diff --git a/index.html b/index.html index 7c3122d..7a7e9ae 100644 --- a/index.html +++ b/index.html @@ -460,17 +460,177 @@ .comparison-frame { margin-top: 18px; + padding: 18px; + color: #f8faf7; background: #0b1018; border: 1px solid var(--line-strong); border-radius: 8px; - overflow: hidden; box-shadow: var(--shadow); } - .comparison-frame img { - display: block; - width: 100%; - background: #0b1018; + .comparison-head { + display: flex; + justify-content: space-between; + gap: 18px; + align-items: end; + margin-bottom: 16px; + } + + .comparison-heading span { + color: #f4c542; + font-family: var(--mono); + font-size: 0.72rem; + font-weight: 700; + text-transform: uppercase; + } + + .comparison-heading h3 { + margin: 6px 0 0; + font-family: var(--display); + font-size: 2rem; + line-height: 1; + letter-spacing: 0; + } + + .comparison-legend { + display: flex; + flex-wrap: wrap; + justify-content: flex-end; + gap: 8px; + max-width: 560px; + } + + .legend-chip { + display: inline-flex; + align-items: center; + gap: 7px; + min-height: 26px; + padding: 5px 8px; + color: rgba(248, 250, 247, 0.82); + background: rgba(248, 250, 247, 0.06); + border: 1px solid rgba(248, 250, 247, 0.14); + border-radius: 6px; + font-family: var(--mono); + font-size: 0.67rem; + } + + .legend-chip::before { + content: ""; + width: 10px; + height: 10px; + background: var(--color); + border-radius: 3px; + } + + .comparison-grid { + display: grid; + grid-template-columns: repeat(3, minmax(0, 1fr)); + gap: 12px; + } + + .benchmark-panel { + min-width: 0; + padding: 14px; + background: + linear-gradient(180deg, rgba(248, 250, 247, 0.055), rgba(248, 250, 247, 0.02)), + rgba(13, 17, 23, 0.72); + border: 1px solid rgba(248, 250, 247, 0.13); + border-radius: 8px; + } + + .benchmark-top { + display: flex; + justify-content: space-between; + gap: 10px; + align-items: start; + margin-bottom: 10px; + } + + .benchmark-top h4 { + margin: 0; + font-family: var(--display); + font-size: 1.28rem; + line-height: 1; + letter-spacing: 0; + } + + .delta-pill { + flex: none; + padding: 5px 8px; + color: #87d996; + background: rgba(46, 123, 63, 0.2); + border: 1px solid rgba(135, 217, 150, 0.5); + border-radius: 6px; + font-family: var(--mono); + font-size: 0.67rem; + font-weight: 700; + white-space: nowrap; + } + + .bar-stage { + position: relative; + display: flex; + align-items: flex-end; + gap: 6px; + height: 170px; + padding: 24px 8px 22px 34px; + background: + linear-gradient(rgba(248, 250, 247, 0.08) 1px, transparent 1px) 0 24px / 100% 42px, + rgba(0, 0, 0, 0.16); + border-left: 1px solid rgba(248, 250, 247, 0.24); + border-bottom: 1px solid rgba(248, 250, 247, 0.24); + border-radius: 6px; + } + + .axis-range { + position: absolute; + left: 7px; + bottom: 6px; + color: rgba(248, 250, 247, 0.46); + font-family: var(--mono); + font-size: 0.58rem; + writing-mode: vertical-rl; + transform: rotate(180deg); + text-transform: uppercase; + } + + .method-bar { + position: relative; + flex: 1; + min-width: 0; + height: max(8px, var(--h)); + background: var(--color); + border-radius: 4px 4px 2px 2px; + opacity: 0.86; + } + + .method-bar.skillopt { + border: 2px solid rgba(248, 250, 247, 0.82); + box-shadow: 0 0 16px rgba(135, 217, 150, 0.62); + opacity: 1; + } + + .method-bar span { + position: absolute; + left: 50%; + bottom: calc(100% + 6px); + transform: translateX(-50%); + padding: 2px 5px; + color: #0b1018; + background: #87d996; + border-radius: 5px; + font-family: var(--mono); + font-size: 0.62rem; + font-weight: 800; + white-space: nowrap; + } + + .comparison-caption { + margin: 14px 0 0; + color: rgba(248, 250, 247, 0.62); + font-family: var(--mono); + font-size: 0.72rem; + line-height: 1.55; } .caption { @@ -1059,6 +1219,19 @@ grid-template-columns: 1fr; } + .comparison-head { + align-items: flex-start; + flex-direction: column; + } + + .comparison-legend { + justify-content: flex-start; + } + + .comparison-grid { + grid-template-columns: repeat(2, minmax(0, 1fr)); + } + .hero h1 { font-size: 4.1rem; } @@ -1103,6 +1276,7 @@ .method-grid, .ablation-summary .mini-list, + .comparison-grid, .transfer-grid, .evolution-footnotes, .steps { @@ -1132,6 +1306,10 @@ border-left: 0; } + .bar-stage { + height: 150px; + } + .chart-toolbar, .chart-caption { flex-direction: column; @@ -1474,12 +1652,19 @@ -
- Bar charts comparing SkillOpt with no skill, human skill, LLM skill, Trace2Skill, TextGrad, and GEPA across SearchQA, SpreadsheetBench, OfficeQA, DocVQA, LiveMath, and ALFWorld. -
- Method comparison from the project video. Bars report per-benchmark direct-chat accuracy averaged over seven target models; SkillOpt is best or tied-best in every panel. -
-
+
+
+
+ Method comparison +

SkillOpt clears the strongest baseline on every benchmark.

+
+
+
+
+

+ Rendered from the raw values used in the project-video results scene. Bars report direct-chat benchmark means across seven target models; each panel uses a zoomed y-axis to keep method gaps readable. +

+
@@ -1758,6 +1943,129 @@