log scale

.gitignore

@@ -176,8 +176,4 @@ cython_debug/
 evaluation_results/
 images/
 hf_cache/
-*.lock
-
-# macOS
-.DS_Store
-
+*.lock

README.md

@@ -30,6 +30,8 @@ pip install "gradio==5.19.0" pandas -r requirements.txt
 python app.py
 ```
 
+The app is served at `http://127.0.0.1:7860`.
+
 `requirements.txt` lists Plotly. Gradio and pandas are required locally;
 Hugging Face Spaces installs Gradio from the YAML `sdk_version` above.
 

app.py

@@ -62,9 +62,6 @@ custom_css = """
     --pruna-accordion-bg: rgba(255, 255, 255, 0.02);
     --pruna-accordion-border: rgba(216, 180, 254, 0.15);
     --pruna-dropdown-hover: #2a1844;
-    --pruna-toggle-track: var(--pruna-bg-header);
-    --pruna-toggle-thumb: var(--pruna-bg-elevated);
-    --pruna-toggle-thumb-shadow: 0 1px 2px rgba(0, 0, 0, 0.45), inset 0 1px rgba(255, 255, 255, 0.06);
     color-scheme: dark;
 }
 
@@ -108,27 +105,13 @@ custom_css = """
     --pruna-accordion-bg: var(--pruna-bg-card);
     --pruna-accordion-border: var(--pruna-border);
     --pruna-dropdown-hover: #f3e8ff;
-    --pruna-toggle-track: var(--pruna-bg-header);
-    --pruna-toggle-thumb: var(--pruna-bg-card);
-    --pruna-toggle-thumb-shadow: 0 1px 2px rgba(88, 28, 135, 0.12);
     color-scheme: light;
 }
 
-html {
+html, body {
     width: 100% !important;
     max-width: 100% !important;
-    min-width: 0 !important;
-    overflow-x: hidden;
-    overflow-y: auto;
-    -webkit-text-size-adjust: 100%;
-    text-size-adjust: 100%;
-    -webkit-tap-highlight-color: transparent;
-}
-body, gradio-app {
-    width: 100% !important;
-    max-width: 100% !important;
-    min-width: 0 !important;
-    overflow: visible;
+    overflow-x: clip;
     -webkit-tap-highlight-color: transparent;
 }
 html, body, .gradio-container, .main {
@@ -151,25 +134,18 @@ button, a, label, input, select, textarea,
 /* Subtle depth — not a marketing-site hero glow */
 body, .gradio-container {
     background-image: var(--pruna-glow) !important;
-    background-repeat: no-repeat !important;
-    background-attachment: scroll !important;
-}
-@media (min-width: 701px) and (hover: hover) and (pointer: fine) {
-    body, .gradio-container {
-        background-attachment: fixed !important;
-    }
+    background-attachment: fixed !important;
 }
 
 .gradio-container {
     width: 100% !important;
     max-width: 1200px !important;
-    min-width: 0 !important;
     margin: 0 auto !important;
     padding-top: 0 !important;
     padding-left: 20px !important;
     padding-right: 20px !important;
     box-sizing: border-box !important;
-    overflow-x: hidden;
+    overflow-x: clip;
 }
 .gradio-container .main,
 .gradio-container .wrap,
@@ -190,7 +166,6 @@ body, .gradio-container {
 .workspace-filters,
 .view-filters {
     max-width: 100% !important;
-    min-width: 0 !important;
 }
 
 /* —— App header (P-Bench only) —— */
@@ -325,9 +300,6 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
     background: transparent !important;
     box-shadow: none !important;
 }
-/* Flatten tabs so the bar sits above shared filters. display:contents is
-   the fallback; Safari can drop or mis-order those children, so browsers
-   with subgrid use the grid layout below instead. */
 .workspace-shell > .tabs,
 .workspace-shell > .main-tabs,
 .workspace-shell > .block:not(.workspace-filters),
@@ -351,49 +323,12 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
     margin: 0 0 16px !important;
     justify-content: center !important;
     width: 100% !important;
-    overflow: visible !important;
-}
-@supports (grid-template-rows: subgrid) {
-    .workspace-shell,
-    .workspace-shell.block,
-    .workspace-shell.column,
-    .workspace-shell.gap {
-        display: grid !important;
-        grid-template-columns: minmax(0, 1fr) !important;
-        grid-template-rows: auto auto auto !important;
-        align-content: start !important;
-    }
-    .workspace-shell > .tabs,
-    .workspace-shell > .main-tabs,
-    .workspace-shell .tabs.main-tabs {
-        display: grid !important;
-        grid-template-columns: minmax(0, 1fr) !important;
-        grid-template-rows: subgrid !important;
-        grid-column: 1 !important;
-        grid-row: 1 / 4 !important;
-        position: static !important;
-    }
-    .main-tabs > .tab-wrapper {
-        grid-row: 1 !important;
-        order: 0 !important;
-    }
-    .workspace-filters {
-        grid-column: 1 !important;
-        grid-row: 2 !important;
-        order: 0 !important;
-    }
-    .main-tabs .tabitem {
-        grid-row: 3 !important;
-        order: 0 !important;
-        min-width: 0 !important;
-    }
 }
 .main-tabs .tab-container {
     height: auto !important;
-    min-height: 0 !important;
     justify-content: center !important;
     flex-wrap: wrap !important;
-    overflow: visible !important;
+    overflow: hidden !important;
     max-width: 100% !important;
     gap: 2px;
 }
@@ -510,8 +445,6 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
     .app-header-brand h1,
     .gradio-container .app-header-brand h1 {
         font-size: 1.55rem !important;
-        width: auto !important;
-        max-width: 100% !important;
     }
     .app-header-tagline {
         font-size: 0.88rem !important;
@@ -581,43 +514,24 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
     .prose .ranking-table td {
         padding: 8px 10px !important;
     }
-    .ranking-table,
-    .prose .ranking-table {
-        --rank-col-width: 3.25rem;
-    }
     .ranking-table .rank,
     .prose .ranking-table .rank,
     .ranking-table th.rank {
         position: sticky !important;
         left: 0 !important;
-        width: var(--rank-col-width);
-        min-width: var(--rank-col-width);
-        max-width: var(--rank-col-width);
-        box-shadow: none;
+        width: 2.4rem;
+        min-width: 2.4rem;
     }
     .ranking-table .model-cell,
-    .prose .ranking-table .model-cell {
-        position: static !important;
-        left: auto !important;
-        z-index: auto;
-        min-width: 140px;
-        max-width: none;
-        background: transparent !important;
-        box-shadow: none !important;
-    }
-    .ranking-table th.model-cell,
-    .prose .ranking-table th.model-cell {
+    .prose .ranking-table .model-cell,
+    .ranking-table th.model-cell {
         position: sticky !important;
-        top: 0 !important;
-        left: auto !important;
-        z-index: 3;
-        min-width: 140px;
-        max-width: none;
-        background: var(--pruna-bg-header) !important;
-        box-shadow: 0 1px 0 var(--pruna-hairline) !important;
+        left: 2.4rem !important;
+        min-width: 108px;
+        max-width: 36vw;
     }
-    .ranking-table tbody tr:hover .model-cell {
-        background: var(--pruna-table-hover) !important;
+    .ranking-table .model-cell strong {
+        white-space: nowrap;
     }
     .compare-controls,
     .compare-controls.row,
@@ -690,7 +604,7 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
 .view-filters {
     display: flex !important;
     flex-wrap: wrap !important;
-    align-items: flex-end !important;
+    align-items: end !important;
     gap: 12px !important;
     margin: 0;
     overflow: visible !important;
@@ -698,7 +612,7 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
 .view-filters > div,
 .view-filters > .block,
 .view-filters > .form {
-    flex: 1 1 0% !important;
+    flex: 1 1 0 !important;
     min-width: 0 !important;
 }
 .view-filters > .block,
@@ -850,10 +764,6 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
     line-height: 1.45 !important;
     font-weight: 400 !important;
 }
-.view-help + .view-help,
-.prose .view-help + .view-help {
-    margin-top: 0.45rem !important;
-}
 .view-filters span[data-testid="block-info"],
 .view-filters .info,
 .view-filters .block-info {
@@ -991,7 +901,7 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
 .leaderboard-controls {
     display: flex !important;
     flex-wrap: wrap !important;
-    align-items: flex-end !important;
+    align-items: end !important;
     gap: 10px !important;
     margin-bottom: 12px;
     overflow: visible !important;
@@ -1114,11 +1024,11 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
     max-height: min(70vh, 720px);
     overflow-x: auto;
     overflow-y: auto;
-    overscroll-behavior: none;
+    -webkit-overflow-scrolling: touch;
+    overscroll-behavior-x: contain;
 }
 .ranking-table,
 .prose .ranking-table {
-    --rank-col-width: 4.25rem;
     width: 100%;
     margin: 0 !important;
     overflow: visible;
@@ -1194,14 +1104,10 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
 .prose .ranking-table .rank {
     position: sticky;
     left: 0;
-    z-index: 2;
+    z-index: 1;
     box-sizing: border-box;
-    width: var(--rank-col-width);
-    min-width: var(--rank-col-width);
-    max-width: var(--rank-col-width);
-    padding-left: 0.5rem !important;
-    padding-right: 0.5rem !important;
-    overflow: hidden;
+    width: 3.25rem;
+    min-width: 3.25rem;
     color: var(--pruna-lavender) !important;
     font-weight: 700 !important;
     text-align: center;
@@ -1218,21 +1124,18 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
 .ranking-table .model-cell,
 .prose .ranking-table .model-cell {
     position: sticky;
-    left: var(--rank-col-width);
-    z-index: 2;
-    box-sizing: border-box;
+    left: 3.25rem;
+    z-index: 1;
     min-width: 180px;
     max-width: 260px;
     background: var(--pruna-table-sticky) !important;
-    box-shadow: 8px 0 10px -8px rgba(0, 0, 0, 0.35) !important;
 }
 .ranking-table th.model-cell,
 .prose .ranking-table th.model-cell {
     top: 0;
-    left: var(--rank-col-width);
+    left: 3.25rem;
     z-index: 5;
     background: var(--pruna-bg-header) !important;
-    box-shadow: 0 1px 0 var(--pruna-hairline), 8px 0 10px -8px rgba(0, 0, 0, 0.35) !important;
 }
 .ranking-table tbody tr:hover .rank,
 .ranking-table tbody tr:hover .model-cell {
@@ -1414,9 +1317,7 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
     box-shadow: none !important;
 }
 .compare-controls .compare-prompt-count .head {
-    display: block !important;
-    grid-column: 1 / -1;
-    grid-row: 1;
+    display: contents;
     margin: 0 !important;
 }
 .compare-controls .compare-prompt-count .head label {
@@ -1590,182 +1491,7 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
     margin: 0 !important;
 }
 .compare-row { display: grid; gap: 12px; min-width: 0; width: 100%; }
-.compare-prompt-text { overflow-wrap: anywhere; word-break: break-word; }
-.pareto-heading-row,
-.pareto-heading-row.row,
-.pareto-heading-row .form {
-    display: flex !important;
-    flex-wrap: wrap !important;
-    align-items: center !important;
-    gap: 8px 12px !important;
-    width: 100% !important;
-    margin-bottom: 0.4rem !important;
-}
-.pareto-heading-row .pareto-subhead,
-.pareto-heading-row > div:first-child,
-.pareto-heading-row .form > div:first-child {
-    flex: 1 1 240px !important;
-    min-width: 0 !important;
-    margin: 0 !important;
-}
-.pareto-heading-row .pareto-scale-control {
-    display: flex !important;
-    flex-direction: row !important;
-    align-items: center !important;
-    justify-content: flex-end !important;
-    flex: 0 0 auto !important;
-    gap: 0 !important;
-    margin-left: auto !important;
-    max-width: 168px !important;
-    padding: 0 !important;
-}
-.pareto-scale-all-row,
-.pareto-scale-all-row.row,
-.pareto-scale-all-row .form {
-    display: flex !important;
-    flex-direction: row !important;
-    flex-wrap: wrap !important;
-    align-items: center !important;
-    justify-content: flex-start !important;
-    gap: 8px 12px !important;
-    width: 100% !important;
-    margin: 2px 0 14px !important;
-}
-.pareto-scale-all-row .html-container,
-.pareto-scale-all-row .block {
-    border: none !important;
-    background: transparent !important;
-    box-shadow: none !important;
-    padding: 0 !important;
-    margin: 0 !important;
-    width: auto !important;
-    flex: 0 0 auto !important;
-}
-.pareto-scale-all-row .html-container {
-    flex: 1 1 auto !important;
-    min-width: 0 !important;
-}
-.pareto-scale-all-label {
-    color: var(--pruna-text-muted);
-    font-size: 0.8rem;
-    font-weight: 500;
-    white-space: nowrap;
-}
-.pareto-scale-all-row .pareto-scale-toggle {
-    margin-left: auto !important;
-}
-.pareto-scale-toggle,
-.pareto-scale-toggle.block {
-    min-width: 0 !important;
-    width: auto !important;
-    border: none !important;
-    background: transparent !important;
-    box-shadow: none !important;
-    padding: 0 !important;
-    margin: 0 !important;
-}
-.pareto-scale-toggle .wrap,
-.pareto-scale-toggle .form {
-    display: block !important;
-    width: auto !important;
-    margin: 0 !important;
-    padding: 0 !important;
-    border: none !important;
-    background: transparent !important;
-    box-shadow: none !important;
-}
-.pareto-scale-toggle legend {
-    display: none !important;
-}
-.pareto-scale-toggle fieldset,
-.pareto-scale-toggle .wrap:has(> label),
-.pareto-scale-toggle .form:has(> label) {
-    position: relative !important;
-    display: grid !important;
-    grid-template-columns: 1fr 1fr !important;
-    align-items: stretch !important;
-    isolation: isolate;
-    box-sizing: border-box !important;
-    width: max-content !important;
-    min-width: 0 !important;
-    padding: 4px !important;
-    gap: 4px !important;
-    border: 1px solid var(--pruna-input-border) !important;
-    border-radius: 10px !important;
-    background: var(--pruna-toggle-track) !important;
-    box-shadow: none !important;
-}
-.pareto-scale-toggle fieldset::before,
-.pareto-scale-toggle .wrap:has(> label)::before,
-.pareto-scale-toggle .form:has(> label)::before {
-    content: none !important;
-}
-.pareto-scale-toggle label {
-    position: relative !important;
-    z-index: 1 !important;
-    display: flex !important;
-    flex: 1 1 auto !important;
-    align-items: center !important;
-    justify-content: center !important;
-    gap: 0 !important;
-    box-sizing: border-box !important;
-    min-width: 58px !important;
-    min-height: 26px !important;
-    margin: 0 !important;
-    padding: 5px 12px !important;
-    border: none !important;
-    border-radius: 6px !important;
-    background: transparent !important;
-    box-shadow: none !important;
-    color: var(--pruna-text-body) !important;
-    font-size: 0.72rem !important;
-    font-weight: 600 !important;
-    line-height: 1.2 !important;
-    letter-spacing: 0.01em;
-    white-space: nowrap;
-    cursor: pointer !important;
-}
-.pareto-scale-toggle-all label {
-    min-width: 68px !important;
-    min-height: 30px !important;
-    padding: 6px 14px !important;
-    font-size: 0.85rem !important;
-}
-.pareto-scale-toggle label span {
-    margin: 0 !important;
-    padding: 0 !important;
-    color: inherit !important;
-    opacity: 1 !important;
-}
-.pareto-scale-toggle label > * + * {
-    margin-left: 0 !important;
-}
-.pareto-scale-toggle label + label::before,
-.pareto-scale-toggle label + label {
-    content: none !important;
-    border-left: none !important;
-}
-.pareto-scale-toggle input[type="radio"] {
-    position: absolute !important;
-    appearance: none !important;
-    opacity: 0 !important;
-    width: 0 !important;
-    height: 0 !important;
-    margin: 0 !important;
-    pointer-events: none !important;
-}
-.pareto-scale-toggle label:hover {
-    background: transparent !important;
-    color: var(--pruna-text-primary) !important;
-}
-.pareto-scale-toggle label.selected,
-.pareto-scale-toggle label:has(input:checked) {
-    background: var(--pruna-toggle-thumb) !important;
-    color: var(--pruna-lavender) !important;
-    font-weight: 700 !important;
-    border-color: transparent !important;
-    box-shadow: var(--pruna-toggle-thumb-shadow) !important;
-}
+.compare-prompt-text { overflow-wrap: anywhere; }
 .pareto-layout,
 .pareto-layout.row,
 .pareto-layout .form {
@@ -1783,9 +1509,6 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
     max-width: 100% !important;
 }
 .compare-cell { min-width: 0; }
-.compare-cell.compare-source .compare-model-label {
-    color: var(--pruna-text-muted);
-}
 .compare-model-label {
     margin-bottom: 6px;
     color: var(--pruna-lavender);
@@ -1793,23 +1516,15 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
     font-weight: 700;
     word-break: break-word;
 }
-.compare-cell img,
-.compare-cell video {
+.compare-cell img {
     display: block;
     width: 100%;
+    aspect-ratio: 1 / 1;
+    object-fit: cover;
     border-radius: 12px;
     border: 1px solid var(--pruna-border);
     background: var(--pruna-bg-elevated);
 }
-.compare-cell img {
-    aspect-ratio: 1 / 1;
-    object-fit: cover;
-}
-.compare-cell video {
-    aspect-ratio: 16 / 9;
-    max-height: 360px;
-    object-fit: contain;
-}
 .compare-empty,
 .pareto-note-copy {
     margin: 0;
@@ -1834,7 +1549,7 @@ button.theme-toggle[data-mode="light"] .theme-icon-moon { display: block !import
 .app-header .app-header-brand h1 {
     display: block !important;
     width: max-content !important;
-    max-width: 100% !important;
+    max-width: none !important;
     flex: 0 0 auto !important;
     margin: 0 !important;
     padding: 0 !important;
@@ -2168,20 +1883,14 @@ def load_sample_comparison_data(folder):
         return None
 
     prompts = {}
-    source_videos = {}
     with prompts_path.open() as handle:
         for line in handle:
             if not line.strip():
                 continue
             row = json.loads(line)
-            prompt_id = row["prompt_id"]
-            prompts[prompt_id] = row.get("text", "")
-            source_video = row.get("source_video")
-            if source_video:
-                source_videos[prompt_id] = source_video
+            prompts[row["prompt_id"]] = row.get("text", "")
 
     images = defaultdict(dict)
-    kinds = set()
     with generations_path.open() as handle:
         for line in handle:
             if not line.strip():
@@ -2189,33 +1898,20 @@ def load_sample_comparison_data(folder):
             row = json.loads(line)
             model_id = row["model_id"]
             prompt_id = row["prompt_id"]
-            media_url = row.get("image") or row.get("video")
-            if row.get("video"):
-                kinds.add("video")
-            elif row.get("image"):
-                kinds.add("image")
-            if model_id and prompt_id and media_url:
-                images[model_id][prompt_id] = media_url
-            if prompt_id and prompt_id not in source_videos:
-                params = row.get("params") or {}
-                input_video = params.get("input_video") or row.get("input_video")
-                if input_video:
-                    source_videos[prompt_id] = input_video
+            image_url = row.get("image")
+            if model_id and prompt_id and image_url:
+                images[model_id][prompt_id] = image_url
 
     models = sorted(images)
     if not models or not prompts:
         return None
 
-    loaded = {
+    return {
         "prompts": prompts,
         "images": {model: dict(prompt_map) for model, prompt_map in images.items()},
         "models": models,
         "prompt_ids": sorted(prompts),
-        "kind": "video" if "video" in kinds else "image",
     }
-    if source_videos:
-        loaded["source_videos"] = source_videos
-    return loaded
 
 
 def _as_numeric(df, columns):
@@ -2330,7 +2026,7 @@ def load_qwen_combined_dataframe(path):
     df = df[~df["Model"].astype(str).str.startswith("#")].copy()
     df["Model"] = df["Model"].astype(str).str.strip()
 
-    df = _as_numeric(
+    return _as_numeric(
         df,
         [
             "Price / Image (USD)",
@@ -2340,126 +2036,7 @@ def load_qwen_combined_dataframe(path):
             "Rapidata Elo",
             "Datapoint Elo",
         ],
-    )
-    df = df.drop(columns=["Raw Win Rate"], errors="ignore")
-    return df.reset_index(drop=True)
-
-
-def _is_pruna_video_model(series):
-    models = series.astype(str).str.casefold()
-    return models.str.startswith("p_video") | models.str.startswith("p-video")
-
-
-def _apply_video_timings(df, *, replace_displayed_time=False):
-    """Mix Fal wall time with Pruna model execution time."""
-    fal = df.get("Time / Output Video Second (s)")
-    execution = df.get("Execution Time / Output Video Second (s)")
-    if fal is None:
-        return df
-
-    if "model_id" in df.columns:
-        is_ours = _is_pruna_video_model(df["model_id"])
-    else:
-        is_ours = _is_pruna_video_model(df["Model"])
-
-    if execution is None:
-        mixed = fal
-    else:
-        ours_time = execution.where(execution.notna(), fal)
-        mixed = fal.where(~is_ours, ours_time)
-    if replace_displayed_time:
-        df["Time / Output Video Second (s)"] = mixed
-    df["Pareto Time / Output Video Second (s)"] = mixed
-    return df
-
-
-def load_video_editing_dataframe(path):
-    """Load the video-to-video editing leaderboard."""
-    df = pd.read_csv(path, na_values=["N/A", "n/a", ""])
-    df = df.rename(
-        columns={
-            "display_name": "Model",
-            "elo": "Datapoint Elo",
-            "min_generation_s": "Min Generation Time (s)",
-            "median_generation_s": "Median Generation Time (s)",
-            "p20_generation_s": "P20 Generation Time (s)",
-            "generation_s_per_output_video_s": "Time / Output Video Second (s)",
-            "predict_time_s_per_output_video_s": "Predict Time / Output Video Second (s)",
-            "model_execution_time_s_per_output_video_s": (
-                "Execution Time / Output Video Second (s)"
-            ),
-            "price": "Price / Second of Video (USD)",
-        }
-    )
-    df = df.drop(columns=["wandb_run_ids", "n_generations"], errors="ignore")
-    df["Model"] = df["Model"].astype(str).str.strip()
-    df = _as_numeric(
-        df,
-        [
-            "Datapoint Elo",
-            "Min Generation Time (s)",
-            "Median Generation Time (s)",
-            "P20 Generation Time (s)",
-            "Time / Output Video Second (s)",
-            "Predict Time / Output Video Second (s)",
-            "Execution Time / Output Video Second (s)",
-            "Price / Second of Video (USD)",
-        ],
-    )
-    df = _apply_video_timings(df)
-    df = df.drop(columns=["model_id"], errors="ignore")
-    return df.reset_index(drop=True)
-
-
-def load_text_to_video_dataframe(path):
-    """Load the text-to-video leaderboard (P-Video-2 and Fal models)."""
-    df = pd.read_csv(path, na_values=["N/A", "n/a", ""])
-    model_column = "model_id" if "model_id" in df.columns else "Model"
-    df = df[~df[model_column].astype(str).str.casefold().str.startswith("agnes")].copy()
-    df = df.rename(
-        columns={
-            "model_id": "Model",
-            "datapoint_elo": "Datapoint Elo",
-            "rapidata_elo": "Rapidata Elo",
-            "min_generation_s": "Min Generation Time (s)",
-            "median_generation_s": "Median Generation Time (s)",
-            "p20_generation_s": "P20 Generation Time (s)",
-            "generation_s_per_output_video_s": "Time / Output Video Second (s)",
-            "model_execution_s_per_output_video_s": (
-                "Execution Time / Output Video Second (s)"
-            ),
-            "price_usd_per_second": "Price / Second of Video (USD)",
-        }
-    )
-    df = df.drop(columns=["wandb_run_ids", "n_generations"], errors="ignore")
-    df["Model"] = df["Model"].astype(str).str.strip()
-    df = _as_numeric(
-        df,
-        [
-            "Datapoint Elo",
-            "Rapidata Elo",
-            "Min Generation Time (s)",
-            "Median Generation Time (s)",
-            "P20 Generation Time (s)",
-            "Time / Output Video Second (s)",
-            "Execution Time / Output Video Second (s)",
-            "Price / Second of Video (USD)",
-        ],
-    )
-    # Pruna rows use model execution time; Fal rows keep Fal wall time.
-    df = _apply_video_timings(df, replace_displayed_time=True)
-    df = df.drop(
-        columns=["Execution Time / Output Video Second (s)"],
-        errors="ignore",
-    )
-    elo_columns = [
-        column
-        for column in ("Datapoint Elo", "Rapidata Elo")
-        if column in df.columns
-    ]
-    if elo_columns:
-        df = df.dropna(subset=elo_columns, how="all")
-    return df.reset_index(drop=True)
+    ).reset_index(drop=True)
 
 
 df = load_oneig_dataframe(oneig_path)
@@ -2509,17 +2086,9 @@ qwen_combined_dir = _resolve_data_path(
     data_dir / "qwen_image_bench_combined",
     space_root.parent / "qwen_image_bench_combined",
 )
-video_combined_dir = _resolve_data_path(
-    data_dir / "video_editing_combined",
-    space_root.parent / "video_editing_combined",
-)
-text_to_video_combined_dir = _resolve_data_path(
-    data_dir / "video_generation_combined",
-    space_root.parent / "video_generation_combined",
-)
 qwen_path = _resolve_data_path(
-    data_dir / "qwen_image_bench_model_price_and_median_generation_time_10_august.csv",
-    space_root.parent / "qwen_image_bench_model_price_and_median_generation_time_10_august.csv",
+    data_dir / "qwen_image_bench_model_price_and_median_generation_time.csv",
+    space_root.parent / "qwen_image_bench_model_price_and_median_generation_time.csv",
 )
 aa_path = _resolve_data_path(
     data_dir / "artificial_analysis_text_to_image_leaderboard.csv",
@@ -2529,20 +2098,10 @@ arena_path = _resolve_data_path(
     data_dir / "arena_ai_text_to_image_leaderboard.csv",
     space_root.parent / "arena_ai_text_to_image_leaderboard.csv",
 )
-video_path = _resolve_data_path(
-    data_dir / "video-editing-leaderboard.csv",
-    space_root.parent / "video-editing-leaderboard.csv",
-)
-text_to_video_path = _resolve_data_path(
-    data_dir / "p-video-2-leaderboard.csv",
-    space_root.parent / "p-video-2-leaderboard.csv",
-)
 
 qwen_df = load_qwen_combined_dataframe(qwen_path)
 aa_df = load_artificial_analysis_dataframe(aa_path)
 arena_df = load_arena_ai_dataframe(arena_path)
-video_df = load_video_editing_dataframe(video_path)
-text_to_video_df = load_text_to_video_dataframe(text_to_video_path)
 qwen_display_columns = [
     col
     for col in [
@@ -2550,6 +2109,7 @@ qwen_display_columns = [
         "Datapoint Elo",
         "Rapidata Elo",
         "P-Judge Overall",
+        "Raw Win Rate",
         "Median Generation Time (s)",
         "Min Generation Time (s)",
         "Price / Image (USD)",
@@ -2574,36 +2134,9 @@ arena_display_columns = [
     ]
     if col in arena_df.columns
 ]
-video_display_columns = [
-    col
-    for col in [
-        "Model",
-        "Datapoint Elo",
-        "Time / Output Video Second (s)",
-        "Median Generation Time (s)",
-        "Min Generation Time (s)",
-        "Price / Second of Video (USD)",
-    ]
-    if col in video_df.columns
-]
-text_to_video_display_columns = [
-    col
-    for col in [
-        "Model",
-        "Datapoint Elo",
-        "Rapidata Elo",
-        "Time / Output Video Second (s)",
-        "Median Generation Time (s)",
-        "Min Generation Time (s)",
-        "Price / Second of Video (USD)",
-    ]
-    if col in text_to_video_df.columns
-]
 
 oneig_samples = load_sample_comparison_data(oneig_combined_dir)
 qwen_samples = load_sample_comparison_data(qwen_combined_dir)
-video_samples = load_sample_comparison_data(video_combined_dir)
-text_to_video_samples = load_sample_comparison_data(text_to_video_combined_dir)
 
 metrics = [
     {"id": "datapoint_elo", "column": "Datapoint Elo"},
@@ -2662,45 +2195,11 @@ arena_metric_ids = _metric_ids_for(
         "arena_text",
     ],
 )
-video_metric_ids = _metric_ids_for(video_df, ["datapoint_elo"])
-text_to_video_metric_ids = _metric_ids_for(
-    text_to_video_df, ["datapoint_elo", "rapidata_elo"]
-)
 
 datasets = [
-    {
-        "id": "text_to_video",
-        "name": "VBench-2.0 Dataset",
-        "modality": "text_to_video",
-        "data": text_to_video_df,
-        "columns": text_to_video_display_columns,
-        "metric_ids": text_to_video_metric_ids,
-        "note": (
-            "Datapoint Elo and Rapidata Elo from pairwise text-to-video "
-            "preference. Price is USD per second of output video. Time per "
-            "second of video is Fal wall time, except Pruna models which use "
-            "model execution time."
-        ),
-        "samples": text_to_video_samples,
-    },
-    {
-        "id": "video_editing",
-        "name": "Pruna Internal Video-Edit Benchmark",
-        "modality": "video_to_video",
-        "data": video_df,
-        "columns": video_display_columns,
-        "metric_ids": video_metric_ids,
-        "note": (
-            "Datapoint Elo from pairwise video-edit preference. Price is USD "
-            "per second of output video. Generation time per second of video "
-            "is end-to-end wall time to produce one second of output."
-        ),
-        "samples": video_samples,
-    },
     {
         "id": "qwen",
         "name": "Qwen Image Dataset",
-        "modality": "text_to_image",
         "data": qwen_df,
         "columns": qwen_display_columns,
         "metric_ids": qwen_metric_ids,
@@ -2710,7 +2209,6 @@ datasets = [
     {
         "id": "oneig",
         "name": "OneIG Alignment Dataset",
-        "modality": "text_to_image",
         "data": oneig_df,
         "columns": oneig_display_columns,
         "metric_ids": oneig_metric_ids,
@@ -2723,7 +2221,6 @@ datasets = [
     {
         "id": "artificial_analysis",
         "name": "Artificial Analysis Dataset",
-        "modality": "text_to_image",
         "data": aa_df,
         "columns": aa_display_columns,
         "metric_ids": aa_metric_ids,
@@ -2733,7 +2230,6 @@ datasets = [
     {
         "id": "arena_ai",
         "name": "Arena AI Dataset",
-        "modality": "text_to_image",
         "data": arena_df,
         "columns": arena_display_columns,
         "metric_ids": arena_metric_ids,
@@ -2744,14 +2240,8 @@ datasets = [
 datasets = [dataset for dataset in datasets if dataset["metric_ids"]]
 
 DEFAULT_DATASET_ID = next(
-    (dataset["id"] for dataset in datasets if dataset["id"] == "text_to_video"),
-    next(
-        (dataset["id"] for dataset in datasets if dataset["id"] == "video_editing"),
-        next(
-            (dataset["id"] for dataset in datasets if dataset["id"] == "qwen"),
-            datasets[0]["id"] if datasets else None,
-        ),
-    ),
+    (dataset["id"] for dataset in datasets if dataset["id"] == "qwen"),
+    datasets[0]["id"] if datasets else None,
 )
 DEFAULT_METRIC_ID = None
 
@@ -2827,23 +2317,11 @@ custom_head = """
     return found;
   };
 
-  const applyPlotTheme = (gd, mode) => {
-    const layout = PLOT_LAYOUT[mode];
-    if (!layout || typeof Plotly === "undefined" || !gd) return;
-    if (gd.layout && gd.layout.paper_bgcolor === layout.paper_bgcolor) return;
-    try { Plotly.relayout(gd, layout); } catch (e) {}
-  };
-
-  const watchPlotTheme = (gd) => {
-    if (!gd || gd.__inferbenchThemeBound) return;
-    gd.__inferbenchThemeBound = true;
-    gd.addEventListener("plotly_afterplot", () => applyPlotTheme(gd, currentMode()));
-  };
-
   const restylePlots = (mode) => {
+    const layout = PLOT_LAYOUT[mode];
+    if (!layout || typeof Plotly === "undefined") return;
     queryAll(".js-plotly-plot").forEach((gd) => {
-      watchPlotTheme(gd);
-      applyPlotTheme(gd, mode);
+      try { Plotly.relayout(gd, layout); } catch (e) {}
     });
   };
 

data/p-video-2-leaderboard.csv

@@ -1,19 +0,0 @@
-model_id,wandb_run_ids,n_generations,min_generation_s,median_generation_s,p20_generation_s,generation_s_per_output_video_s,model_execution_s_per_output_video_s,datapoint_elo,rapidata_elo,price_usd_per_second
-gemini_omni_1_1_flash,mfajagum,89,26.276599962002365,31.884095050001633,29.0771403791965,6.399738597206617,,1037,1197,0.100
-grok_imagine_video,q39r8v9e,90,68.9776645039965,99.17614604350092,78.69151808420138,16.13514411181308,,958,994,0.050
-grok_imagine_video_v1_5,z2fp7hd7,90,38.42492011199647,52.268091876499966,46.47850653079659,11.452002471104521,,959,1016,0.140
-ltx_2_5_fast,h6se1hfi,90,21.695961473000352,26.684429597007693,24.118977216200438,5.270079915707363,,1002,1035,0.090
-ltx_2_5_pro,kjbgxbab,90,23.997383733993047,27.897931821993552,25.263995286799037,4.936325968288616,,1007,978,0.120
-minimax_h3,xpcv2m58,89,97.61419671699696,117.51345164199665,106.57570995900024,24.968780374361877,,1026,1108,0.100
-minimax_h3_max,0d2lkmp6,90,4.173863318999793,4.49869444649994,4.4671880990012145,1.3807388451000264,0.644,1031,1197,0.080
-minimax_h3_max_turbo__prompt_expansion_mode_balanced,kxuwq31c,90,2.891819470001792,3.487746603501364,3.134695611400821,1.35231252479333,0.61,1040,1214,0.040
-p_video_2__draft_false__prompt_upsampling_false__resolution_1080p,9hgkhujw,90,13.33159556199098,16.702878594005597,14.793676785795833,4.033411242884629,2.151682222222222,,,0.050
-p_video_2__draft_false__prompt_upsampling_false__resolution_720p,m2jf4z7h,89,5.859760620005545,8.162274362999597,7.147269867203431,2.1525601170473116,0.9053123595505614,,,0.025
-p_video_2__draft_false__prompt_upsampling_true__resolution_1080p,uoqpv10b,90,14.615217412007041,18.316011337505188,16.441080821005745,4.5638632221758275,2.152566666666667,979,991,0.050
-p_video_2__draft_false__prompt_upsampling_true__resolution_720p,2q5jydr8,90,7.227749306999613,9.880705762494472,9.154919161600992,2.4789854674801206,0.9079755555555558,985,1039,0.025
-p_video_2__draft_true__prompt_upsampling_false__resolution_1080p,qbt0atz4,89,6.250065040003392,9.528199804000906,7.504525098402519,2.431133616615736,0.7517640449438197,,,0.030
-p_video_2__draft_true__prompt_upsampling_false__resolution_720p,9ur57h54,88,3.379928643000312,5.454214365498046,4.208131522199255,1.4960965825225272,0.40954545454545455,,,0.015
-p_video_2__draft_true__prompt_upsampling_true__resolution_1080p,gcbyy4wn,90,8.047415203996934,11.128662197996164,9.22702455239487,2.8557857865532665,0.7514066666666668,,,0.030
-p_video_2__draft_true__prompt_upsampling_true__resolution_720p,6hi994vk,90,4.985804016003385,7.4922584965024726,6.472871531004785,2.4024102996157146,0.40678444444444434,982,1048,0.015
-seedance_2_5_turbo,l9zjnv7n,90,168.997338743,252.07832621250054,211.93920676639829,67.60899374696221,,1037,1080,0.200
-veo_3_1_lite,jjbobes4,90,33.894167409991496,37.83729844300251,37.009501291197374,6.745116482181308,,968,986,0.050

data/qwen_image_bench_model_price_and_median_generation_time.csv

@@ -0,0 +1,60 @@
+Model,Price / Image (USD),Median Generation Time (s),Min Generation Time (s),P-Judge Overall,Raw Win Rate,Rapidata Elo,Datapoint Elo,Benchmark.ai Elo
+reve_2_1,N/A,N/A,28.3,,,,,1173.2
+ideogram_4_0_quality,N/A,N/A,66.6,,,,,1131.0
+gpt_image_2,0.21,77.8,77.8,59.02083099999998,,1172.58,1116,1124.5
+nano_banana_2_0,N/A,N/A,N/A,56.362956,59.4%,1071.79,1067,1056.3
+gpt_image_1_5,0.135,38.0,38.0,57.864434999999986,64.5%,1102.09,1064,910.3
+hidream_i1_dev,0.0086,2.823569217998738,2.06,49.17682099999999,51.7%,999.17,984,
+gpt_image_1,0.167,38.8,38.8,54.975741000000006,,1095.39,987,
+imagen_4_0_ultra,0.06,11.4,11.4,53.385603999999965,59.7%,1075.22,1023,
+flux_2_flex,0.06,10.856533817990567,8.07,53.9245676767677,57.8%,1054.34,1018,921.3
+qwen_image,0.025,4.8,4.8,51.561746,51.5%,1073.72,1005,975.6
+seedream_5_0,N/A,N/A,N/A,55.715153,,1070.35,1012,
+nano_banana_pro,0.134,17.2,17.2,56.452189898989914,58.0%,1029.51,1045,1102.4
+hidream_i1_fast,0.0051,9.920469530501578,1.40,48.89298600000002,50.1%,1024.67,984,
+imagen_4_fast,0.02,3.7531301500021073,2.71,50.155055208333344,,972.6,981,
+seedream_4_5,0.04,16.6,16.6,55.66328800000001,,1048.96,1035,964.2
+seedream_4_0,0.03,12.2,12.2,55.241443999999994,,1050.41,1035,
+p_image_2_ideogram_low_1k,0.0075,2.59,1.55,54.827397,53.2%,1000.07,1009,1103.9
+#p_image_2_ideogram_low_2k,0.016,5.17,4.25,53.971723,48.4%,1074.91,1003,
+qwen_image_2_0_pro,0.035,35.5,35.5,56.181776,,1033.25,1017,959.9
+juggernaut_base_flux,0.035,4.115834823496698,3.84,49.474676,46.4%,1038.22,972,
+flux_2_pro,N/A,N/A,N/A,54.43423900000002,,993.54,1019,1011.6
+z_image,0.005,1.5122045120006078,1.22,49.946227999999984,48.1%,1028.1,1001,
+p_image_2_ideogram_high_1k,0.015,4.28,3.07,55.754507999999994,52.1%,972.92,1022,1104.0
+#p_image_2_ideogram_high_2k,0.03,8.69,7.12,54.757842000000004,49.4%,1074.91,1007,
+qwen_image_2512,0.02,19.1,19.1,51.677326,,1029.6,1009,
+flux_2_max,0.07,26.4,26.4,54.047976,60.9%,955.19,1028,1001.6
+flux_1_1_pro_ultra,0.06,9.026992494000297,6.20,50.358445,52.0%,979.56,995,
+juggernaut_pro_flux,0.055,3.699636150500737,3.36,49.741183,46.4%,972.14,964,
+flux_dev,0.025,1.6931055715031107,1.49,48.241479,42.1%,925.04,940,
+p_image_2_ideogram_very_low_1k,0.003,2.56,1.49,53.51894200000001,48.1%,959.86,995,1071.7
+#p_image_2_ideogram_very_low_2k,0.006,3.94,3.05,53.68248699999998,47.2%,962.38,980,
+#p_image_2_ideogram_very_low_1k_no_upsampling,0.005,0.82,,46.33,,,,1071.7
+#p_image_2_ideogram_very_low_2k_no_upsampling,0.005,2.18,,46.39,,,,1071.7
+#p_image_2_ideogram_low_1k_no_upsampling,0.01,3.33,,45.32,,,,1103.9
+#p_image_2_ideogram_low_2k_no_upsampling,0.01,3.34,,46.25,,,,1103.9
+#p_image_2_ideogram_medium_1k_no_upsampling,0.015,2.13,,46.88,,,,1115.1
+#p_image_2_ideogram_medium_2k_no_upsampling,0.015,5.55,,47.11,,,,1115.1
+#p_image_2_ideogram_high_1k_no_upsampling,0.03,3.11,,46.88,,,,1104.0
+#p_image_2_ideogram_high_2k_no_upsampling,0.03,5.55,,47.06,,,,1104.0
+hidream_i1_full,0.014,6.008430051002506,5.69,46.82559300000002,36.9%,955.46,944,
+flux_2_dev,0.025,4.310278721997747,4.03,52.71644489795918,53.1%,1007.6,1021,942.0
+imagen_4_0,0.04,14.1,14.1,52.08996199999999,53.5%,979.62,1005,
+wan_2_2_image,0.02,3.005390542501118,2.96,48.19959399999999,,944.87,960,
+flux_krea,0.025,1.7150160090022837,1.7150160090022837,50.35734,50.0%,919.73,975,
+p_image_2_ideogram_medium_1k,0.01,3.06,2.05,54.719193000000004,51.7%,941.46,1002,1115.1
+#p_image_2_ideogram_medium_2k,0.02,7.44,6.47,54.23124444444446,48.1%,949.35,1000,
+glm_image,0.05,188.2,188.2,51.42623399999999,,923.35,953,
+p_image,0.005,1.0640762715011078,0.95,48.75217099999999,44.8%,924.37,961,1098.7
+hunyuanimage_3_0,0.09,41.0,41.0,52.32440099999998,52.4%,1009.61,979,765.3
+juggernaut_lightning_flux,0.006,1.1787893719956628,0.93,48.30471699999998,40.5%,916.68,929,
+flux_1_1_pro,0.04,3.0104645500032348,2.34,49.92882700000001,50.6%,925.04,984,
+flux_schnell,0.003,0.8411653029907029,0.80,46.685981818181816,34.8%,892.32,915,
+kling_v2_1,N/A,N/A,N/A,51.044512,,870.04,981,
+#p_image_2_ideogram_very_high_high_1k,0.075,9.34,7.25,58.45,,,1025,
+#p_image_2_ideogram_very_high_low_1k,0.0375,26.17,10.44,57.59,,,1024,
+#p_image_2_ideogram_very_high_medium_1k,0.05,9.28,5.96,57.88,,,1020,
+#p_image_2_ideogram_very_high_very_low_1k,0.015,26.52,10.60,56.92,,,1002,
+#p_image_2_ideogram_final_1k,0.0375,11.17,5.55,58.28,,,,
+#p_image_2_ideogram_final_2k,0.075,14.34,9.96,57.68,,,,

data/qwen_image_bench_model_price_and_median_generation_time_10_august.csv

@@ -1,64 +0,0 @@
-Rapidata Model,Price / Image (USD),Median Generation Time (s),Min Generation Time (s),P-Judge Overall,Raw Win Rate,Rapidata Elo,Datapoint Elo,Benchmark.ai Elo
-reve_2_1,N/A,N/A,28.3,,,,,1173.2
-ideogram_4_0_quality,N/A,N/A,66.6,,,,,1131.0
-gpt_image_2,0.21,77.8,77.8,59.02083099999998,65.6%,1172.58,1110,1124.5
-nano_banana_2_0,N/A,N/A,N/A,56.362956,59.2%,1071.79,1063,1056.3
-gpt_image_1_5,0.135,38.0,38.0,57.864434999999986,58.9%,1102.09,1060,910.3
-hidream_i1_dev,0.0086,2.823569217998738,2.06,49.17682099999999,47.7%,999.17,983,
-gpt_image_1,0.167,38.8,38.8,54.975741000000006,47.4%,1095.39,984,
-imagen_4_0_ultra,0.06,11.4,11.4,53.385603999999965,52.9%,1075.22,1019,
-flux_2_flex,0.06,10.856533817990567,8.07,53.9245676767677,52.4%,1054.34,1014,921.3
-#flux_2_turbo,0.008,2.14,1.81,53.02,,,1003,
-#flux_2_flash,0.005,1.43,1.03,52.03,,,1001,
-qwen_image,0.025,4.8,4.8,51.561746,50.3%,1073.72,1002,975.6
-seedream_5_0,N/A,N/A,N/A,55.715153,51.6%,1070.35,1010,
-nano_banana_pro,0.134,17.2,17.2,56.452189898989914,56.4%,1029.51,1041,1102.4
-hidream_i1_fast,0.0051,9.920469530501578,1.40,48.89298600000002,47.2%,1024.67,980,
-imagen_4_fast,0.02,3.7531301500021073,2.71,50.155055208333344,47.2%,972.6,980,
-seedream_4_5,0.04,16.6,16.6,55.66328800000001,54.7%,1048.96,1032,964.2
-seedream_4_0,0.03,12.2,12.2,55.241443999999994,54.9%,1050.41,1032,
-p_image_2_ideogram_low_1k,0.0075,2.59,1.55,54.827397,50.8%,1000.07,1006,1103.9
-#p_image_2_ideogram_low_2k,0.016,5.17,4.25,53.971723,50.0%,1074.91,1000,1103.9
-qwen_image_2_0_pro,0.035,35.5,35.5,56.181776,52.2%,1033.25,1014,959.9
-juggernaut_base_flux,0.035,4.115834823496698,3.84,49.474676,46.1%,1038.22,971,
-flux_2_pro,N/A,N/A,N/A,54.43423900000002,52.2%,993.54,1014,1011.6
-z_image,0.005,1.5122045120006078,1.22,49.946227999999984,49.9%,1028.1,999,
-p_image_2_ideogram_high_1k,0.015,4.28,3.07,55.754507999999994,52.5%,972.92,1017,1104.0
-#p_image_2_ideogram_high_2k,0.06,8.69,7.12,54.757842000000004,50.7%,1074.91,1006,1104.0
-qwen_image_2512,0.02,19.1,19.1,51.677326,50.9%,1029.6,1005,
-flux_2_max,0.07,26.4,26.4,54.047976,53.9%,955.19,1026,1001.6
-flux_1_1_pro_ultra,0.06,9.026992494000297,6.20,50.358445,48.6%,979.56,990,
-juggernaut_pro_flux,0.055,3.699636150500737,3.36,49.741183,44.7%,972.14,963,
-flux_dev,0.025,1.6931055715031107,1.49,48.241479,41.5%,925.04,941,
-p_image_2_ideogram_very_low_1k,0.003,2.56,1.49,53.51894200000001,49.1%,959.86,994,1071.7
-#p_image_2_ideogram_very_low_2k,0.006,3.94,3.05,53.68248699999998,46.3%,962.38,976,1071.7
-#p_image_2_ideogram_very_low_1k_no_upsampling,0.005,0.82,,46.33,,,,1071.7
-#p_image_2_ideogram_very_low_2k_no_upsampling,0.005,2.18,,46.39,,,,1071.7
-#p_image_2_ideogram_low_1k_no_upsampling,0.01,3.33,,45.32,,,,1103.9
-#p_image_2_ideogram_low_2k_no_upsampling,0.01,3.34,,46.25,,,,1103.9
-#p_image_2_ideogram_medium_1k_no_upsampling,0.015,2.13,,46.88,,,,1115.1
-#p_image_2_ideogram_medium_2k_no_upsampling,0.015,5.55,,47.11,,,,1115.1
-#p_image_2_ideogram_high_1k_no_upsampling,0.03,3.11,,46.88,,,,1104.0
-#p_image_2_ideogram_high_2k_no_upsampling,0.03,5.55,,47.06,,,,1104.0
-hidream_i1_full,0.014,6.008430051002506,5.69,46.82559300000002,41.4%,955.46,941,
-flux_2_dev,0.025,4.310278721997747,4.03,52.71644489795918,52.5%,1007.6,1016,942.0
-imagen_4_0,0.04,14.1,14.1,52.08996199999999,50.4%,979.62,1002,
-wan_2_2_image,0.02,3.005390542501118,2.96,48.19959399999999,44.1%,944.87,958,
-flux_krea,0.025,1.7150160090022837,1.7150160090022837,50.35734,46.2%,919.73,973,
-p_image_2_ideogram_medium_1k,0.01,3.06,2.05,54.719193000000004,49.8%,941.46,999,1115.1
-#p_image_2_ideogram_medium_2k,0.02,7.44,6.47,54.23124444444446,49.3%,949.35,996,1115.1
-glm_image,0.05,188.2,188.2,51.42623399999999,42.9%,923.35,950,
-p_image,0.005,1.0640762715011078,0.95,48.75217099999999,44.3%,924.37,960,1098.7
-hunyuanimage_3_0,0.09,41.0,41.0,52.32440099999998,46.6%,1009.61,977,765.3
-juggernaut_lightning_flux,0.006,1.1787893719956628,0.93,48.30471699999998,40.0%,916.68,930,
-flux_1_1_pro,0.04,3.0104645500032348,2.34,49.92882700000001,47.7%,925.04,984,
-flux_schnell,0.003,0.8411653029907029,0.80,46.685981818181816,37.6%,892.32,914,
-kling_v2_1,N/A,N/A,N/A,51.044512,47.2%,870.04,980,
-#p_image_2_ideogram_very_high_high_1k,0.075,9.34,7.25,58.45,52.9%,,1024,
-#p_image_2_ideogram_very_high_low_1k,0.0375,26.17,10.44,57.59,52.9%,,1023,
-#p_image_2_ideogram_very_high_medium_1k,0.05,9.28,5.96,57.88,52.3%,,1019,
-#p_image_2_ideogram_very_high_very_low_1k,0.015,26.52,10.60,56.92,49.4%,,1001,
-p_image_2_ideogram_very_high_1k,0.033,11.17,5.55,58.28,55.5%,,1034,
-#p_image_2_ideogram_very_high_2k,0.066,14.34,9.96,57.68,53.4%,,1020,
-#p_image_2_ideogram_very_high_john_1k,,8.53,6.43,58.19,,,1015,
-#p_image_2_ideogram_very_high_john_2k,,13.86,10.64,57.75,,,1002,

data/video-editing-leaderboard.csv

@@ -1,12 +0,0 @@
-model_id,display_name,elo,wandb_run_ids,n_generations,min_generation_s,median_generation_s,p20_generation_s,generation_s_per_output_video_s,predict_time_s_per_output_video_s,model_execution_time_s_per_output_video_s,price
-gemini_omni_flash_edit__fal,Gemini Omni Flash Edit,1054,3iuctwj9,56,32.574746752001374,58.2327566820004,50.44402900500063,13.78,,,0.13
-grok_imagine_video__replicate,Grok Imagine Video,974,q0ogvekx,59,31.640928319000523,43.75649321700257,32.88782280539963,11.04,,,0.05
-happyhorse_1_0__wavespeed,HappyHorse 1.0,1048,k4zj5wqk,71,106.5824136010051,201.2431439649954,141.46595527700265,46.73,,,0.14
-ltx_2_3_quality_reference_video_to_video__fal,LTX 2.3 Video Edit,911,p19ks8ad,73,60.768441981999786,73.62109239100027,70.20137865180223,15.83,,,0.054
-lucy_edit_pro__fal,Lucy Edit Pro,879,na6bz9xx,72,118.3459930579993,136.44442766549764,124.13885582720104,27.18,,,0.15
-minimax_h3_reference_to_video__fal,MiniMax H3 Reference-to-Video,1060,639xmakz,63,180.8385945170012,308.29198316100155,248.2394383729996,58.31,,,0.06
-p_video_edit_preview__replicate_final,P-Video-Edit,1000,m8irpsa6,73,31.41135125700021,86.86665409700072,57.21095684959946,23.18,17.99,12.06,0.045
-p_video_edit_preview__replicate_final__draft,P-Video-Edit Draft,994,6f88e6mx,73,23.04809326099712,42.9797806409988,33.19193872759861,11.48,11.28,4.46,0.025
-seedance_2_5_video_edit_turbo__wavespeed,Seedance 2.5 Video Edit Turbo,1063,2nr274vp,68,142.49452170499717,293.7401115540015,223.11569286320045,59.93,,,0.24
-wan_2_7_video_edit__wavespeed,Wan 2.7 Video Edit,1055,9lq803be,66,134.3744560209998,313.957433804002,219.2491260079987,68.04,,,0.2
-

data/video_editing_combined/generations.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:3ec4ea0a431484e65f30990c8cc4292b00c665be96db587775426da1c84a4ab8
-size 762227

data/video_editing_combined/prompts.jsonl

@@ -1,78 +0,0 @@
-{"prompt_id": "video_edit_internal__advertising__p0000", "text": "Replace the white bottle with an orange sunscreen bottle.", "dataset": "video_edit_internal", "category": "advertising", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/advertising/4620329ad65820ce.mp4"}
-{"prompt_id": "video_edit_internal__advertising__p0001", "text": "Replace the woman with an asian woman riding on a donkey through a chinese small town.", "dataset": "video_edit_internal", "category": "advertising", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/advertising/44db5689846be901.mp4"}
-{"prompt_id": "video_edit_internal__advertising__p0002", "text": "Place the couple on a snowy mountain next to a mountain hut.", "dataset": "video_edit_internal", "category": "advertising", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/advertising/481db4a047688bbe.mp4"}
-{"prompt_id": "video_edit_internal__advertising__p0003", "text": "Turn this into a scene in the summer with birds flying in the sky and butterflies in the foreground.", "dataset": "video_edit_internal", "category": "advertising", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/advertising/32b9f56763c56482.mp4"}
-{"prompt_id": "video_edit_internal__advertising__p0004", "text": "Replace the robot arm with a dancing hamster on a small table.", "dataset": "video_edit_internal", "category": "advertising", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/advertising/22d02ae2df7f9045.mp4"}
-{"prompt_id": "video_edit_internal__anonymization__p0000", "text": "Anonymize the womans face", "dataset": "video_edit_internal", "category": "anonymization", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/anonymization/7c410085f2fd939d.mp4"}
-{"prompt_id": "video_edit_internal__anonymization__p0001", "text": "Anonymize the mans face", "dataset": "video_edit_internal", "category": "anonymization", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/anonymization/7b495805c0c270ea.mp4"}
-{"prompt_id": "video_edit_internal__anonymization__p0002", "text": "Anonymize the mans face", "dataset": "video_edit_internal", "category": "anonymization", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/anonymization/736b6fb504bd32be.mp4"}
-{"prompt_id": "video_edit_internal__anonymization__p0003", "text": "Anonymize the girls face", "dataset": "video_edit_internal", "category": "anonymization", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/anonymization/2c82413d39579cd5.mp4"}
-{"prompt_id": "video_edit_internal__artificial_analysis__p0000", "text": "Change the rainforest setting to a neon-lit urban alley at night with steam from vents and wet reflective asphalt, and move into an ariel shot of the character after the initial camera movement to focus on the character", "dataset": "video_edit_internal", "category": "artificial_analysis", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/artificial_analysis/e2f6121269a206d3.mp4"}
-{"prompt_id": "video_edit_internal__artificial_analysis__p0001", "text": "Replace the magenta hover-car with a chrome-blue car, keeping the drift and neon light trails.", "dataset": "video_edit_internal", "category": "artificial_analysis", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/artificial_analysis/55d1baa7f3f71b26.mp4"}
-{"prompt_id": "video_edit_internal__artificial_analysis__p0002", "text": "Make the footage look like it was shot on a 1970s film camera, with grainy film texture, faded warm colors, slight softness, and slightly darker corners around the edges.", "dataset": "video_edit_internal", "category": "artificial_analysis", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/artificial_analysis/fd1ab436de020152.mp4"}
-{"prompt_id": "video_edit_internal__artificial_analysis__p0003", "text": "A person appears at the top of the waterfall, leaps into the lake below, and disappears into the misty water beneath the falls.", "dataset": "video_edit_internal", "category": "artificial_analysis", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/artificial_analysis/7ee1623303a8da68.mp4"}
-{"prompt_id": "video_edit_internal__artificial_analysis__p0004", "text": "A fluffy orange cat leaps gracefully from the floor onto the couch, paws sinking into the soft cushions. It circles once in place, tail swaying gently, before curling up tightly on one of the cushions and settling in comfortably.", "dataset": "video_edit_internal", "category": "artificial_analysis", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/artificial_analysis/9a73c608b343bbbe.mp4"}
-{"prompt_id": "video_edit_internal__camera_editing__p0000", "text": "Zoom in on the man's face to show his focused expression", "dataset": "video_edit_internal", "category": "camera_editing", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/camera_editing/0336e64b0594bd7a.mp4"}
-{"prompt_id": "video_edit_internal__camera_editing__p0001", "text": "Perform an arc shot around the tram as it arrives at the station", "dataset": "video_edit_internal", "category": "camera_editing", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/camera_editing/204aa93703da117b.mp4"}
-{"prompt_id": "video_edit_internal__camera_editing__p0002", "text": "Change the view to a high angle.", "dataset": "video_edit_internal", "category": "camera_editing", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/camera_editing/c2ccb5351be5c718.mp4"}
-{"prompt_id": "video_edit_internal__camera_editing__p0003", "text": "Change the view to a high angle.", "dataset": "video_edit_internal", "category": "camera_editing", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/camera_editing/b725aaded02d65e9.mp4"}
-{"prompt_id": "video_edit_internal__camera_editing__p0004", "text": "Gradually move the camera away from the doctor", "dataset": "video_edit_internal", "category": "camera_editing", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/camera_editing/33f24f4d083df743.mp4"}
-{"prompt_id": "video_edit_internal__design_arena__p0000", "text": "Create a transition of the video of the product whihc is the jeans on the lady model with an atitiude that goes into a zoom out camera that she is in a party .Focus on the vibes", "dataset": "video_edit_internal", "category": "design_arena", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/design_arena/b0b4642272a481b1.mp4"}
-{"prompt_id": "video_edit_internal__design_arena__p0001", "text": "add ethnic urban women dancing at the bottem and at the bar wearing DMI Tshirts", "dataset": "video_edit_internal", "category": "design_arena", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/design_arena/0f730f98f3c51356.mp4"}
-{"prompt_id": "video_edit_internal__design_arena__p0002", "text": "Generate a commercial of this dog drinking beer at an electronic music party in a world where dogs and humans are mixed together.", "dataset": "video_edit_internal", "category": "design_arena", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/design_arena/575fc4b9b6d7b3ab.mp4"}
-{"prompt_id": "video_edit_internal__design_arena__p0003", "text": "everything is the same excpet the winter season, snowing", "dataset": "video_edit_internal", "category": "design_arena", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/design_arena/8674fd022867a736.mp4"}
-{"prompt_id": "video_edit_internal__design_arena__p0004", "text": "A cinematic character introduction of Jason, framed in a medium close-up with subtle camera movement, confident body language, and expressive facial detail. Moody, high-contrast lighting with a cool-toned color palette, shallow depth of field, and a slow dramatic reveal that builds intrigue over a few seconds. I want a video in a 9:16 aspect ratio for tiktok. remove all text and have him in a mid day campus setting", "dataset": "video_edit_internal", "category": "design_arena", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/design_arena/91873bbafe4173da.mp4"}
-{"prompt_id": "video_edit_internal__e_commerce__p0000", "text": "Remove the black blazer.", "dataset": "video_edit_internal", "category": "e_commerce", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/e_commerce/b89faaa5080c31db.mp4"}
-{"prompt_id": "video_edit_internal__e_commerce__p0001", "text": "Replace her skirt with a jeans skirt.", "dataset": "video_edit_internal", "category": "e_commerce", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/e_commerce/c34a33d07e843804.mp4"}
-{"prompt_id": "video_edit_internal__e_commerce__p0002", "text": "Replace the black leggings and the black shirt with a white dress.", "dataset": "video_edit_internal", "category": "e_commerce", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/e_commerce/489d2439cbaeed79.mp4"}
-{"prompt_id": "video_edit_internal__e_commerce__p0003", "text": "Replace the white woman with a lebanese-looking woman.", "dataset": "video_edit_internal", "category": "e_commerce", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/e_commerce/0ad8fb773f303a11.mp4"}
-{"prompt_id": "video_edit_internal__e_commerce__p0004", "text": "Remove all items on the table and place an eyeshadow pallete on the table next to the woman.", "dataset": "video_edit_internal", "category": "e_commerce", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/e_commerce/f28f3aa12b1ec220.mp4"}
-{"prompt_id": "video_edit_internal__lighting__p0000", "text": "After the sun sets behind the mountains, the scene transitions into nighttime.", "dataset": "video_edit_internal", "category": "lighting", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/lighting/229c165e0ac97daf.mp4"}
-{"prompt_id": "video_edit_internal__lighting__p0001", "text": "Change the weather to a dazzling starry night.", "dataset": "video_edit_internal", "category": "lighting", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/lighting/f42c41ab2f3adc26.mp4"}
-{"prompt_id": "video_edit_internal__lighting__p0002", "text": "Change the weather to a thunderstorm with heavy rain.", "dataset": "video_edit_internal", "category": "lighting", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/lighting/daa36567634ca6be.mp4"}
-{"prompt_id": "video_edit_internal__lighting__p0003", "text": "Change the weather to a torrential downpour.", "dataset": "video_edit_internal", "category": "lighting", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/lighting/6817a4142b69d602.mp4"}
-{"prompt_id": "video_edit_internal__lighting__p0004", "text": "Change the weather to a torrential downpour.", "dataset": "video_edit_internal", "category": "lighting", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/lighting/ac01488a685a1096.mp4"}
-{"prompt_id": "video_edit_internal__long__p0000", "text": "Change the weather to a dazzling starry night.", "dataset": "video_edit_internal", "category": "long", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/long/e9af2a7039e43f16.mp4"}
-{"prompt_id": "video_edit_internal__long__p0001", "text": "Adjust the color of book to blue", "dataset": "video_edit_internal", "category": "long", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/long/eab696d1555be316.mp4"}
-{"prompt_id": "video_edit_internal__long__p0002", "text": "Make the young woman turn into sand and blow away.", "dataset": "video_edit_internal", "category": "long", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/long/d18378f1d7e4ca0e.mp4"}
-{"prompt_id": "video_edit_internal__long__p0003", "text": "Make the bird flap its wings", "dataset": "video_edit_internal", "category": "long", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/long/3835b37b43f5e5bf.mp4"}
-{"prompt_id": "video_edit_internal__long__p0004", "text": "Transform the video into a ukiyo-e style", "dataset": "video_edit_internal", "category": "long", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/long/09fe5bb4b5304b66.mp4"}
-{"prompt_id": "video_edit_internal__movie_concept_art__p0000", "text": "Add more blood and wounds to the mans face.", "dataset": "video_edit_internal", "category": "movie_concept_art", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/movie_concept_art/1d1a99167475720b.mp4"}
-{"prompt_id": "video_edit_internal__movie_concept_art__p0001", "text": "Make the scene less dark.", "dataset": "video_edit_internal", "category": "movie_concept_art", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/movie_concept_art/10cb44d8c0bf4116.mp4"}
-{"prompt_id": "video_edit_internal__movie_concept_art__p0002", "text": "Replace the female warrior with a male warrior with ginger hair.", "dataset": "video_edit_internal", "category": "movie_concept_art", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/movie_concept_art/9e45454390c20b8a.mp4"}
-{"prompt_id": "video_edit_internal__movie_concept_art__p0003", "text": "Turn the man's hand into a robotic hand.", "dataset": "video_edit_internal", "category": "movie_concept_art", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/movie_concept_art/99977e7a91a950b4.mp4"}
-{"prompt_id": "video_edit_internal__real_estate__p0000", "text": "Replace the interior with a gold-black interior heavy luxury style.", "dataset": "video_edit_internal", "category": "real_estate", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/real_estate/869064ab06bb5daa.mp4"}
-{"prompt_id": "video_edit_internal__real_estate__p0001", "text": "Replace the grey wallpaper with a beige painted wall and turn the grey curtains a dark brown.", "dataset": "video_edit_internal", "category": "real_estate", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/real_estate/1c2f358bfe9b8299.mp4"}
-{"prompt_id": "video_edit_internal__real_estate__p0002", "text": "Remove all decoration from the walls and keep only the furniture.", "dataset": "video_edit_internal", "category": "real_estate", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/real_estate/2058bcf38389f8fe.mp4"}
-{"prompt_id": "video_edit_internal__real_estate__p0003", "text": "Exchange the wooden floor for a marble floor.", "dataset": "video_edit_internal", "category": "real_estate", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/real_estate/6a4c09a23dda57d9.mp4"}
-{"prompt_id": "video_edit_internal__real_estate__p0004", "text": "Replace the bedframe with a modern wooden bed.", "dataset": "video_edit_internal", "category": "real_estate", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/real_estate/3e0f6dc5df891001.mp4"}
-{"prompt_id": "video_edit_internal__style_transfer__p0000", "text": "Transform the video into a cyberpunk style", "dataset": "video_edit_internal", "category": "style_transfer", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/style_transfer/2253f6ed3674f006.mp4"}
-{"prompt_id": "video_edit_internal__style_transfer__p0001", "text": "Convert to different shades of orange", "dataset": "video_edit_internal", "category": "style_transfer", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/style_transfer/edb6a105be03bae0.mp4"}
-{"prompt_id": "video_edit_internal__style_transfer__p0002", "text": "Convert to black and white", "dataset": "video_edit_internal", "category": "style_transfer", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/style_transfer/c3dc058b3be26cbd.mp4"}
-{"prompt_id": "video_edit_internal__style_transfer__p0003", "text": "Apply Ghibli-style editing to the video", "dataset": "video_edit_internal", "category": "style_transfer", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/style_transfer/83a638223a20911c.mp4"}
-{"prompt_id": "video_edit_internal__style_transfer__p0004", "text": "Relight the scene as if it were shot during golden hour, with warm low-angle sunlight, soft shadows, and natural highlights on faces and surfaces.", "dataset": "video_edit_internal", "category": "style_transfer", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/style_transfer/daa36567634ca6be.mp4"}
-{"prompt_id": "video_edit_internal__subject_editing__p0000", "text": "Replace the rainbow colors of the logo with different shades of purple", "dataset": "video_edit_internal", "category": "subject_editing", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/subject_editing/6358f26b46baadf5.mp4"}
-{"prompt_id": "video_edit_internal__subject_editing__p0001", "text": "Add a small dog running beside the scooter", "dataset": "video_edit_internal", "category": "subject_editing", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/subject_editing/052c0bfd7b68fea7.mp4"}
-{"prompt_id": "video_edit_internal__subject_editing__p0002", "text": "Replace the splashing waves with a calm water surface", "dataset": "video_edit_internal", "category": "subject_editing", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/subject_editing/c3f79578a39f415d.mp4"}
-{"prompt_id": "video_edit_internal__subject_editing__p0003", "text": "Add a violinist in the background", "dataset": "video_edit_internal", "category": "subject_editing", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/subject_editing/a0eebf32a076e2fc.mp4"}
-{"prompt_id": "video_edit_internal__subject_editing__p0004", "text": "Add a group of people walking on the pathway", "dataset": "video_edit_internal", "category": "subject_editing", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/subject_editing/3c2b7e3a5b284e1d.mp4"}
-{"prompt_id": "video_edit_internal__subject_motion_editing__p0000", "text": "Make the bird hop around instead of walking and foraging", "dataset": "video_edit_internal", "category": "subject_motion_editing", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/subject_motion_editing/ac38b2a97d1a5c3a.mp4"}
-{"prompt_id": "video_edit_internal__subject_motion_editing__p0001", "text": "The male colleague is walking around to observe.", "dataset": "video_edit_internal", "category": "subject_motion_editing", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/subject_motion_editing/8aaab7d304c5ad99.mp4"}
-{"prompt_id": "video_edit_internal__subject_motion_editing__p0002", "text": "Make the static spider-man in the mural dynamic and make him swing faster", "dataset": "video_edit_internal", "category": "subject_motion_editing", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/subject_motion_editing/3412b39686fd5879.mp4"}
-{"prompt_id": "video_edit_internal__subject_motion_editing__p0003", "text": "Change the woman's jogging to taking off.", "dataset": "video_edit_internal", "category": "subject_motion_editing", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/subject_motion_editing/17c4dc0514772321.mp4"}
-{"prompt_id": "video_edit_internal__subject_motion_editing__p0004", "text": "Make the knight lunging forward and the creature swiping at the knight", "dataset": "video_edit_internal", "category": "subject_motion_editing", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/subject_motion_editing/980354550011911b.mp4"}
-{"prompt_id": "video_edit_internal__synthetic_data__p0000", "text": "Turn the scene into nighttime.", "dataset": "video_edit_internal", "category": "synthetic_data", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/synthetic_data/e3b1e603bfaa1aec.mp4"}
-{"prompt_id": "video_edit_internal__synthetic_data__p0001", "text": "Remove the crosswalk.", "dataset": "video_edit_internal", "category": "synthetic_data", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/synthetic_data/30df5fd5fe28af65.mp4"}
-{"prompt_id": "video_edit_internal__synthetic_data__p0002", "text": "Add a bicycle riding in front of the car.", "dataset": "video_edit_internal", "category": "synthetic_data", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/synthetic_data/d093491e25682b3a.mp4"}
-{"prompt_id": "video_edit_internal__synthetic_data__p0003", "text": "Turn the scene into a snowstorm scene.", "dataset": "video_edit_internal", "category": "synthetic_data", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/synthetic_data/1da5913f13362f8d.mp4"}
-{"prompt_id": "video_edit_internal__synthetic_data__p0004", "text": "Make it rain in the scene.", "dataset": "video_edit_internal", "category": "synthetic_data", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/synthetic_data/936bdb89558f5a7f.mp4"}
-{"prompt_id": "video_edit_internal__text__p0000", "text": "Add the text \"True Love\" in the foreground in pink romantic font.", "dataset": "video_edit_internal", "category": "text", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/text/c763407793ca9b0e.mp4"}
-{"prompt_id": "video_edit_internal__text__p0001", "text": "Replace any mention of \"Fanta\" with the branding \"Cola\".", "dataset": "video_edit_internal", "category": "text", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/text/7b878ea6d99834b2.mp4"}
-{"prompt_id": "video_edit_internal__text__p0002", "text": "Remove all text from the video.", "dataset": "video_edit_internal", "category": "text", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/text/76763c31b2609109.mp4"}
-{"prompt_id": "video_edit_internal__text__p0003", "text": "Replace the branding \"Royalty\" with the phrasing \"Princess\".", "dataset": "video_edit_internal", "category": "text", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/text/d221d299c9e49064.mp4"}
-{"prompt_id": "video_edit_internal__text__p0004", "text": "Remove all text from the video.", "dataset": "video_edit_internal", "category": "text", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/text/5c169901af8e4ed6.mp4"}
-{"prompt_id": "video_edit_internal__transitions__p0000", "text": "After a black screen transition, the road transforms into lush grass.", "dataset": "video_edit_internal", "category": "transitions", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/transitions/964c35b19ee1fc1f.mp4"}
-{"prompt_id": "video_edit_internal__transitions__p0001", "text": "After a wave-foam transition, the small fishing boat is eaten by a giant whale.", "dataset": "video_edit_internal", "category": "transitions", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/transitions/b8b88f43316363f3.mp4"}
-{"prompt_id": "video_edit_internal__transitions__p0002", "text": "Add a cut transition, then show a bowl of ramen topped with parsley.", "dataset": "video_edit_internal", "category": "transitions", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/transitions/c823d6c77bdc87f4.mp4"}
-{"prompt_id": "video_edit_internal__transitions__p0003", "text": "Add a smoke transition, then show the circuit board burning.", "dataset": "video_edit_internal", "category": "transitions", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/transitions/edb9927ff06c1175.mp4"}
-{"prompt_id": "video_edit_internal__transitions__p0004", "text": "After a smoke transition, the Basilica of the Sacred Heart of Paris catches fire.", "dataset": "video_edit_internal", "category": "transitions", "source_video": "https://d2j1a65dna040x.cloudfront.net/benchmark_inputs/video_edit_internal/transitions/77bb900383d4f370.mp4"}

data/video_generation_combined/generations.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:8395b068f818174bbbbd0af450c6cedcc76142352f35ccf565f55c4837cb43b4
-size 1135969

data/video_generation_combined/prompts.jsonl

@@ -1,90 +0,0 @@
-{"prompt_id": "v-bench2__all__p0000", "text": "Garden, zoom in.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0001", "text": "Garden, zoom out.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0002", "text": "Garden, tilt up.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0003", "text": "The camera starts at the top of the forest, where thick morning mist drifts between the treetops, with sunlight filtering through the leaves and casting dappled spots of light, the air fresh and moist. As the camera slowly moves downward, dewdrops glisten on the leaves, a gentle breeze rustling the leaves, making a soft, soothing sound. The camera moves along a woodland path, occasionally capturing a squirrel leaping between trees, its agile figure darting across the branches. The shot continues, eventually arriving at a tranquil lake. The surface of the water is like a mirror, reflecting the towering trees and distant mountains. In the distance, a heron takes off from the lakeshore, breaking the stillness of the water, ripples forming as the bird skims the surface. The camera follows its flight and gradually pulls back, showing the entire serene beauty of the lake.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0004", "text": "The scene opens in the vast desert, with the camera angled low over the sand, fine grains drifting in the gentle breeze, the undulating dunes in the distance bathed in a faint orange glow. As the camera moves forward, the outlines of the dunes become clearer, and the early morning sky begins to glow with warm hues, transitioning the desert from deep brown to golden yellow. The scene cuts to the top of a dune, where the camera follows a caravan of camels making their way across the endless desert, their bells jingling softly in the wind. The first rays of the sun break through the gaps between the dunes, casting golden light on the sand, as the desert comes to life, glowing in the morning's embrace. Finally, the camera pulls back, revealing the vast expanse of the desert merging with the horizon, the sun climbing higher, illuminating the entire landscape in a mystical, warm glow.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0005", "text": "The camera begins in a vast grassland, where the lush green grass sways gently in the breeze, the air fresh, and the soft rustling of the leaves fills the space. As the camera moves forward, the grass turns from a vibrant green to golden hues, sunlight pouring over the land, the textures of the grass becoming more pronounced in the play of light and shadow. In the distance, a herd of cattle grazes peacefully. Their figures become clearer as the camera moves closer, occasionally lifting their heads and gazing into the distance. The scene shifts to a lakeshore, where the lake reflects the sky, grasslands, and distant mountains, the water clear and calm, with gentle ripples created by the breeze. Finally, the camera pulls back, following the herd as they move off into the distance, the vastness of the grassland merging with the expansive sky, creating a feeling of peaceful openness.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0006", "text": "The camera starts at a tranquil coastline, where the waves gently crash against the rocks, the salty sea breeze fills the air, and the atmosphere feels fresh and alive. As the camera moves forward, the surface of the water glistens with golden light from the setting sun, the sound of the waves becoming more distinct as the tide retreats, revealing a moist sandy shore. The scene shifts to the beach, where a few seagulls peck at the sand, occasionally flying up and breaking the stillness of the sky. The sun slowly sinks below the horizon, painting the sky in vibrant shades of orange and red, as the golden light on the water fades, and the waves begin to intensify. Finally, the camera pulls back, and the entire coastline fades into twilight, with the sea and sky blending together in a serene, solitary embrace.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0007", "text": "The race began, and the first runner quickly took off, leading the other teams. Everyone was focused on his performance. During the baton handoff of the first runner, Team A made a slight mistake, and the baton nearly fell. Despite this, the runner quickly passed it to the second runner. The second runner showed great determination and chased hard, finally overtaking Team B on the bend. However, due to the earlier mistake, Team A's lead was not significant. During the third runner's handoff, the runner from Team A nervously accelerated and managed to maintain the lead, but Team C quickly caught up thanks to their excellent pace control. During the fourth runner's handoff, the final sprint became crucial. The Team A runner started to accelerate in the second-to-last turn and, with a burst of strength, successfully widened the gap, steadily running toward the finish line and winning the race.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0008", "text": "The match began, and Team A quickly organized an attack, breaking through Team B's defense with fast passes. The forward accurately kicked the ball into the goal, taking a 1-0 lead. Team B did not panic; they used a long pass to quickly counterattack. The forward calmly shot past the goalkeeper, and the ball went straight into the net, making it 1-1. At this point, the game entered a stalemate. Both teams' defenses were solid, and the match became a fierce battle. In the second half, Team A had a corner kick. The high ball delivered by the player was headed in by the center-back who jumped high, putting Team A in the lead at 2-1. In the final moments of the game, Team B got a penalty kick. The forward took the shot without hesitation, sending the ball into the net to make it 2-2. Finally, in the last moments of extra time, Team A used a quick counterattack and a long-range shot from outside the box flew straight into the corner, securing the win with a 3-2 score.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0009", "text": "The race began, and all the runners sprinted quickly. The runner from Team A quickly took the lead. After a period of steady running, the runners from Team A and Team B gradually created a gap, almost running shoulder to shoulder. However, at the 25 km mark, the runner from Team A suddenly began to feel unwell, slowing down noticeably. The runner from Team B seized the opportunity and caught up, eventually overtaking Team A to take the lead at the 30 km mark. As the race entered the second half, the Team A runner gave it their all, regained their rhythm, and caught up with Team B at the 35 km mark. The two were neck and neck in the final sprint, and with only 200 meters remaining, the Team A runner pushed through with determination and accelerated to cross the finish line first by a narrow margin, winning the tough race.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0010", "text": "The race began, and the runners quickly started. The Team A runner took the lead initially due to a powerful start. However, the Team B runner did not rush to chase but instead steadily adjusted their pace, ensuring they had more energy for the latter part of the race. By the third lap, the Team A runner began to tire, and the Team B runner gradually reduced the gap, eventually overtaking Team A in the fourth lap. The Team C runner, meanwhile, began to accelerate, and in the last two laps, with strong willpower, they overtook Team B and moved into the lead. In the final sprint, the Team A runner gritted their teeth and tried to close the gap, but with only 50 meters left, the Team C runner exploded with speed, crossing the finish line with a clear lead and winning the race.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0011", "text": "The match began, and Team A quickly found their rhythm. A series of fast attacks put pressure on Team B’s defense. Team A's forward made two three-pointers, quickly increasing the lead. After the first quarter, Team B adjusted their tactics, strengthened their defense, and gradually started to counterattack, narrowing the gap. In the third quarter, Team A’s key player was injured and forced to leave the game. Team B took advantage of this by speeding up their attacks, overtaking the score. In the fourth quarter, Team A's bench players caught up with a series of fast breaks, and in the final moments, a player from Team A calmly made a game-winning three-pointer from beyond the arc, securing a narrow 1-point victory.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0012", "text": "The match began, and Team A quickly gained the upper hand with several precise kills, taking a 11-6 lead. The Team B player stayed calm and gradually adjusted, finding ways to respond. In the second set, Team B improved their serving quality, scoring several powerful smashes to win a set, leveling the score at 1-1. In the deciding set, the Team A player showed signs of fatigue, but with precise net control and quick reflexes, they pulled ahead. In the crucial final point, a lightning-fast smash left Team B with no chance to return, and Team A won the match 21-19.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0013", "text": "A lion with the wings of an eagle, soaring through the sky with majestic ease.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0014", "text": "A giraffe with the scales of a fish, able to glide smoothly through the water.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0015", "text": "A wolf with the body of a horse, galloping across a vast plains with wild abandon.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0016", "text": "A bear with the antlers of a deer, roaming the forest with a regal presence.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0017", "text": "A cheetah with the shell of a tortoise, moving quickly but with a protective outer layer.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0018", "text": "A wooden toy is placed gently on the surface of a small bowl of water.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0019", "text": "A river changes from blue to brown.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0020", "text": "The leaves gradually change from red to green.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0021", "text": "A river changes from brown to blue.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0022", "text": "A car changes from white to red.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0023", "text": "A car changes from red to white.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0024", "text": "A dog is on the left of a table, then the dog runs to the front of the table.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0025", "text": "A dog is on the left of a sofa, then the dog runs to the front of the sofa.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0026", "text": "A dog is on the right of a table, then the dog runs to the left of the table.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0027", "text": "A dog is on the right of a sofa, then the dog runs to the front of the sofa.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0028", "text": "A dog is on the right of a rock, then the dog runs to the left of the rock.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0029", "text": "A dog is behind a chair, then the dog runs to the right of the chair.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0030", "text": "A man is doing yoga.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0031", "text": "A woman is doing yoga.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0032", "text": "people are doing yoga.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0033", "text": "A man is running.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0034", "text": "A woman is running.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0035", "text": "people are running.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0036", "text": "A man is walking.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0037", "text": "A woman is walking.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0038", "text": "people are walking.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0039", "text": "A man is dancing.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0040", "text": "A woman is dancing.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0041", "text": "A man is playing basketball.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0042", "text": "A woman is playing basketball.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0043", "text": "One person hands a cup of water to another.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0044", "text": "One person passes a ball to another.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0045", "text": "Two people shake hands.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0046", "text": "One person ties the shoelaces of another person.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0047", "text": "One person opens the door for another person.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0048", "text": "Two people exchange a book.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0049", "text": "One person puts a coat on another person.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0050", "text": "One person places a chair for another to sit in.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0051", "text": "An orange dog is running.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0052", "text": "Two orange dogs are running.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0053", "text": "A brown dog is on the left of an apple, then the dog moves to the right of the apple.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0054", "text": "An orange cat is running.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0055", "text": "Equal amounts of yellow and blue paint are rapidly combined, with the mixture being vigorously stirred until fully blended.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0056", "text": "Equal amounts of black and white paint are rapidly combined, with the mixture being vigorously stirred until fully blended.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0057", "text": "Equal amounts of white and black paint are rapidly combined, with the mixture being vigorously stirred until fully blended.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0058", "text": "Equal amounts of red and purple paint are rapidly combined, with the mixture being vigorously stirred until fully blended.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0059", "text": "A bowl of soup is tilted in the space station, with the liquid slowly spreading in all directions.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0060", "text": "A jar of peanut butter is opened in the space station, with the viscous liquid slowly dispersing.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0061", "text": "A container of cream is opened in the space station, with the liquid slowly dispersing into the air.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0062", "text": "A cup of tea is carefully tilted in the space station, and the liquid floats in various directions.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0063", "text": "A bottle of ketchup is gently squeezed in the space station, with the thick liquid spreading into the environment.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0064", "text": "A bottle of water is opened in the space station, and the water starts to float out in irregular shapes.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0065", "text": "A person is sitting on the couch, then suddenly they get up and start sweeping the floor.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0066", "text": "A dog is playing with a ball, then it suddenly starts lying down on the carpet.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0067", "text": "A person is cooking dinner, then they suddenly start organizing the pantry.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0068", "text": "A cat is watching birds through the window, then it suddenly starts grooming itself.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0069", "text": "A person is typing on a keyboard, then they suddenly get up and start making the bed.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0070", "text": "A horse is trotting in the field, then it suddenly starts drinking from a stream.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0071", "text": "A person is drinking a glass of water, then they suddenly start cleaning the windows.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0072", "text": "A dog is sitting in the yard, then it suddenly starts running in circles.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0073", "text": "A person is slurping noodles from a steaming bowl.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0074", "text": "A person is eating hamburger.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0075", "text": "A person is eating ice cream.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0076", "text": "A person is drinking coffee from a cup.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0077", "text": "A person is spreading butter on toast.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0078", "text": "A person is eating spaghetti with a fork.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0079", "text": "A person is biting into an apple.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0080", "text": "A person is peeling a banana.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0081", "text": "The camera orbits around. Castle, the camera circles around.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0082", "text": "The camera orbits around. Volcano, the camera circles around.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0083", "text": "The camera orbits around. Statue, the camera circles around.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0084", "text": "The camera orbits around. Clock Tower, the camera circles around.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0085", "text": "The camera orbits around. Playground, the camera circles around.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0086", "text": "A timelapse captures the transformation of water in an untextured bottle as the temperature significantly drops below 0°C.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0087", "text": "A timelapse captures the transformation of a river as the temperature significantly drops below 0°C.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0088", "text": "A timelapse captures the transformation of juice in an untextured bottle as the temperature significantly drops below 0°C.", "dataset": "v-bench2", "category": "all"}
-{"prompt_id": "v-bench2__all__p0089", "text": "A timelapse captures the transformation of milk in an untextured bottle as the temperature significantly drops below 0°C.", "dataset": "v-bench2", "category": "all"}

model_display.py

@@ -87,35 +87,6 @@ MODEL_DISPLAY_NAMES = {
     "p_image_2_ideogram_high_1k": "P-Image-Ideogram High 1K",
     "p_image_2_ideogram_high_2k": "P-Image-Ideogram High 2K",
     "P-Image-Ideogram (High)": "P-Image-Ideogram High",
-    "p_image_2_ideogram_very_high_1k": "P-Image-Ideogram Very High 1K",
-    # P-Video-Edit
-    "P-Video-Edit": "P-Video-Edit",
-    "P-Video-Edit Draft": "P-Video-Edit Draft",
-    "P-Video Edit Final": "P-Video-Edit",
-    "P-Video Edit Final (draft)": "P-Video-Edit Draft",
-    "p_video_edit_preview__replicate_final": "P-Video-Edit",
-    "p_video_edit_preview__replicate_final__draft": "P-Video-Edit Draft",
-    # Text-to-video leaderboard
-    "gemini_omni_1_1_flash": "Gemini Omni 1.1 Flash",
-    "grok_imagine_video": "Grok Imagine Video",
-    "grok_imagine_video_v1_5": "Grok Imagine Video 1.5",
-    "ltx_2_5_fast": "LTX 2.5 Fast",
-    "ltx_2_5_pro": "LTX 2.5 Pro",
-    "minimax_h3": "MiniMax H3",
-    "minimax_h3_max": "MiniMax H3 Max",
-    "minimax_h3_max_turbo": "MiniMax H3 Max Turbo",
-    "minimax_h3_max_turbo__prompt_expansion_mode_balanced": "MiniMax H3 Max Turbo",
-    "seedance_2_5_turbo": "Seedance 2.5 Turbo",
-    "veo_3_1_lite": "Veo 3.1 Lite",
-    # Video-to-video leaderboard
-    "gemini_omni_flash_edit__fal": "Gemini Omni Flash Edit",
-    "grok_imagine_video__replicate": "Grok Imagine Video",
-    "happyhorse_1_0__wavespeed": "HappyHorse 1.0",
-    "ltx_2_3_quality_reference_video_to_video__fal": "LTX 2.3 Video Edit",
-    "lucy_edit_pro__fal": "Lucy Edit Pro",
-    "minimax_h3_reference_to_video__fal": "MiniMax H3 Reference-to-Video",
-    "seedance_2_5_video_edit_turbo__wavespeed": "Seedance 2.5 Video Edit Turbo",
-    "wan_2_7_video_edit__wavespeed": "Wan 2.7 Video Edit",
     # Others overlapping P-Bench
     "z_image": "Z-Image",
     "glm_image": "GLM-Image",
@@ -213,35 +184,6 @@ MODEL_DISPLAY_NAMES = {
 }
 
 
-def _p_video_2_display_name(model_id: str):
-    """Turn p_video_2 variant ids into P-Video-2 Draft 720p labels."""
-    raw = str(model_id).strip()
-    if raw != "p_video_2" and not raw.startswith("p_video_2__"):
-        return None
-    if raw == "p_video_2":
-        return "P-Video-2"
-
-    draft = False
-    upsample = None
-    resolution = None
-    for part in raw.split("__")[1:]:
-        if part.startswith("draft_"):
-            draft = part.endswith("true")
-        elif part.startswith("prompt_upsampling_"):
-            upsample = part.endswith("true")
-        elif part.startswith("resolution_"):
-            resolution = part[len("resolution_") :]
-
-    label = "P-Video-2"
-    if draft:
-        label += " Draft"
-    if resolution:
-        label += f" {resolution}"
-    if upsample is False:
-        label += " (no prompt upsampling)"
-    return label
-
-
 def _prettify_snake_case(model_id: str) -> str:
     parts = [part for part in str(model_id).split("_") if part]
     pretty = []
@@ -266,10 +208,7 @@ def display_model_name(model_id) -> str:
         return ""
     if raw in MODEL_DISPLAY_NAMES:
         return MODEL_DISPLAY_NAMES[raw]
-    p_video_2 = _p_video_2_display_name(raw)
-    if p_video_2:
-        return p_video_2
     # Already a human label (spaces / punctuation) — keep as-is.
-    if re.search(r"[\s.\[\]()-]", raw):
+    if re.search(r"[\s.\[\]()]", raw):
         return raw
     return _prettify_snake_case(raw)

ui.py

@@ -1,5 +1,4 @@
 from html import escape
-from math import ceil, floor, log10
 from pathlib import Path
 import base64
 import random
@@ -25,43 +24,13 @@ MAX_COMPARE_PROMPTS = 8
 MAX_PARETO_METRICS = 8
 _PARETO_SLOT_COUNT = 1 + MAX_PARETO_METRICS * 8
 _PARETO_PRICE_COLUMN = "Price / Image (USD)"
-_PARETO_VIDEO_PRICE_COLUMN = "Price / Second of Video (USD)"
-_PARETO_PRICE_COLUMNS = (_PARETO_PRICE_COLUMN, _PARETO_VIDEO_PRICE_COLUMN)
 _PARETO_TIME_COLUMN = "Min Generation Time (s)"
-_PARETO_VIDEO_TIME_COLUMN = "Pareto Time / Output Video Second (s)"
-_PARETO_TIME_COLUMNS = (_PARETO_VIDEO_TIME_COLUMN, _PARETO_TIME_COLUMN)
-_PARETO_PRICE_TITLES = {
-    _PARETO_PRICE_COLUMN: "Price per image (USD)",
-    _PARETO_VIDEO_PRICE_COLUMN: "Price per second of video (USD)",
-}
-_PARETO_TIME_TITLES = {
-    _PARETO_TIME_COLUMN: "Min generation time (s)",
-    _PARETO_VIDEO_TIME_COLUMN: "Generation time per second of video",
-}
-_PARETO_SCALE_CHOICES = [
-    ("Log", "Logarithmic"),
-    ("Linear", "Linear"),
-]
-_PARETO_SCALE_VALUES = {value for _, value in _PARETO_SCALE_CHOICES}
-_PARETO_SCALE_DEFAULT = "Logarithmic"
-_PARETO_PRUNA_COLOR = "#c084fc"
-_PARETO_OTHER_COLOR = "#9aa3b5"
-_PARETO_FRONTIER_OUTLINE = "#3fa87e"
 
 TAB_LEADERBOARDS = "leaderboards"
 TAB_PARETO = "pareto"
 TAB_SAMPLES = "samples"
 TAB_ABOUT = "about"
 
-MODALITY_TEXT_TO_VIDEO = "text_to_video"
-MODALITY_VIDEO_TO_VIDEO = "video_to_video"
-MODALITY_TEXT_TO_IMAGE = "text_to_image"
-MODALITY_CHOICES = [
-    ("Text to Video", MODALITY_TEXT_TO_VIDEO),
-    ("Video to Video", MODALITY_VIDEO_TO_VIDEO),
-    ("Text to Image", MODALITY_TEXT_TO_IMAGE),
-]
-
 _MODEL_CHOICES_CACHE = {}
 _VIEW_EVENTS = {
     "show_progress": "hidden",
@@ -73,24 +42,21 @@ _VIEW_EVENTS = {
 ABOUT_OVERVIEW_CONTENT = """
 # About P-Bench
 
-P-Bench compares **text-to-video**, **video-to-video**, and **text-to-image**
-models, including optimized or accelerated endpoints, on **quality, speed,
-and price**. Each view is a **dataset** scored with a **metric**, written as
-`Dataset | Metric`. There is no single score across P-Bench.
+P-Bench compares **text-to-image models**, including optimized or accelerated
+endpoints, on **quality, speed, and price**. Each view is a **dataset** scored
+with a **metric**, written as `Dataset | Metric`. There is no single score
+across P-Bench.
 
 ## How to read it
 
-1. Pick a **model type** (Text to Video, Video to Video, or Text to Image),
-   then a **dataset** and a **metric**.
+1. Pick a **dataset** and a **metric**.
 2. **Leaderboards**: ranked by that metric. Price and generation time sit in
    the same table when the source publishes them.
 3. **Pareto plots**: mark models that are not beaten on both higher score
    and lower price (or time). Only datasets with price or generation time
    can open this tab (not Arena AI).
 4. **Samples**: the same prompts, side by side. Only for datasets we
-   generated (VBench-2.0 Dataset, Qwen Image Dataset, OneIG Alignment
-   Dataset, and the Pruna Internal Video-Edit Benchmark). Video-edit
-   samples show the source clip first, then each model's edit.
+   generated (Qwen Image Dataset and OneIG Alignment Dataset).
 
 ## How a score is made
 
@@ -109,24 +75,6 @@ prompt suites, so samples are not shown.
 
 ## Current datasets
 
-### VBench-2.0 Dataset
-VBench-2.0 prompts, comparing P-Video-2 variants with Fal-hosted models.
-Quality is Datapoint Elo and Rapidata Elo from pairwise preference. Price
-is USD per second of output video. Time per second of video is Fal wall
-time, except Pruna models which use model execution time. Samples are
-available.
-
-### Pruna Internal Video-Edit Benchmark
-Pruna's internal video-to-video editing benchmark, collected by our
-research engineers. It combines prompts from public video-editing
-benchmarks with use-case examples we gathered for advertisement,
-e-commerce, real estate, concept art, and similar work. The suite also
-covers camera-angle and movement changes, lighting, and text in video
-(altering, adding, or removing it). Quality is Datapoint Elo from
-pairwise preference. Price is USD per second of output video;
-generation time is wall time per second of output video. Samples show
-the source clip beside each model's edit.
-
 ### Qwen Image Dataset
 100 prompts from the 1,000-prompt Qwen Image Bench set, sampled for coverage
 across its fine-grained (L3) categories. Metrics include Datapoint Elo,
@@ -169,17 +117,12 @@ ABOUT_DETAILS_CONTENT = """
 - **Arena Elo**: Elo published by Arena AI on their own dataset, plus
   category Elos (branding, 3D, cartoon/anime, photorealistic, art, portraits,
   text rendering).
-- **Generation time**: median and minimum generation time in seconds for
-  images, as reported in the evaluation table. For video, generation time
-  per second of output video is the more informative figure. On the
-  text-to-video benchmark this is Fal wall time, except Pruna models which
-  use model execution time. On video-edit it is end-to-end wall time. This
-  is not a p95, and we do not state warm vs cold or concurrent load. Not
+- **Generation time**: median and minimum generation time in seconds, as
+  reported in the evaluation table. This is not a p95, and we do not state
+  warm vs cold or concurrent load. Not available for Arena AI.
+- **Price**: USD per image in the evaluation table. We do not state list
+  price vs amount paid, or whether failed generations are included. Not
   available for Arena AI.
-- **Price**: USD per image for text-to-image, or USD per second of output
-  video for text-to-video and video-to-video. We do not state list price vs
-  amount paid, or whether failed generations are included. Not available
-  for Arena AI.
 
 Scores from different datasets or metrics are **not interchangeable**. A high
 OneIG alignment score is not the same quantity as a Datapoint Elo. Compare
@@ -193,21 +136,14 @@ models *within* a Dataset | Metric view.
 - **Prompt counts:** OneIG Alignment uses 100 anime, 100 human, and 99 object
   prompts (299 total). Qwen Image Dataset uses 100 prompts sampled from the
   1,000-prompt pool for roughly even coverage of its fine-grained (L3)
-  categories. The VBench-2.0 Dataset uses about 90
-  generations per model. The Pruna Internal Video-Edit Benchmark uses 78
-  prompts across advertising, e-commerce, real estate, camera, lighting,
-  text, and related categories. Artificial Analysis and Arena AI use their
-  own private prompt sets.
+  categories. Artificial Analysis and Arena AI use their own private prompt
+  sets.
 - **Generation (Qwen and OneIG):** one image per prompt per endpoint when
   the run exists. Default resolution is 1024×1024. Exceptions: FLUX 1.1 Pro
   Ultra at 2K, FLUX 2 Flex at 1008×1008, and any endpoint labeled 2K. The
   seed is derived from the prompt, so every model gets the same seed for the
   same prompt. Steps, CFG, prompt rewrite, and safety filters follow each
   endpoint's default. This does not describe Artificial Analysis or Arena AI.
-- **Generation (Text-to-Video):** one clip per prompt per endpoint when the
-  run exists. About 90 generations per model.
-- **Generation (Video-Edit):** one edited clip per prompt per endpoint when
-  the run exists. Every model sees the same source video for a prompt.
 - **Datapoint (Qwen and OneIG):** every model pair is compared on every
   prompt, with 10 votes per battle.
 - **Rapidata (Qwen and OneIG):** prompts longer than 400 characters are
@@ -279,7 +215,7 @@ def render_header():
               </svg>
             </button>
           </div>
-          <p class="app-header-tagline">Compare models on quality, speed, and price</p>
+          <p class="app-header-tagline">Compare text-to-image models on quality, speed, and price</p>
         </header>
         """,
         padding=False,
@@ -294,42 +230,10 @@ def _item(items, item_id):
     return items[0] if items else None
 
 
-def _dataset_modality(dataset):
-    return (dataset or {}).get("modality") or MODALITY_TEXT_TO_IMAGE
-
-
-def _datasets_for_modality(datasets, modality):
-    if not modality:
-        return list(datasets)
-    scoped = [
-        dataset
-        for dataset in datasets
-        if _dataset_modality(dataset) == modality
-    ]
-    return scoped or list(datasets)
-
-
-def _modality_choices(datasets):
-    present = {_dataset_modality(dataset) for dataset in datasets}
-    return [
-        (label, value) for label, value in MODALITY_CHOICES if value in present
-    ]
-
-
-def _default_dataset_id(datasets, modality, preferred=None):
-    scoped = _datasets_for_modality(datasets, modality)
-    if preferred and any(dataset["id"] == preferred for dataset in scoped):
-        return preferred
-    return scoped[0]["id"] if scoped else None
-
-
-def _dataset_choices(
-    datasets, *, modality=None, require_samples=False, require_pareto=False
-):
-    scoped = _datasets_for_modality(datasets, modality)
+def _dataset_choices(datasets, *, require_samples=False, require_pareto=False):
     return [
         (dataset["name"], dataset["id"])
-        for dataset in scoped
+        for dataset in datasets
         if (not require_samples or dataset.get("samples"))
         and (not require_pareto or _dataset_has_pareto(datasets, dataset["id"]))
     ]
@@ -340,77 +244,17 @@ def _dataset_has_samples(datasets, dataset_id):
     return bool(dataset and dataset.get("samples"))
 
 
-def _sample_model_ids(datasets, dataset_id):
-    dataset = _item(datasets, dataset_id)
-    samples = dataset.get("samples") if dataset else None
-    if not samples:
-        return set()
-    models = set(samples.get("models") or [])
-    return models | {display_model_name(model) for model in models}
-
-
-def _sample_media_map(samples):
-    return (samples or {}).get("images") or {}
-
-
-def _resolve_sample_model(samples, model):
-    media = _sample_media_map(samples)
-    if model in media:
-        return model
-    wanted = {str(model or "").strip(), display_model_name(model)}
-    wanted.discard("")
-    for key in media:
-        if key in wanted or display_model_name(key) in wanted:
-            return key
-    return None
-
-
-def _default_sample_models(samples):
-    models = list((samples or {}).get("models") or [])
-    preferred = [model for model in models if _is_pruna_model(model)]
-    preferred.sort(
-        key=lambda model: (
-            "draft" in str(model).casefold()
-            or "draft" in display_model_name(model).casefold(),
-            display_model_name(model).casefold(),
-        )
-    )
-    return (preferred or models)[:2]
-
-
-def _pareto_price_column(data):
-    columns = getattr(data, "columns", []) if data is not None else []
-    for column in _PARETO_PRICE_COLUMNS:
-        if column in columns:
-            return column
-    return None
-
-
-def _pareto_time_column(data):
-    columns = getattr(data, "columns", []) if data is not None else []
-    for column in _PARETO_TIME_COLUMNS:
-        if column in columns:
-            return column
-    return None
-
-
 def _dataset_has_pareto(datasets, dataset_id):
     dataset = _item(datasets, dataset_id)
-    data = dataset.get("data") if dataset else None
-    return (
-        _pareto_price_column(data) is not None
-        or _pareto_time_column(data) is not None
-    )
+    columns = getattr(dataset.get("data") if dataset else None, "columns", [])
+    return _PARETO_PRICE_COLUMN in columns or _PARETO_TIME_COLUMN in columns
 
 
-def _dataset_dropdown_update(datasets, tab, dataset_id, modality=None):
+def _dataset_dropdown_update(datasets, tab, dataset_id):
     """Limit the dataset list to what the current tab can show."""
-    if modality is None:
-        modality = _dataset_modality(_item(datasets, dataset_id))
     return gr.update(
         choices=_dataset_choices(
             datasets,
-            modality=modality,
             require_samples=tab == TAB_SAMPLES
             and _dataset_has_samples(datasets, dataset_id),
             require_pareto=tab == TAB_PARETO
@@ -470,25 +314,23 @@ def _metric_dropdown_value(metric_id):
     ]
 
 
-def _model_choices(datasets, dataset_id, *, require_samples=False):
+def _model_choices(datasets, dataset_id):
     cached = _MODEL_CHOICES_CACHE.get(dataset_id)
-    if cached is None:
-        dataset = _item(datasets, dataset_id)
-        data = dataset.get("data") if dataset else None
-        if data is None or "Model" not in getattr(data, "columns", []):
-            cached = []
-        else:
-            models = data["Model"].dropna().astype(str).unique().tolist()
-            # (label, value) so the UI shows the shared name but filters on the raw id.
-            cached = sorted(
-                ((display_model_name(model), model) for model in models),
-                key=lambda item: item[0].casefold(),
-            )
-        _MODEL_CHOICES_CACHE[dataset_id] = cached
-    if not require_samples:
+    if cached is not None:
         return cached
-    allowed = _sample_model_ids(datasets, dataset_id)
-    return [choice for choice in cached if choice[1] in allowed]
+    dataset = _item(datasets, dataset_id)
+    data = dataset.get("data") if dataset else None
+    if data is None or "Model" not in getattr(data, "columns", []):
+        _MODEL_CHOICES_CACHE[dataset_id] = []
+        return []
+    models = data["Model"].dropna().astype(str).unique().tolist()
+    # (label, value) so the UI shows the shared name but filters on the raw id.
+    choices = sorted(
+        ((display_model_name(model), model) for model in models),
+        key=lambda item: item[0].casefold(),
+    )
+    _MODEL_CHOICES_CACHE[dataset_id] = choices
+    return choices
 
 
 def _model_choice_values(choices):
@@ -516,11 +358,9 @@ _LEADERBOARD_IDENTITY_COLUMNS = [
     "Optimized",
 ]
 _LEADERBOARD_META_COLUMNS = [
-    "Time / Output Video Second (s)",
     "Median Generation Time (s)",
     "Min Generation Time (s)",
     "Price / Image (USD)",
-    "Price / Second of Video (USD)",
     "Evaluation Date (UTC)",
     "Date",
 ]
@@ -745,11 +585,10 @@ def _display_label(column):
         "Arena Art Elo": "Art",
         "Arena Portraits Elo": "Portraits",
         "Arena Text Rendering Elo": "Text Rendering",
+        "Raw Win Rate": "Raw win rate",
         "Median Generation Time (s)": "Median generation time",
         "Min Generation Time (s)": "Min generation time",
-        "Time / Output Video Second (s)": "Generation time per second of video",
         "Price / Image (USD)": "Price per image",
-        "Price / Second of Video (USD)": "Price per second of video",
         "Evaluation Date (UTC)": "Date",
         "Date": "Date",
     }
@@ -824,23 +663,6 @@ def _applied_key(view_state):
     )
 
 
-def _is_pruna_model(model_id) -> bool:
-    raw = str(model_id or "").casefold()
-    label = display_model_name(model_id).casefold()
-    return any(
-        value.startswith(prefix)
-        for value in (raw, label)
-        for prefix in ("p-image", "p_image", "p-video", "p_video")
-    )
-
-
-def _pareto_fill_colors(models):
-    return [
-        _PARETO_PRUNA_COLOR if _is_pruna_model(model) else _PARETO_OTHER_COLOR
-        for model in models
-    ]
-
-
 def _build_pareto_figure(
     data,
     score_column,
@@ -848,7 +670,6 @@ def _build_pareto_figure(
     x_title,
     x_hover_prefix="",
     x_hover_suffix="",
-    x_axis_type="linear",
 ):
     scatter = (
         data[["Model", score_column, x_column]]
@@ -865,8 +686,6 @@ def _build_pareto_figure(
 
     dominated = scatter.loc[[not flag for flag in on_frontier]].copy()
     frontier = scatter.loc[on_frontier].sort_values(x_column).copy()
-    dominated_colors = _pareto_fill_colors(dominated["Model"]) if not dominated.empty else []
-    frontier_colors = _pareto_fill_colors(frontier["Model"]) if not frontier.empty else []
     if not dominated.empty:
         dominated["Model"] = dominated["Model"].map(display_model_name)
     if not frontier.empty:
@@ -887,11 +706,10 @@ def _build_pareto_figure(
                 name="Below frontier",
                 text=dominated["Model"],
                 hovertemplate=hover,
-                showlegend=False,
                 marker={
                     "size": 9,
-                    "color": dominated_colors,
-                    "opacity": 0.85,
+                    "color": "#d8b4fe",
+                    "opacity": 0.8,
                     "line": {"width": 0},
                 },
             )
@@ -905,51 +723,14 @@ def _build_pareto_figure(
                 name="On frontier",
                 text=frontier["Model"],
                 hovertemplate=hover,
-                showlegend=False,
-                line={"color": _PARETO_FRONTIER_OUTLINE, "width": 2.5},
+                line={"color": "#69a45c", "width": 2.5},
                 marker={
                     "size": 12,
-                    "color": frontier_colors,
-                    "line": {"width": 2.5, "color": _PARETO_FRONTIER_OUTLINE},
+                    "color": "#69a45c",
+                    "line": {"width": 1.5, "color": "#86c077"},
                 },
             )
         )
-    for name, marker in (
-        (
-            "Pruna",
-            {
-                "size": 10,
-                "color": _PARETO_PRUNA_COLOR,
-                "line": {"width": 0},
-            },
-        ),
-        (
-            "Other models",
-            {
-                "size": 10,
-                "color": _PARETO_OTHER_COLOR,
-                "line": {"width": 0},
-            },
-        ),
-        (
-            "On frontier",
-            {
-                "size": 12,
-                "color": "rgba(0,0,0,0)",
-                "line": {"width": 2.5, "color": _PARETO_FRONTIER_OUTLINE},
-            },
-        ),
-    ):
-        fig.add_trace(
-            go.Scatter(
-                x=[None],
-                y=[None],
-                mode="markers",
-                name=name,
-                marker=marker,
-                hoverinfo="skip",
-            )
-        )
 
     score_label = _display_label(score_column)
     fig.update_layout(
@@ -974,29 +755,7 @@ def _build_pareto_figure(
     )
     axis_font = {"color": "#fafafa", "size": 13}
     tick_font = {"color": "#a3a3a3", "size": 12}
-    x_axis_ticks = {}
-    if x_axis_type == "log":
-        positive_x = scatter.loc[scatter[x_column] > 0, x_column].astype(float)
-        if not positive_x.empty:
-            minimum = positive_x.min()
-            maximum = positive_x.max()
-            tick_values = [
-                factor * (10**exponent)
-                for exponent in range(
-                    floor(log10(minimum)),
-                    ceil(log10(maximum)) + 1,
-                )
-                for factor in (1, 2, 5)
-                if minimum * 0.8 <= factor * (10**exponent) <= maximum * 1.2
-            ]
-            x_axis_ticks = {
-                "tickmode": "array",
-                "tickvals": tick_values,
-                "ticktext": [f"{value:g}" for value in tick_values],
-            }
     fig.update_xaxes(
-        type=x_axis_type,
-        **x_axis_ticks,
         showgrid=True,
         gridcolor="rgba(74, 57, 98, 0.55)",
         zeroline=False,
@@ -1015,55 +774,6 @@ def _build_pareto_figure(
     return fig
 
 
-def _is_log_scale(scale):
-    return scale == "Logarithmic"
-
-
-def _pareto_axis_type(scale):
-    return "log" if _is_log_scale(scale) else "linear"
-
-
-def _pareto_scale_radio(*extra_classes):
-    return gr.Radio(
-        choices=_PARETO_SCALE_CHOICES,
-        value=_PARETO_SCALE_DEFAULT,
-        show_label=False,
-        container=False,
-        elem_classes=["pareto-scale-toggle", *extra_classes],
-    )
-
-
-def _pareto_plot_heading(title):
-    with gr.Row(equal_height=False, elem_classes="pareto-heading-row"):
-        gr.Markdown(f"#### {title}", elem_classes="pareto-subhead")
-        with gr.Column(min_width=140, elem_classes="pareto-scale-control"):
-            return _pareto_scale_radio()
-
-
-def _default_pareto_scales():
-    return [_PARETO_SCALE_DEFAULT] * MAX_PARETO_METRICS
-
-
-def _normalize_pareto_scales(scales):
-    values = list(scales or [])
-    if len(values) < MAX_PARETO_METRICS:
-        values.extend(
-            [_PARETO_SCALE_DEFAULT] * (MAX_PARETO_METRICS - len(values))
-        )
-    return values[:MAX_PARETO_METRICS]
-
-
-def _uniform_pareto_scales(scale):
-    return [scale] * MAX_PARETO_METRICS
-
-
-def _pareto_master_scale_update(price_scales, time_scales):
-    values = list(price_scales) + list(time_scales)
-    if values and all(value == values[0] for value in values):
-        return gr.update(value=values[0])
-    return gr.update(value=None)
-
-
 def _pareto_axis(data, score_column, x_column, x_title, missing_message, empty_message, **hover):
     if x_column not in data.columns:
         return None, missing_message
@@ -1079,83 +789,56 @@ def _pareto_axis(data, score_column, x_column, x_title, missing_message, empty_m
     return fig, None
 
 
-def _pareto_pair(
-    data,
-    score_column,
-    latency_scale=_PARETO_SCALE_DEFAULT,
-    price_scale=_PARETO_SCALE_DEFAULT,
-):
+def _pareto_pair(data, score_column):
     score_missing = "No score data is available for this metric."
     if data is None or not score_column or score_column not in data.columns:
         return None, score_missing, None, score_missing
 
-    price_column = _pareto_price_column(data) or _PARETO_PRICE_COLUMN
-    price_title = _PARETO_PRICE_TITLES.get(price_column, "Price (USD)")
-    price_missing = (
-        "Price per second of video isn't available for this dataset."
-        if price_column == _PARETO_VIDEO_PRICE_COLUMN
-        else "Price per image isn't available for this dataset."
-    )
     price_fig, price_message = _pareto_axis(
         data,
         score_column,
-        price_column,
-        price_title,
-        price_missing,
+        _PARETO_PRICE_COLUMN,
+        "Price per image (USD)",
+        "Price per image isn't available for this dataset.",
         "No models have both a score and a price for this metric.",
         x_hover_prefix="$",
-        x_axis_type=_pareto_axis_type(price_scale),
-    )
-    time_column = _pareto_time_column(data) or _PARETO_TIME_COLUMN
-    time_title = _PARETO_TIME_TITLES.get(time_column, "Generation time (s)")
-    time_missing = (
-        "Generation time per second of video isn't available for this dataset."
-        if time_column == _PARETO_VIDEO_TIME_COLUMN
-        else "Min generation time isn't available for this dataset."
-    )
-    time_empty = (
-        "No models have both a score and generation time per second of "
-        "video for this metric."
-        if time_column == _PARETO_VIDEO_TIME_COLUMN
-        else "No models have both a score and a min generation time for this metric."
     )
     time_fig, time_message = _pareto_axis(
         data,
         score_column,
-        time_column,
-        time_title,
-        time_missing,
-        time_empty,
+        _PARETO_TIME_COLUMN,
+        "Min generation time (s)",
+        "Min generation time isn't available for this dataset.",
+        "No models have both a score and a min generation time for this metric.",
         x_hover_suffix="s",
-        x_axis_type=_pareto_axis_type(latency_scale),
     )
     return price_fig, price_message, time_fig, time_message
 
 
 def _pareto_dataset_message(data):
-    has_price = _pareto_price_column(data) is not None
-    has_time = _pareto_time_column(data) is not None
+    has_price = data is not None and _PARETO_PRICE_COLUMN in data.columns
+    has_time = data is not None and _PARETO_TIME_COLUMN in data.columns
     if has_price or has_time:
         return None
     return (
-        "Price and generation time aren't available for "
+        "Price per image and min generation time aren't available for "
         "this dataset, so these plots can't be drawn."
     )
 
 
 def _pareto_slot_note(price_fig, price_message, time_fig, time_message, data):
-    has_price = _pareto_price_column(data) is not None
-    has_time = _pareto_time_column(data) is not None
+    has_price = data is not None and _PARETO_PRICE_COLUMN in data.columns
+    has_time = data is not None and _PARETO_TIME_COLUMN in data.columns
     notes = []
     if has_price and not has_time:
         notes.append(
-            "Generation time isn't available for this dataset, so only "
+            "Min generation time isn't available for this dataset, so only "
             "price vs score is shown."
         )
     elif has_time and not has_price:
         notes.append(
-            "Price isn't available for this dataset, so only "
-            "time vs score is shown."
+            "Price per image isn't available for this dataset, so only min "
+            "generation time vs score is shown."
         )
     if price_fig is None and has_price:
         notes.append(price_message)
@@ -1166,18 +849,11 @@ def _pareto_slot_note(price_fig, price_message, time_fig, time_message, data):
     return " ".join(notes)
 
 
-def _pareto_slot_updates(
-    data,
-    score_columns,
-    price_scales=None,
-    time_scales=None,
-):
+def _pareto_slot_updates(data, score_columns):
     """Updates for a fixed bank of Gradio Plot slots (visible/hidden)."""
     score_columns = [column for column in (score_columns or []) if column]
-    price_scales = _normalize_pareto_scales(price_scales)
-    time_scales = _normalize_pareto_scales(time_scales)
-    has_price = _pareto_price_column(data) is not None
-    has_time = _pareto_time_column(data) is not None
+    has_price = data is not None and _PARETO_PRICE_COLUMN in data.columns
+    has_time = data is not None and _PARETO_TIME_COLUMN in data.columns
     dataset_note = _pareto_dataset_message(data)
     updates = [_pareto_note_update(dataset_note)]
     hide_all_slots = not has_price and not has_time
@@ -1197,10 +873,7 @@ def _pareto_slot_updates(
             continue
         score_column = score_columns[index]
         price_fig, price_message, time_fig, time_message = _pareto_pair(
-            data,
-            score_column,
-            latency_scale=time_scales[index],
-            price_scale=price_scales[index],
+            data, score_column
         )
         show_price = price_fig is not None
         show_time = time_fig is not None
@@ -1227,77 +900,20 @@ def _pareto_slot_updates(
     return updates
 
 
-def _pareto_all_scale_updates(data, score_columns, scale):
-    """Apply one scale to every Pareto plot and radio."""
-    score_columns = [column for column in (score_columns or []) if column]
-    price_updates = []
-    time_updates = []
-    for index in range(MAX_PARETO_METRICS):
-        if index >= len(score_columns):
-            price_updates.append(gr.skip())
-            time_updates.append(gr.skip())
-            continue
-        price_fig, _, time_fig, _ = _pareto_pair(
-            data,
-            score_columns[index],
-            latency_scale=scale,
-            price_scale=scale,
-        )
-        price_updates.append(_pareto_plot_update(price_fig))
-        time_updates.append(_pareto_plot_update(time_fig))
-    radio_updates = [
-        gr.update(value=scale) for _ in range(MAX_PARETO_METRICS * 2)
-    ]
-    return price_updates + time_updates + radio_updates
-
-
 def _samples_html(samples, selected_models, num_prompts, seed=0):
     if not samples:
         return _pareto_unavailable_html(
             "Samples aren't available for this dataset."
         )
-    models = [
-        resolved
-        for model in (selected_models or [])
-        if (resolved := _resolve_sample_model(samples, model))
-    ]
+    images = samples.get("images", {})
+    models = [model for model in (selected_models or []) if model in images]
     if not models:
-        models = _default_sample_models(samples)
+        models = (samples.get("models") or [])[:2]
     return _build_compare_samples_html(samples, models, num_prompts, seed)
 
 
-def _compare_media_html(url, label, *, kind):
-    safe_url = escape(url, quote=True)
-    safe_label = escape(label)
-    if kind == "video":
-        return (
-            f'<video src="{safe_url}" controls preload="metadata" '
-            f'playsinline></video>'
-        )
-    return (
-        f'<a href="{safe_url}" target="_blank" rel="noopener noreferrer">'
-        f'<img src="{safe_url}" alt="{safe_label} sample" loading="lazy" />'
-        f"</a>"
-    )
-
-
-def _compare_cell_html(label, url, *, kind, extra_class=""):
-    classes = "compare-cell"
-    if extra_class:
-        classes = f"{classes} {extra_class}"
-    return f"""
-                <div class="{classes}">
-                    <div class="compare-model-label">{escape(label)}</div>
-                    {_compare_media_html(url, label, kind=kind)}
-                </div>
-                """
-
-
 def _build_compare_samples_html(samples, selected_models, num_prompts, seed=0):
     selected_models = list(selected_models or [])[:MAX_COMPARE_MODELS]
-    media = _sample_media_map(samples)
-    kind = (samples or {}).get("kind") or "image"
-    source_videos = (samples or {}).get("source_videos") or {}
 
     if not selected_models:
         return (
@@ -1308,7 +924,7 @@ def _build_compare_samples_html(samples, selected_models, num_prompts, seed=0):
 
     shared_prompt_ids = None
     for model in selected_models:
-        model_prompt_ids = set(media.get(model) or [])
+        model_prompt_ids = set(samples["images"][model])
         shared_prompt_ids = (
             model_prompt_ids
             if shared_prompt_ids is None
@@ -1328,31 +944,29 @@ def _build_compare_samples_html(samples, selected_models, num_prompts, seed=0):
     rng.shuffle(prompt_pool)
     chosen = prompt_pool[: max(1, min(int(num_prompts), len(prompt_pool)))]
 
+    columns = len(selected_models)
     blocks = []
     for index, prompt_id in enumerate(chosen, start=1):
         prompt_text = escape(samples["prompts"].get(prompt_id, ""))
         cells = []
-        source_url = source_videos.get(prompt_id)
-        if source_url:
-            cells.append(
-                _compare_cell_html(
-                    "Source", source_url, kind="video", extra_class="compare-source"
-                )
-            )
         for model in selected_models:
+            image_url = escape(samples["images"][model][prompt_id], quote=True)
             cells.append(
-                _compare_cell_html(
-                    display_model_name(model),
-                    media[model][prompt_id],
-                    kind=kind,
-                )
+                f"""
+                <div class="compare-cell">
+                    <div class="compare-model-label">{escape(display_model_name(model))}</div>
+                    <a href="{image_url}" target="_blank" rel="noopener noreferrer">
+                        <img src="{image_url}" alt="{escape(display_model_name(model))} sample" loading="lazy" />
+                    </a>
+                </div>
+                """
             )
-        columns = len(cells)
         blocks.append(
             f"""
             <div class="compare-prompt-block">
                 <div class="compare-prompt-meta">
                     <span>Prompt {index}</span>
+                    <span>{escape(prompt_id)}</span>
                 </div>
                 <p class="compare-prompt-text">{prompt_text}</p>
                 <div class="compare-row" style="grid-template-columns: repeat({columns}, minmax(0, 1fr));">
@@ -1380,19 +994,9 @@ def _filter_row(datasets, metrics, default_dataset_id, default_metric_id=None):
     metric_id = _coerce_metric(
         datasets, metrics, default_dataset_id, default_metric_id
     )
-    default_modality = _dataset_modality(_item(datasets, default_dataset_id))
     with gr.Row(elem_classes="view-filters"):
-        modality_dd = gr.Dropdown(
-            choices=_modality_choices(datasets),
-            value=default_modality,
-            label="Model Type",
-            type="value",
-            filterable=False,
-            scale=1,
-            min_width=170,
-        )
         dataset_dd = gr.Dropdown(
-            choices=_dataset_choices(datasets, modality=default_modality),
+            choices=_dataset_choices(datasets),
             value=default_dataset_id,
             label="Dataset",
             type="value",
@@ -1424,7 +1028,7 @@ def _filter_row(datasets, metrics, default_dataset_id, default_metric_id=None):
             min_width=180,
             elem_classes="filter-chips",
         )
-    return modality_dd, dataset_dd, metric_dd, models_dd
+    return dataset_dd, metric_dd, models_dd
 
 
 def render_image_workspace(datasets, metrics, default_dataset_id, default_metric_id):
@@ -1439,17 +1043,15 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
         with gr.Column(elem_classes="workspace-filters") as filters_host:
             gr.Markdown(
                 "<p class='filter-help'>"
-                "Start with Model Type to switch modalities. The rest of "
-                "the filters follow you across Leaderboards, Pareto plots, "
-                "and Samples. "
-                "Samples only lists datasets and models we have generations "
-                "for; Pareto plots only lists datasets with price or "
-                "generation time. Search in Models, or leave it empty to "
-                "include every model."
+                "These filters apply to Leaderboards, Pareto plots, and Samples. "
+                "On Samples, only datasets we have generations for are listed. "
+                "On Pareto plots, only datasets with price or generation time "
+                "are listed. Search in Models, or leave it empty to include "
+                "every model."
                 "</p>",
                 elem_classes="filter-help-host",
             )
-            modality_dd, dataset_dd, metric_dd, models_dd = _filter_row(
+            dataset_dd, metric_dd, models_dd = _filter_row(
                 datasets, metrics, default_dataset_id, None
             )
         with gr.Tabs(elem_classes="main-tabs") as main_tabs:
@@ -1515,27 +1117,12 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
             ) as pp_tab:
                 gr.Markdown(
                     "<p class='view-help'>"
-                    "Score against price and generation time. Green points are on "
-                    "the frontier; lavender points sit below it. Hover a point to "
-                    "see which model it is."
-                    "</p>"
-                    "<p class='view-help'>"
-                    "Plots use a logarithmic scale by default. You can switch "
-                    "to linear for all plots, or individually for each plot."
+                    "Score against price and generation time. Green points are on the "
+                    "frontier; lavender points sit below it. Hover a point to see "
+                    "which model it is."
                     "</p>",
                     elem_classes="view-help-host",
                 )
-                with gr.Row(
-                    equal_height=False,
-                    elem_classes="pareto-scale-all-row",
-                ):
-                    gr.HTML(
-                        "<span class='pareto-scale-all-label'>All plots</span>",
-                        padding=False,
-                    )
-                    pareto_all_scale = _pareto_scale_radio(
-                        "pareto-scale-toggle-all",
-                    )
                 pareto_dataset_note = gr.HTML(
                     "",
                     padding=False,
@@ -1561,8 +1148,9 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
                                 min_width=320,
                                 elem_classes="pareto-col",
                             ) as slot_price_col:
-                                slot_price_scale = _pareto_plot_heading(
-                                    "Price vs score"
+                                gr.Markdown(
+                                    "#### Price vs score",
+                                    elem_classes="pareto-subhead",
                                 )
                                 slot_price = gr.Plot(
                                     value=None,
@@ -1574,8 +1162,9 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
                                 min_width=320,
                                 elem_classes="pareto-col",
                             ) as slot_time_col:
-                                slot_time_scale = _pareto_plot_heading(
-                                    "Time vs score"
+                                gr.Markdown(
+                                    "#### Min generation time vs score",
+                                    elem_classes="pareto-subhead",
                                 )
                                 slot_time = gr.Plot(
                                     value=None,
@@ -1596,10 +1185,8 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
                             slot_layout,
                             slot_price_col,
                             slot_price,
-                            slot_price_scale,
                             slot_time_col,
                             slot_time,
-                            slot_time_scale,
                         )
                     )
 
@@ -1611,8 +1198,7 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
                 with gr.Column(visible=bool(initial_samples)) as samples_panel:
                     gr.Markdown(
                         f"<p class='view-help'>"
-                        f"The same prompts, side by side. Video edits show the "
-                        f"source clip first. Select up to "
+                        f"The same prompts, side by side. Select up to "
                         f"<strong>{MAX_COMPARE_MODELS}</strong> models above, or leave "
                         f"Models empty for two defaults."
                         f"</p>",
@@ -1649,16 +1235,12 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
             with gr.TabItem("About", id=TAB_ABOUT) as about_tab:
                 render_about()
 
-    def _synced_filters(
-        dataset_id, metric_id, models, *, clear_metric=False, require_samples=False
-    ):
+    def _synced_filters(dataset_id, metric_id, models, *, clear_metric=False):
         if clear_metric:
             metric_id = []
         else:
             metric_id = _coerce_metric(datasets, metrics, dataset_id, metric_id)
-        model_choices = _model_choices(
-            datasets, dataset_id, require_samples=require_samples
-        )
+        model_choices = _model_choices(datasets, dataset_id)
         model_values = set(_model_choice_values(model_choices))
         models = [model for model in (models or []) if model in model_values]
         metric_choices = _metric_dropdown_choices(datasets, metrics, dataset_id)
@@ -1730,14 +1312,8 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
     ):
         prev = dict(view_state or {})
         extras = extras or {}
-        modality = _dataset_modality(_item(datasets, dataset_id))
-        last_by_modality = dict(prev.get("dataset_by_modality") or {})
-        if dataset_id:
-            last_by_modality[modality] = dataset_id
         return {
             "dataset_id": dataset_id,
-            "modality": modality,
-            "dataset_by_modality": last_by_modality,
             "metric_id": metric_id,
             "models": list(models or []),
             "current_tab": tab,
@@ -1748,8 +1324,6 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
             "optimized": list(
                 extras.get("optimized", prev.get("optimized") or [])
             ),
-            "price_scales": _normalize_pareto_scales(prev.get("price_scales")),
-            "time_scales": _normalize_pareto_scales(prev.get("time_scales")),
             "stale": {
                 TAB_LEADERBOARDS: not flags["include_leaderboard"],
                 TAB_PARETO: not flags["include_pareto"],
@@ -1809,8 +1383,6 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
         include_leaderboard=True,
         include_pareto=False,
         include_samples=False,
-        price_scales=None,
-        time_scales=None,
     ):
         view = resolve_view(datasets, metrics, dataset_id, metric_id)
         data = view["data"]
@@ -1831,12 +1403,7 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
             ranking_html = gr.skip()
         if include_pareto:
             pareto_data = _filter_leaderboard(data, [], [], [], models=models)
-            pareto_updates = _pareto_slot_updates(
-                pareto_data,
-                view["score_columns"],
-                price_scales=price_scales,
-                time_scales=time_scales,
-            )
+            pareto_updates = _pareto_slot_updates(pareto_data, view["score_columns"])
         else:
             pareto_updates = _pareto_skip_updates()
         if include_samples:
@@ -1875,24 +1442,13 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
         tab = view_state.get("current_tab") or TAB_LEADERBOARDS
         selected_raw = _normalize_metric_ids(metric_id)
         incoming_models = list(models or [])
-        filter_changed = source in {"dataset", "modality"}
-        dataset_changed = filter_changed and dataset_id != view_state.get(
+        dataset_changed = source == "dataset" and dataset_id != view_state.get(
             "dataset_id"
         )
-        can_pareto = _dataset_has_pareto(datasets, dataset_id)
-        can_samples = _dataset_has_samples(datasets, dataset_id)
-        selected_tab = tab
-        if filter_changed:
-            if tab == TAB_SAMPLES and not can_samples:
-                selected_tab = TAB_LEADERBOARDS
-            elif tab == TAB_PARETO and not can_pareto:
-                selected_tab = TAB_LEADERBOARDS
+
+        if source == "dataset":
             synced = _synced_filters(
-                dataset_id,
-                metric_id,
-                models,
-                clear_metric=source == "modality" or dataset_changed,
-                require_samples=selected_tab == TAB_SAMPLES,
+                dataset_id, metric_id, models, clear_metric=dataset_changed
             )
             dataset_id, metric_id, models = synced[:3]
             metric_update, models_update = synced[3], synced[4]
@@ -1930,13 +1486,20 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
         ):
             return None
 
+        selected_tab = tab
         extras = (
             list(platform_value or []),
             list(owner_value or []),
             list(optimized_value or []),
         )
         extra_updates = None
-        if filter_changed:
+        can_pareto = _dataset_has_pareto(datasets, dataset_id)
+        can_samples = _dataset_has_samples(datasets, dataset_id)
+        if source == "dataset":
+            if tab == TAB_SAMPLES and not can_samples:
+                selected_tab = TAB_LEADERBOARDS
+            elif tab == TAB_PARETO and not can_pareto:
+                selected_tab = TAB_LEADERBOARDS
             view = resolve_view(datasets, metrics, dataset_id, metric_id)
             extra_updates = _leaderboard_extras(
                 view["data"] if view else None,
@@ -1974,8 +1537,6 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
                 extras[2],
                 num_prompts,
                 seed,
-                price_scales=view_state.get("price_scales"),
-                time_scales=view_state.get("time_scales"),
                 **flags,
             ),
             "state": _commit_state(
@@ -1989,67 +1550,6 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
             ),
         }
 
-    def on_modality(
-        modality,
-        dataset_id,
-        metric_id,
-        models,
-        platform_value,
-        owner_value,
-        optimized_value,
-        num_prompts,
-        seed,
-        view_state,
-    ):
-        view_state = dict(view_state or {})
-        last_by_modality = dict(view_state.get("dataset_by_modality") or {})
-        current_modality = view_state.get("modality") or _dataset_modality(
-            _item(datasets, dataset_id)
-        )
-        if dataset_id:
-            last_by_modality[current_modality] = dataset_id
-        dataset_id = _default_dataset_id(
-            datasets, modality, last_by_modality.get(modality)
-        )
-        view_state["modality"] = modality
-        view_state["dataset_by_modality"] = last_by_modality
-        result = _apply_filter_change(
-            "modality",
-            dataset_id,
-            metric_id,
-            models,
-            platform_value,
-            owner_value,
-            optimized_value,
-            num_prompts,
-            seed,
-            view_state,
-        )
-        if result is None:
-            return _skip_all(len(dataset_outputs))
-        extras = result["extra_updates"]
-        return (
-            _dataset_dropdown_update(
-                datasets,
-                result["selected_tab"],
-                result["dataset_id"],
-                modality=modality,
-            ),
-            result["metric_update"],
-            result["models_update"],
-            extras[6],
-            extras[0],
-            extras[1],
-            extras[2],
-            *result["views"],
-            gr.update(interactive=result["can_pareto"]),
-            gr.update(interactive=result["can_samples"]),
-            gr.update(selected=result["selected_tab"])
-            if result["selected_tab"] != result["tab"]
-            else gr.skip(),
-            result["state"],
-        )
-
     def on_dataset(
         dataset_id,
         metric_id,
@@ -2177,17 +1677,7 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
             )
         dataset_update = _dataset_dropdown_update(datasets, tab, dataset_id)
         metric_id = _coerce_metric(datasets, metrics, dataset_id, metric_id)
-        require_samples = tab == TAB_SAMPLES
-        model_choices = _model_choices(
-            datasets, dataset_id, require_samples=require_samples
-        )
-        allowed_models = set(_model_choice_values(model_choices))
-        models = [model for model in (models or []) if model in allowed_models]
-        models_update = (
-            gr.update(choices=model_choices, value=models)
-            if (prev_tab == TAB_SAMPLES) != require_samples
-            else gr.skip()
-        )
+        models = list(models or [])
         view_state["current_tab"] = tab
         view_state["dataset_id"] = dataset_id
         view_state["metric_id"] = metric_id
@@ -2224,7 +1714,7 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
             filters_vis,
             dataset_update,
             metric_vis,
-            models_update,
+            gr.skip(),
             *lb_filters,
         )
         tab_select = (
@@ -2249,8 +1739,6 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
             optimized_value,
             num_prompts,
             seed,
-            price_scales=view_state.get("price_scales"),
-            time_scales=view_state.get("time_scales"),
             **flags,
         )
         stale[tab] = False
@@ -2307,70 +1795,6 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
             next_seed,
         )
 
-    def _on_pareto_plot_scale(slot_index, axis):
-        def handler(dataset_id, metric_id, models, scale, view_state):
-            view_state = dict(view_state or {})
-            price_scales = _normalize_pareto_scales(
-                view_state.get("price_scales")
-            )
-            time_scales = _normalize_pareto_scales(
-                view_state.get("time_scales")
-            )
-            if axis == "price":
-                if price_scales[slot_index] == scale:
-                    return gr.skip(), gr.skip(), gr.skip()
-                price_scales[slot_index] = scale
-            else:
-                if time_scales[slot_index] == scale:
-                    return gr.skip(), gr.skip(), gr.skip()
-                time_scales[slot_index] = scale
-            view_state["price_scales"] = price_scales
-            view_state["time_scales"] = time_scales
-            master_scale = _pareto_master_scale_update(
-                price_scales, time_scales
-            )
-            view = resolve_view(datasets, metrics, dataset_id, metric_id)
-            score_columns = [
-                column for column in (view["score_columns"] or []) if column
-            ]
-            if slot_index >= len(score_columns):
-                return gr.skip(), master_scale, view_state
-            data = _filter_leaderboard(
-                view["data"], [], [], [], models=list(models or [])
-            )
-            price_fig, _, time_fig, _ = _pareto_pair(
-                data,
-                score_columns[slot_index],
-                latency_scale=time_scales[slot_index],
-                price_scale=price_scales[slot_index],
-            )
-            fig = price_fig if axis == "price" else time_fig
-            return _pareto_plot_update(fig), master_scale, view_state
-
-        handler.__name__ = f"on_pareto_{axis}_scale_{slot_index}"
-        return handler
-
-    def on_pareto_all_scale(dataset_id, metric_id, models, scale, view_state):
-        if scale not in _PARETO_SCALE_VALUES:
-            return (*_skip_all(MAX_PARETO_METRICS * 4), gr.skip())
-        view_state = dict(view_state or {})
-        scales = _uniform_pareto_scales(scale)
-        if (
-            _normalize_pareto_scales(view_state.get("price_scales")) == scales
-            and _normalize_pareto_scales(view_state.get("time_scales")) == scales
-        ):
-            return (*_skip_all(MAX_PARETO_METRICS * 4), gr.skip())
-        view_state["price_scales"] = scales
-        view_state["time_scales"] = scales
-        view = resolve_view(datasets, metrics, dataset_id, metric_id)
-        data = _filter_leaderboard(
-            view["data"], [], [], [], models=list(models or [])
-        )
-        return (
-            *_pareto_all_scale_updates(data, view["score_columns"], scale),
-            view_state,
-        )
-
     def _on_tab(tab):
         def handler(
             dataset_id,
@@ -2399,20 +1823,15 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
         handler.__name__ = f"on_tab_{tab}"
         return handler
 
-    default_modality = _dataset_modality(_item(datasets, default_dataset_id))
     view_state = gr.State(
         {
             "dataset_id": default_dataset_id,
-            "modality": default_modality,
-            "dataset_by_modality": {default_modality: default_dataset_id},
             "metric_id": None,
             "models": [],
             "current_tab": TAB_LEADERBOARDS,
             "platform": [],
             "owner": [],
             "optimized": [],
-            "price_scales": _default_pareto_scales(),
-            "time_scales": _default_pareto_scales(),
             "stale": {
                 TAB_LEADERBOARDS: False,
                 TAB_PARETO: True,
@@ -2424,7 +1843,7 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
         pareto_dataset_note,
         *[
             component
-            for slot_group, slot_title, slot_note, slot_layout, slot_price_col, slot_price, slot_price_scale, slot_time_col, slot_time, slot_time_scale in pareto_slots
+            for slot_group, slot_title, slot_note, slot_layout, slot_price_col, slot_price, slot_time_col, slot_time in pareto_slots
             for component in (
                 slot_group,
                 slot_title,
@@ -2437,24 +1856,6 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
             )
         ],
     ]
-    pareto_all_scale_outputs = [
-        *[
-            slot_price
-            for _, _, _, _, _, slot_price, _, _, _, _ in pareto_slots
-        ],
-        *[
-            slot_time
-            for _, _, _, _, _, _, _, _, slot_time, _ in pareto_slots
-        ],
-        *[
-            slot_price_scale
-            for _, _, _, _, _, _, slot_price_scale, _, _, _ in pareto_slots
-        ],
-        *[
-            slot_time_scale
-            for _, _, _, _, _, _, _, _, _, slot_time_scale in pareto_slots
-        ],
-    ]
     view_inputs = [
         platform,
         owner,
@@ -2486,12 +1887,6 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
         main_tabs,
         view_state,
     ]
-    modality_dd.change(
-        on_modality,
-        inputs=[modality_dd, *filter_inputs],
-        outputs=dataset_outputs,
-        **_VIEW_EVENTS,
-    )
     dataset_dd.change(
         on_dataset,
         inputs=filter_inputs,
@@ -2565,56 +1960,6 @@ def render_image_workspace(datasets, metrics, default_dataset_id, default_metric
             show_progress="hidden",
         )
 
-    pareto_all_scale.change(
-        on_pareto_all_scale,
-        inputs=[
-            dataset_dd,
-            metric_dd,
-            models_dd,
-            pareto_all_scale,
-            view_state,
-        ],
-        outputs=[*pareto_all_scale_outputs, view_state],
-        **_VIEW_EVENTS,
-    )
-
-    for slot_index, (
-        _,
-        _,
-        _,
-        _,
-        _,
-        slot_price,
-        slot_price_scale,
-        _,
-        slot_time,
-        slot_time_scale,
-    ) in enumerate(pareto_slots):
-        slot_price_scale.change(
-            _on_pareto_plot_scale(slot_index, "price"),
-            inputs=[
-                dataset_dd,
-                metric_dd,
-                models_dd,
-                slot_price_scale,
-                view_state,
-            ],
-            outputs=[slot_price, pareto_all_scale, view_state],
-            **_VIEW_EVENTS,
-        )
-        slot_time_scale.change(
-            _on_pareto_plot_scale(slot_index, "time"),
-            inputs=[
-                dataset_dd,
-                metric_dd,
-                models_dd,
-                slot_time_scale,
-                view_state,
-            ],
-            outputs=[slot_time, pareto_all_scale, view_state],
-            **_VIEW_EVENTS,
-        )
-
     prompt_count.change(
         on_samples_controls,
         inputs=[dataset_dd, models_dd, prompt_count, seed_state],