Broaden categorical color palettes

.gitignore

@@ -6,6 +6,7 @@ __pycache__/
 *ipynb
 .vscode/
 
+Backup/
 eval-queue/
 eval-results/
 eval-queue-bk/

app.py

@@ -63,7 +63,7 @@ REPO_ID = "taagarwa/coding-agent-leaderboard"
 TOKEN = os.environ.get("HF_TOKEN")
 API = HfApi(token=TOKEN)
 COLOR_BY_CHOICES = ["Model", "Harness"]
-COLOR_PALETTE_CHOICES = ["Citrus", "Okabe-Ito", "High contrast"]
+COLOR_PALETTE_CHOICES = ["Citrus", "Okabe-Ito", "High contrast", "Rainbow"]
 DEFAULT_COLOR_PALETTE = "Citrus"
 
 

src/charts.py

@@ -9,45 +9,53 @@ import plotly.graph_objects as go
 from plotly.graph_objs._figure import Figure
 
 ColorBy = Literal["Model", "Harness"]
-PaletteName = Literal["Citrus", "Okabe-Ito", "High contrast"]
+PaletteName = Literal["Citrus", "Okabe-Ito", "High contrast", "Rainbow"]
 DEFAULT_PALETTE: PaletteName = "Citrus"
 
-# Separate, colorblind-friendly palettes for each grouping dimension.
-# Models use warm citrus-adjacent colors; harnesses use cool complementary colors.
-# Keep the two palettes non-overlapping so switching "Color by" is visually obvious.
+# Separate categorical palettes for each grouping dimension.
+# Model and harness colors intentionally start from different hue families so
+# switching "Color by" remains visually obvious.
 MODEL_COLORS: dict[str, str] = {
     "GPT 5.5 - high": "#F59E0B",  # amber
-    "Opus 4.8": "#EAB308",  # yellow
+    "Opus 4.8": "#84CC16",  # lime
     "RedHatAI/Qwen3.6-35B-A3B-NVFP4": "#F97316",  # orange
-    "Sonnet 4.6": "#D97706",  # dark amber
+    "Sonnet 4.6": "#22C55E",  # green
 }
 
 HARNESS_COLORS: dict[str, str] = {
     "Claude Code": "#06B6D4",  # cyan
     "Codex": "#3B82F6",  # blue
-    "OpenCode": "#14B8A6",  # teal
-    "OpenClaw": "#8B5CF6",  # violet
-    "Pi": "#EC4899",  # pink
-    "Qwen Code": "#64748B",  # slate
+    "OpenCode": "#8B5CF6",  # violet
+    "OpenClaw": "#EC4899",  # pink
+    "Pi": "#14B8A6",  # teal
+    "Qwen Code": "#F43F5E",  # rose
     "internal": "#94A3B8",
 }
 
 MODEL_FALLBACK_PALETTE = [
-    "#F59E0B",
-    "#EAB308",
-    "#F97316",
-    "#D97706",
-    "#FACC15",
-    "#FB923C",
+    "#F59E0B",  # amber
+    "#84CC16",  # lime
+    "#F97316",  # orange
+    "#22C55E",  # green
+    "#EAB308",  # yellow
+    "#FB7185",  # rose
+    "#A3E635",  # light lime
+    "#FACC15",  # gold
+    "#F472B6",  # pink
+    "#2DD4BF",  # teal
 ]
 
 HARNESS_FALLBACK_PALETTE = [
-    "#06B6D4",
-    "#3B82F6",
-    "#14B8A6",
-    "#8B5CF6",
-    "#EC4899",
-    "#64748B",
+    "#06B6D4",  # cyan
+    "#3B82F6",  # blue
+    "#8B5CF6",  # violet
+    "#EC4899",  # pink
+    "#14B8A6",  # teal
+    "#F43F5E",  # rose
+    "#6366F1",  # indigo
+    "#10B981",  # emerald
+    "#A855F7",  # purple
+    "#94A3B8",  # slate
 ]
 
 DARK_PAPER = "#15110F"
@@ -69,14 +77,22 @@ def clean_markdown_link(value: object) -> str:
 
 MODEL_PALETTES: dict[PaletteName, list[str]] = {
     "Citrus": MODEL_FALLBACK_PALETTE,
-    "Okabe-Ito": ["#E69F00", "#F0E442", "#D55E00", "#CC79A7", "#A6761D", "#FB9A99"],
-    "High contrast": ["#FFD166", "#FF9F1C", "#F77F00", "#F4D35E", "#EE964B", "#F95738"],
+    # Full Okabe-Ito palette. It is categorical and colorblind-friendly.
+    "Okabe-Ito": ["#E69F00", "#56B4E9", "#009E73", "#F0E442", "#0072B2", "#D55E00", "#CC79A7", "#999999"],
+    # High-contrast colors are intentionally broad, not just orange/yellow variants.
+    "High contrast": ["#FFD166", "#06D6A0", "#118AB2", "#EF476F", "#A78BFA", "#F97316", "#22D3EE", "#E5E7EB"],
+    # Categorical rainbow-style palette. This is not a continuous colorscale; it is
+    # sampled as discrete colors so each category gets a distinct color.
+    "Rainbow": ["#E6194B", "#F58231", "#FFE119", "#3CB44B", "#42D4F4", "#4363D8", "#911EB4", "#F032E6", "#469990", "#9A6324"],
 }
 
 HARNESS_PALETTES: dict[PaletteName, list[str]] = {
     "Citrus": HARNESS_FALLBACK_PALETTE,
-    "Okabe-Ito": ["#56B4E9", "#009E73", "#0072B2", "#CC79A7", "#999999", "#8DD3C7"],
-    "High contrast": ["#4CC9F0", "#4895EF", "#80FFDB", "#B5179E", "#7209B7", "#ADB5BD"],
+    # Use the same broad Okabe-Ito family but with a different starting point so the
+    # harness mode does not visually mirror the model mode.
+    "Okabe-Ito": ["#0072B2", "#D55E00", "#CC79A7", "#009E73", "#56B4E9", "#E69F00", "#F0E442", "#999999"],
+    "High contrast": ["#38BDF8", "#34D399", "#A78BFA", "#F472B6", "#FACC15", "#FB923C", "#22D3EE", "#E5E7EB"],
+    "Rainbow": ["#4363D8", "#E6194B", "#3CB44B", "#F58231", "#911EB4", "#42D4F4", "#F032E6", "#FFE119", "#469990", "#9A6324"],
 }
 
 
@@ -103,14 +119,42 @@ def get_color(name: str, color_by: ColorBy, palette_name: str | None = DEFAULT_P
     return stable_color(name, color_by, palette_key)
 
 
+def palette_colors_for(color_by: ColorBy, palette_name: str | None = DEFAULT_PALETTE) -> list[str]:
+    palette_key = normalize_palette_name(palette_name)
+    palettes = MODEL_PALETTES if color_by == "Model" else HARNESS_PALETTES
+    return palettes[palette_key]
+
+
 def color_map_for(
     values: pd.Series,
     color_by: ColorBy,
     palette_name: str | None = DEFAULT_PALETTE,
 ) -> dict[str, str]:
+    unique_values = [str(value) for value in sorted(values.dropna().unique())]
+    palette_key = normalize_palette_name(palette_name)
+
+    # For the default Citrus palette, preserve hand-picked colors for known labels.
+    # Unknown labels still get sequential fallback colors to avoid hash collisions.
+    if palette_key == "Citrus":
+        named_colors = MODEL_COLORS if color_by == "Model" else HARNESS_COLORS
+        fallback_colors = palette_colors_for(color_by, palette_key)
+        color_map: dict[str, str] = {}
+        fallback_index = 0
+        for value in unique_values:
+            if value in named_colors:
+                color_map[value] = named_colors[value]
+            else:
+                color_map[value] = fallback_colors[fallback_index % len(fallback_colors)]
+                fallback_index += 1
+        return color_map
+
+    # Non-default palettes are assigned sequentially rather than by hash. Hashing can
+    # map multiple visible categories to the same color, which made the high-contrast
+    # harness palette look like only gray/blue/purple buckets.
+    palette = palette_colors_for(color_by, palette_key)
     return {
-        str(value): get_color(str(value), color_by, palette_name)
-        for value in sorted(values.dropna().unique())
+        value: palette[index % len(palette)]
+        for index, value in enumerate(unique_values)
     }