Skip to content

Clarify baseline timeline and October 21 methodology change #42

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
exe-dev-github-integration merged 1 commit into from
Oct 26, 2026
Merged

Clarify baseline timeline and October 21 methodology change #42

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 19 additions & 4 deletions index.html
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -538,7 +538,7 @@
};

/* ---------------- time-series chart ---------------- */
// Before Oct 21 only the median is visible until hovered or selected; then
// Before Oct 21 the Baseline median is blue until hovered or selected; then
// model answers appear as ink-gray circles. Since Oct 21 each prompt condition
// has a colored dot, with model answers in its details. No connecting lines.
// The ensemble median remains an ink mark, for the ONE horizon
Expand Down Expand Up @@ -569,6 +569,9 @@
const axisEnd = n > 1 ? Math.max(days[n - 1], yearEnd) : yearEnd;
const span = axisEnd > days[0] ? axisEnd - days[0] : 0;
const xDay = day => (span === 0 ? (pad.l + W - pad.r) / 2 : pad.l + ((day - days[0]) / span) * (W - pad.l - pad.r));
const changeDay = Date.parse(TL_CONDITION_START + "T00:00:00Z") / 864e5;
const showChange = n > 0 && changeDay >= days[0] && changeDay <= axisEnd;
const isBaseline = i => snapshots[i].date < TL_CONDITION_START;
const x = i => xDay(days[i]);
const xs = snapshots.map((_, i) => x(i));
const futureTicks = [];
Expand Down Expand Up @@ -694,6 +697,9 @@
))}
{axisEnd > days[n - 1] && W - pad.r - xs[n - 1] >= labelGap && <text x={W - pad.r} y={H - 18} textAnchor="end"
fontSize="12" fill="var(--ink-faint)" fontWeight="600">Dec 31</text>}
{showChange && <line x1={xDay(changeDay)} x2={xDay(changeDay)}
y1={pad.t} y2={H - pad.b} stroke="#cf222e" strokeWidth="1.5"
strokeDasharray="1 4" strokeLinecap="round" />}
{/* Historical model answers are revealed only for the active date. */}
{series.map(s => (
<g key={s.label}>
Expand All @@ -719,12 +725,15 @@
{/* The median's marks, unjoined. The larger transparent target also works on touchscreens. */}
{med.map((v, i) => v == null ? null : (
<g key={"m" + i} role="button" tabIndex={0} style={{ cursor: "pointer" }}
aria-label={`${snapshots[i].date}, ensemble median ${fmtV(v)}; show model forecasts`}
aria-label={`${snapshots[i].date}, ${isBaseline(i) ? "Baseline prompt" : "ensemble median"} ${fmtV(v)}; show model forecasts`}
aria-pressed={pinned && hv === i && !hvar}
onMouseEnter={e => hover(e, i)} onMouseLeave={leave} onFocus={e => hover(e, i)} onBlur={leave}
onClick={e => select(e, i)} onKeyDown={e => keySelect(e, i)}>
<circle cx={x(i)} cy={y(v)} r="10" fill="transparent" />
<circle cx={x(i)} cy={y(v)} r="5" fill="var(--panel)" stroke="var(--ink)" strokeWidth="2.6" />
<circle cx={x(i)} cy={y(v)} r={isBaseline(i) ? "4.5" : "5"}
fill={isBaseline(i) ? TL_CONDITIONS.none.color : "var(--panel)"}
stroke={isBaseline(i) ? "var(--panel)" : "var(--ink)"}
strokeWidth={isBaseline(i) ? "1.5" : "2.6"} />
</g>
))}

Expand All @@ -742,6 +751,10 @@
))}
</svg>

{showChange && <div style={{ color: "#cf222e", fontSize: 11, marginTop: 2 }}>
Oct 21, 2026: Methodology change to add more prompting strategies
</div>}

{hovered && (
<PointTooltip anchor={active.anchor}>
{hoveredModels.map(([label, val]) => (
Expand All @@ -762,7 +775,7 @@
: snapshots[hv].draws > 1 ? ` · ${snapshots[hv].draws} draws per model · point = mean` : ""}{loss ? " · deaths" : ""}
</div>
<div style={{ display: "flex", justifyContent: "space-between", gap: 14, fontSize: 12.5, fontWeight: 600, marginBottom: 4 }}>
<span>Ensemble median</span>
<span>{isBaseline(hv) ? "Baseline prompt · ensemble median" : "Ensemble median"}</span>
<span className="mono">
{fmtV(med[hv])}
</span>
Expand Down Expand Up @@ -900,6 +913,8 @@ <h2 style={{ fontSize: 22, marginTop: 0, marginBottom: 4 }}>{title}</h2>
<div style={{ marginTop: 16, minWidth: 0 }}>
<div style={{ display: "flex", gap: 16, flexWrap: "wrap", marginBottom: 6, alignItems: "center" }}>
<LegendDot color="var(--ink)" label="Ensemble median" hollow />
{!variants.some(v => v.id === "none") && snapshots.some((s, i) => s.date < TL_CONDITION_START && med[i] != null) &&
<LegendDot color={TL_CONDITIONS.none.color} label="Baseline" />}
{variants.map(v => <LegendDot key={v.id} color={v.color} label={v.label} title={v.text || "No added instruction."} />)}
{/* The scale switch at the legend row's right end, where Graph 2 keeps its own. */}
{!loss && <ScaleToggle value={scale} onChange={setScale} style={{ marginLeft: "auto" }} />}
Expand Down
23 changes: 19 additions & 4 deletions web/demo/19-live-timeline.jsx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -50,7 +50,7 @@ const tlVariants = (q, horizon, snapshots) => {
};

/* ---------------- time-series chart ---------------- */
// Before Oct 21 only the median is visible until hovered or selected; then
// Before Oct 21 the Baseline median is blue until hovered or selected; then
// model answers appear as ink-gray circles. Since Oct 21 each prompt condition
// has a colored dot, with model answers in its details. No connecting lines.
// The ensemble median remains an ink mark, for the ONE horizon
Expand Down Expand Up @@ -81,6 +81,9 @@ function TimelineChart({ q, snapshots, horizon, scale }) {
const axisEnd = n > 1 ? Math.max(days[n - 1], yearEnd) : yearEnd;
const span = axisEnd > days[0] ? axisEnd - days[0] : 0;
const xDay = day => (span === 0 ? (pad.l + W - pad.r) / 2 : pad.l + ((day - days[0]) / span) * (W - pad.l - pad.r));
const changeDay = Date.parse(TL_CONDITION_START + "T00:00:00Z") / 864e5;
const showChange = n > 0 && changeDay >= days[0] && changeDay <= axisEnd;
const isBaseline = i => snapshots[i].date < TL_CONDITION_START;
const x = i => xDay(days[i]);
const xs = snapshots.map((_, i) => x(i));
const futureTicks = [];
Expand Down Expand Up @@ -206,6 +209,9 @@ function TimelineChart({ q, snapshots, horizon, scale }) {
))}
{axisEnd > days[n - 1] && W - pad.r - xs[n - 1] >= labelGap && <text x={W - pad.r} y={H - 18} textAnchor="end"
fontSize="12" fill="var(--ink-faint)" fontWeight="600">Dec 31</text>}
{showChange && <line x1={xDay(changeDay)} x2={xDay(changeDay)}
y1={pad.t} y2={H - pad.b} stroke="#cf222e" strokeWidth="1.5"
strokeDasharray="1 4" strokeLinecap="round" />}
{/* Historical model answers are revealed only for the active date. */}
{series.map(s => (
<g key={s.label}>
Expand All @@ -231,12 +237,15 @@ function TimelineChart({ q, snapshots, horizon, scale }) {
{/* The median's marks, unjoined. The larger transparent target also works on touchscreens. */}
{med.map((v, i) => v == null ? null : (
<g key={"m" + i} role="button" tabIndex={0} style={{ cursor: "pointer" }}
aria-label={`${snapshots[i].date}, ensemble median ${fmtV(v)}; show model forecasts`}
aria-label={`${snapshots[i].date}, ${isBaseline(i) ? "Baseline prompt" : "ensemble median"} ${fmtV(v)}; show model forecasts`}
aria-pressed={pinned && hv === i && !hvar}
onMouseEnter={e => hover(e, i)} onMouseLeave={leave} onFocus={e => hover(e, i)} onBlur={leave}
onClick={e => select(e, i)} onKeyDown={e => keySelect(e, i)}>
<circle cx={x(i)} cy={y(v)} r="10" fill="transparent" />
<circle cx={x(i)} cy={y(v)} r="5" fill="var(--panel)" stroke="var(--ink)" strokeWidth="2.6" />
<circle cx={x(i)} cy={y(v)} r={isBaseline(i) ? "4.5" : "5"}
fill={isBaseline(i) ? TL_CONDITIONS.none.color : "var(--panel)"}
stroke={isBaseline(i) ? "var(--panel)" : "var(--ink)"}
strokeWidth={isBaseline(i) ? "1.5" : "2.6"} />
</g>
))}

Expand All @@ -254,6 +263,10 @@ function TimelineChart({ q, snapshots, horizon, scale }) {
))}
</svg>

{showChange && <div style={{ color: "#cf222e", fontSize: 11, marginTop: 2 }}>
Oct 21, 2026: Methodology change to add more prompting strategies
</div>}

{hovered && (
<PointTooltip anchor={active.anchor}>
{hoveredModels.map(([label, val]) => (
Expand All @@ -274,7 +287,7 @@ function TimelineChart({ q, snapshots, horizon, scale }) {
: snapshots[hv].draws > 1 ? ` · ${snapshots[hv].draws} draws per model · point = mean` : ""}{loss ? " · deaths" : ""}
</div>
<div style={{ display: "flex", justifyContent: "space-between", gap: 14, fontSize: 12.5, fontWeight: 600, marginBottom: 4 }}>
<span>Ensemble median</span>
<span>{isBaseline(hv) ? "Baseline prompt · ensemble median" : "Ensemble median"}</span>
<span className="mono">
{fmtV(med[hv])}
</span>
Expand Down Expand Up @@ -412,6 +425,8 @@ function TimelinePanelCurrent({ qid = TL_HEADLINE, title = "Probability of an AI
<div style={{ marginTop: 16, minWidth: 0 }}>
<div style={{ display: "flex", gap: 16, flexWrap: "wrap", marginBottom: 6, alignItems: "center" }}>
<LegendDot color="var(--ink)" label="Ensemble median" hollow />
{!variants.some(v => v.id === "none") && snapshots.some((s, i) => s.date < TL_CONDITION_START && med[i] != null) &&
<LegendDot color={TL_CONDITIONS.none.color} label="Baseline" />}
{variants.map(v => <LegendDot key={v.id} color={v.color} label={v.label} title={v.text || "No added instruction."} />)}
{/* The scale switch at the legend row's right end, where Graph 2 keeps its own. */}
{!loss && <ScaleToggle value={scale} onChange={setScale} style={{ marginLeft: "auto" }} />}
Expand Down
Loading