermaid-1736451893652 .labelBkg{background-color:rgba(143, 159, 169, 0.5);}#mermaid-1736451893652 .cluster rect{fill:#fffaf3;stroke:hsl(342.962962963, 0%, 44.7058823529%);stroke-width:1px;}#mermaid-1736451893652 .cluster text{fill:#4b9c85;}#mermaid-1736451893652 .cluster span{color:#4b9c85;}#mermaid-1736451893652 div.mermaidTooltip{position:absolute;text-align:center;max-width:200px;padding:2px;font-family:"Berkeley Mono","JetBrains Mono",ui-monospace,SFMono-Regular,"SF Mono",Menlo,monospace;font-size:12px;background:#b4637a;border:1px solid hsl(342.962962963, 0%, 44.7058823529%);border-radius:2px;pointer-events:none;z-index:100;}#mermaid-1736451893652 .flowchartTitleText{text-anchor:middle;font-size:18px;fill:#797593;}#mermaid-1736451893652 rect.text{fill:none;stroke-width:0;}#mermaid-1736451893652 .icon-shape,#mermaid-1736451893652 .image-shape{background-color:#8f9fa926;text-align:center;}#mermaid-1736451893652 .icon-shape p,#mermaid-1736451893652 .image-shape p{background-color:#8f9fa926;padding:2px;}#mermaid-1736451893652 .icon-shape rect,#mermaid-1736451893652 .image-shape rect{opacity:0.5;background-color:#8f9fa926;fill:#8f9fa926;}#mermaid-1736451893652 :root{--mermaid-font-family:"trebuchet ms",verdana,arial,sans-serif;}#mermaid-1736451893652 .default&gt;\*{fill:#fff!important;stroke:#333!important;stroke-width:2px!important;}#mermaid-1736451893652 .default span{fill:#fff!important;stroke:#333!important;stroke-width:2px!important;}</style><g><marker id="mermaid-1736451893652_flowchart-v2-pointEnd" class="marker flowchart-v2" viewBox="0 0 10 10" refX="5" refY="5" markerUnits="userSpaceOnUse" markerWidth="8" markerHeight="8" orient="auto"><path d="M 0 0 L 10 5 L 0 10 z" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"></path></marker><marker id="mermaid-1736451893652_flowchart-v2-pointStart" class="marker flowchart-v2" viewBox="0 0 10 10" refX="4.5" refY="5" markerUnits="userSpaceOnUse" markerWidth="8" markerHeight="8" orient="auto"><path d="M 0 5 L 10 10 L 10 0 z" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"></path></marker><marker id="mermaid-1736451893652_flowchart-v2-circleEnd" class="marker flowchart-v2" viewBox="0 0 10 10" refX="11" refY="5" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><circle cx="5" cy="5" r="5" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"></circle></marker><marker id="mermaid-1736451893652_flowchart-v2-circleStart" class="marker flowchart-v2" viewBox="0 0 10 10" refX="-1" refY="5" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><circle cx="5" cy="5" r="5" class="arrowMarkerPath" style="stroke-width: 1; stroke-dasharray: 1, 0;"></circle></marker><marker id="mermaid-1736451893652_flowchart-v2-crossEnd" class="marker cross flowchart-v2" viewBox="0 0 11 11" refX="12" refY="5.2" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><path d="M 1,1 l 9,9 M 10,1 l -9,9" class="arrowMarkerPath" style="stroke-width: 2; stroke-dasharray: 1, 0;"></path></marker><marker id="mermaid-1736451893652_flowchart-v2-crossStart" class="marker cross flowchart-v2" viewBox="0 0 11 11" refX="-1" refY="5.2" markerUnits="userSpaceOnUse" markerWidth="11" markerHeight="11" orient="auto"><path d="M 1,1 l 9,9 M 10,1 l -9,9" class="arrowMarkerPath" style="stroke-width: 2; stroke-dasharray: 1, 0;"></path></marker><g class="root"><g class="clusters"><g class="cluster " id="Decoding" data-look="classic"><rect style="" x="266.5" y="8" width="739" height="144.00390625"></rect><g class="cluster-label " transform="translate(598, 8)"><foreignObject width="76" height="27.203125"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel "><p>Decoding</p></span></div></foreignObject></g></g><g class="cluster " id="subGraph0" data-look="classic"><rect style="" x="8" y="172.00390625" width="501.5" height="127.203125"></rect><g class="cluster-label " transform="translate(163.75, 172.00390625)"><foreignObject width="190" height="27.203125"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel "><p>Structured automaton</p></span></div></foreignObject></g></g></g><g class="edgePaths"><path d="M150,235.605L159.708,235.605C169.417,235.605,188.833,235.605,208.25,235.605C227.667,235.605,247.083,235.605,263.458,235.605C279.833,235.605,293.167,235.605,299.833,235.605L306.5,235.605" id="L_A_B_0" class=" edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style="" marker-end="url(#mermaid-1736451893652_flowchart-v2-pointEnd)"></path><path d="M484.5,71.602L488.667,71.602C492.833,71.602,501.167,71.602,523.75,71.602C546.333,71.602,583.167,71.602,619.335,72.868C655.503,74.134,691.007,76.667,708.758,77.934L726.51,79.2" id="L_C_D_1" class=" edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style="" marker-end="url(#mermaid-1736451893652_flowchart-v2-pointEnd)"></path><path d="M465.5,235.605L472.833,235.605C480.167,235.605,494.833,235.605,520.583,214.705C546.333,193.805,583.167,152.004,619.336,129.46C655.506,106.916,691.011,103.629,708.764,101.986L726.517,100.343" id="L_B_D_2" class=" edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style="" marker-end="url(#mermaid-1736451893652_flowchart-v2-pointEnd)"></path><path d="M980.5,88.402L984.667,88.402C988.833,88.402,997.167,88.402,1005.5,88.402C1013.833,88.402,1022.167,88.402,1029.833,88.402C1037.5,88.402,1044.5,88.402,1048,88.402L1051.5,88.402" id="L_D_E_3" class=" edge-thickness-normal edge-pattern-solid edge-thickness-normal edge-pattern-solid flowchart-link" style="" marker-end="url(#mermaid-1736451893652_flowchart-v2-pointEnd)"></path></g><g class="edgeLabels"><g class="edgeLabel" transform="translate(208.25, 235.60546875)"><g class="label" transform="translate(-33.25, -13.6015625)"><foreignObject width="66.5" height="27.203125"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel "><p>Defines</p></span></div></foreignObject></g></g><g class="edgeLabel" transform="translate(620, 71.6015625)"><g class="label" transform="translate(-85.5, -13.6015625)"><foreignObject width="171" height="27.203125"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel "><p>Token distribution</p></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel "></span></div></foreignObject></g></g><g class="edgeLabel"><g class="label" transform="translate(0, 0)"><foreignObject width="0" height="0"><div xmlns="http://www.w3.org/1999/xhtml" class="labelBkg" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="edgeLabel "></span></div></foreignObject></g></g></g><g class="nodes"><g class="node default  " id="flowchart-A-13" transform="translate(91.5, 235.60546875)"><rect class="basic label-container" style="fill:#e1f3f8 !important;stroke:#333 !important;stroke-width:2px !important" x="-58.5" y="-28.6015625" width="117" height="57.203125"></rect><g class="label" style="" transform="translate(-28.5, -13.6015625)"><rect></rect><foreignObject width="57" height="27.203125"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel "><p>Schema</p></span></div></foreignObject></g></g><g class="node default  " id="flowchart-B-14" transform="translate(388, 235.60546875)"><rect class="basic label-container" style="fill:#e1f3f8 !important;stroke:#333 !important;stroke-width:2px !important" x="-77.5" y="-28.6015625" width="155" height="57.203125"></rect><g class="label" style="" transform="translate(-47.5, -13.6015625)"><rect></rect><foreignObject width="95" height="27.203125"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel "><p>Logit Bias</p></span></div></foreignObject></g></g><g class="node default  " id="flowchart-C-15" transform="translate(388, 71.6015625)"><rect class="basic label-container" style="fill:#f8e1e1 !important;stroke:#333 !important;stroke-width:2px !important" x="-96.5" y="-28.6015625" width="193" height="57.203125"></rect><g class="label" style="" transform="translate(-66.5, -13.6015625)"><rect></rect><foreignObject width="133" height="27.203125"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel "><p>Language Model</p></span></div></foreignObject></g></g><g class="node default  " id="flowchart-D-16" transform="translate(855.5, 88.40234375)"><rect class="basic label-container" style="fill:#f8e1e1 !important;stroke:#333 !important;stroke-width:2px !important" x="-125" y="-28.6015625" width="250" height="57.203125"></rect><g class="label" style="" transform="translate(-95, -13.6015625)"><rect></rect><foreignObject width="190" height="27.203125"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel "><p>Constrained Sampling</p></span></div></foreignObject></g></g><g class="node default  " id="flowchart-E-20" transform="translate(1166.25, 88.40234375)"><rect class="basic label-container" style="fill:#e1f8e1 !important;stroke:#333 !important;stroke-width:2px !important" x="-110.75" y="-28.6015625" width="221.5" height="57.203125"></rect><g class="label" style="" transform="translate(-80.75, -13.6015625)"><rect></rect><foreignObject width="161.5" height="27.203125"><div xmlns="http://www.w3.org/1999/xhtml" style="display: table-cell; white-space: nowrap; line-height: 1.5; max-width: 200px; text-align: center;"><span class="nodeLabel "><p>Structured Output</p></span></div></foreignObject></g></g></g></g></g></svg>
+<img src="/assets/figures/struct-decode-intro/mermaid-intro.svg" alt="top level view of structure decoding" />
 
 From a technical perspective, an inference engine can modify the probability distribution for next-tokens by applying bias (often via logit masks) for all tokens from any given schemas. To apply these biases, [outlines](https://github.com/dottxt-ai/outlines) proposed guided generations via finite-state machine (FSM) for any given schemas (Willard & Louf, 2023). This allows us to track the current state during decoding and filter out invalid tokens by applying logit bias to the output.
 
 <figure>
   <img src="/assets/figures/struct-decode-intro/constrained-json-fsm.webp" />
 <figcaption>
-courtesy of <a href="https://lmsys.org/blog/2024-02-05-compressed-fsm/" target="_blank"><i>LMSys, 2024</i></a>.
+courtesy of <a href="https://lmsys.org/blog/2024-02-05-compressed-fsm/" target="_blank">LMSys, 2024</a>.
 </figcaption>
 </figure>
 

_sass/minima/custom-styles.scss

@@ -1,2 +1,8 @@
 // Placeholder to allow defining custom styles that override everything else.
 // (Use `_sass/minima/custom-variables.scss` to override variable defaults)
+
+figure {
+  & > figcaption {
+    font-style: italic;
+  }
+}