mteb 2.3.1__py3-none-any.whl → 2.3.2__py3-none-any.whl

mteb/benchmarks/benchmarks/__init__.py

@@ -43,6 +43,7 @@ from mteb.benchmarks.benchmarks.benchmarks import (
     VN_MTEB,
     CoIR,
     MTEB_code,
+    MTEB_MAIN_RU_v1_1,
     MTEB_multilingual_v1,
     MTEB_multilingual_v2,
     RAR_b,
@@ -113,6 +114,7 @@ __all__ = [
     "VISUAL_DOCUMENT_RETRIEVAL",
     "VN_MTEB",
     "CoIR",
+    "MTEB_MAIN_RU_v1_1",
     "MTEB_code",
     "MTEB_multilingual_v1",
     "MTEB_multilingual_v2",

mteb/benchmarks/benchmarks/benchmarks.py

@@ -185,7 +185,7 @@ We recommend that you use [MTEB(eng, v2)](http://mteb-leaderboard.hf.space/?benc
 
 MTEB_MAIN_RU = Benchmark(
     name="MTEB(rus, v1)",
-    display_name="Russian",
+    display_name="Russian legacy",
     icon="https://github.com/lipis/flag-icons/raw/260c91531be024944c6514130c5defb2ebb02b7d/flags/4x3/ru.svg",
     tasks=MTEBTasks(
         get_tasks(
@@ -240,6 +240,67 @@ MTEB_MAIN_RU = Benchmark(
   year = {2024},
 }
 """,
+    contacts=["Samoed", "artemsnegirev", "Drozhzhinastya"],
+)
+
+MTEB_MAIN_RU_v1_1 = Benchmark(
+    name="MTEB(rus, v1.1)",
+    display_name="Russian",
+    icon="https://github.com/lipis/flag-icons/raw/260c91531be024944c6514130c5defb2ebb02b7d/flags/4x3/ru.svg",
+    tasks=MTEBTasks(
+        get_tasks(
+            languages=["rus"],
+            tasks=[
+                # Classification
+                "GeoreviewClassification",
+                "HeadlineClassification",
+                "InappropriatenessClassification",
+                "KinopoiskClassification",
+                "MassiveIntentClassification",
+                "MassiveScenarioClassification",
+                "RuReviewsClassification",
+                "RuSciBenchGRNTIClassification",
+                "RuSciBenchOECDClassification",
+                # Clustering
+                "GeoreviewClusteringP2P",
+                "RuSciBenchGRNTIClusteringP2P",
+                "RuSciBenchOECDClusteringP2P",
+                # MultiLabelClassification
+                "CEDRClassification",
+                "SensitiveTopicsClassification",
+                # PairClassification
+                "TERRa",
+                # Reranking
+                "MIRACLReranking",
+                "RuBQReranking",
+                # Retrieval
+                "MIRACLRetrievalHardNegatives.v2",
+                "RiaNewsRetrievalHardNegatives.v2",
+                "RuBQRetrieval",
+                # STS
+                "RUParaPhraserSTS",
+                "STS22",
+            ],
+        )
+        + get_tasks(
+            tasks=["RuSTSBenchmarkSTS"],
+            eval_splits=["test"],
+        )
+    ),
+    description="A Russian version of the Massive Text Embedding Benchmark covering the task categories of classification, clustering, reranking, pair classification, retrieval, and semantic similarity. In v1.1, MIRACLRetrieval and RiaNewsRetrieval were replaced with their HardNegatives variants for improved time-optimization measurement. MIRACLRetrievalHardNegatives and RiaNewsRetrievalHardNegatives are used in their updated versions (v2), both of which include improved default prompts.",
+    reference="https://aclanthology.org/2023.eacl-main.148/",
+    citation=r"""
+@misc{snegirev2024russianfocusedembeddersexplorationrumteb,
+  archiveprefix = {arXiv},
+  author = {Artem Snegirev and Maria Tikhonova and Anna Maksimova and Alena Fenogenova and Alexander Abramov},
+  eprint = {2408.12503},
+  primaryclass = {cs.CL},
+  title = {The Russian-focused embedders' exploration: ruMTEB benchmark and Russian embedding model design},
+  url = {https://arxiv.org/abs/2408.12503},
+  year = {2024},
+}
+""",
+    contacts=["Samoed", "artemsnegirev", "Drozhzhinastya"],
 )
 
 

mteb/leaderboard/app.py

@@ -5,7 +5,7 @@ import tempfile
 import time
 import warnings
 from pathlib import Path
-from typing import Literal, get_args
+from typing import Literal
 from urllib.parse import urlencode
 
 import cachetools
@@ -14,7 +14,6 @@ import pandas as pd
 
 import mteb
 from mteb import BenchmarkResults
-from mteb.abstasks.task_metadata import TaskDomain, TaskType
 from mteb.benchmarks.benchmark import RtebBenchmark
 from mteb.cache import ResultCache
 from mteb.leaderboard.benchmark_selector import (
@@ -29,7 +28,6 @@ from mteb.leaderboard.table import (
     apply_summary_styling_from_benchmark,
 )
 from mteb.leaderboard.text_segments import ACKNOWLEDGEMENT, FAQ
-from mteb.types import Modalities
 
 logger = logging.getLogger(__name__)
 
@@ -139,7 +137,10 @@ def _update_task_info(task_names: str) -> gr.DataFrame:
     df["languages"] = df["languages"].map(_format_list)
     df = df.sort_values("name")
     df["domains"] = df["domains"].map(_format_list)
-    df["name"] = f'<a href="{df["reference"]}" target="_blank">{df["name"]}</a>'
+    df["name"] = df.apply(
+        lambda row: f'<a href="{row["reference"]}" target="_blank">{row["name"]}</a>',
+        axis=1,
+    )
     df["modalities"] = df["modalities"].map(_format_list)
     df = df.rename(
         columns={
@@ -155,9 +156,8 @@ def _update_task_info(task_names: str) -> gr.DataFrame:
     df = df.drop(columns="reference")
     return gr.DataFrame(
         df,
-        datatype=["markdown"] + ["str"] * (len(df.columns) - 1),
-        show_copy_button=True,
-        show_fullscreen_button=True,
+        datatype=["markdown"] + ["str"] * (len(df.columns) - 1),  # type: ignore
+        buttons=["copy", "fullscreen"],
         show_search="filter",
     )
 
@@ -215,6 +215,110 @@ def _should_show_zero_shot_filter(benchmark_name: str) -> bool:
     return True
 
 
+@cachetools.cached(
+    cache={},
+    key=lambda benchmark_name, all_benchmark_results: hash(benchmark_name),
+)
+def _cache_on_benchmark_select(benchmark_name, all_benchmark_results):
+    start_time = time.time()
+    benchmark = mteb.get_benchmark(benchmark_name)
+    languages = [task.languages for task in benchmark.tasks if task.languages]
+    languages = set(itertools.chain.from_iterable(languages))
+    languages = sorted(languages)
+    domains = [
+        task.metadata.domains for task in benchmark.tasks if task.metadata.domains
+    ]
+    domains = set(itertools.chain.from_iterable(domains))
+    types = {task.metadata.type for task in benchmark.tasks if task.metadata.type}
+    modalities = set()
+    for task in benchmark.tasks:
+        modalities.update(task.metadata.modalities)
+    languages, domains, types, modalities = (
+        sorted(languages),
+        sorted(domains),
+        sorted(types),
+        sorted(modalities),
+    )
+    elapsed = time.time() - start_time
+    benchmark_results = all_benchmark_results[benchmark_name]
+    scores = benchmark_results._get_scores(format="long")
+    logger.debug(f"on_benchmark_select callback: {elapsed}s")
+    show_zero_shot = _should_show_zero_shot_filter(benchmark_name)
+
+    # Calculate initial models for this benchmark to avoid race conditions
+    benchmark_tasks = sorted([task.metadata.name for task in benchmark.tasks])
+    all_models_in_scores = list({entry["model_name"] for entry in scores})
+    initial_models = _filter_models(
+        all_models_in_scores,
+        benchmark_tasks,
+        availability=None,
+        compatibility=[],
+        instructions=None,
+        max_model_size=MAX_MODEL_SIZE,
+        zero_shot_setting="allow_all",
+    )
+    # Sort to ensure consistency with update_models
+    initial_models = sorted(initial_models)
+
+    return (
+        languages,
+        domains,
+        types,
+        modalities,
+        benchmark_tasks,
+        scores,
+        show_zero_shot,
+        initial_models,
+    )
+
+
+@cachetools.cached(
+    cache={},
+    key=lambda benchmark_name,
+    type_select,
+    domain_select,
+    lang_select,
+    modality_select: hash(
+        (
+            hash(benchmark_name),
+            hash(tuple(type_select)),
+            hash(tuple(domain_select)),
+            hash(tuple(lang_select)),
+            hash(tuple(modality_select)),
+        )
+    ),
+)
+def _cache_update_task_list(
+    benchmark_name, type_select, domain_select, lang_select, modality_select
+):
+    if not len(lang_select):
+        return []
+    start_time = time.time()
+    benchmark_tasks = []
+    tasks_to_keep = []
+    for task in mteb.get_benchmark(benchmark_name).tasks:
+        benchmark_tasks.append(task.metadata.name)
+        if task.metadata.type not in type_select:
+            continue
+        if task.metadata.domains and not (
+            set(task.metadata.domains) & set(domain_select)
+        ):
+            continue
+        if task.languages and not (set(task.languages) & set(lang_select)):
+            continue
+        if task.metadata.modalities and not (
+            set(task.metadata.modalities) & set(modality_select)
+        ):
+            continue
+        tasks_to_keep.append(task.metadata.name)
+    benchmark_tasks.sort()
+    tasks_to_keep.sort()
+    elapsed = time.time() - start_time
+    logger.debug(f"update_task_list callback: {elapsed}s")
+
+    return benchmark_tasks, tasks_to_keep
+
+
 def get_leaderboard_app(cache: ResultCache = ResultCache()) -> gr.Blocks:
     """Returns a Gradio Blocks app for the MTEB leaderboard."""
     logger.info("Loading all benchmark results")
@@ -227,6 +331,7 @@ def get_leaderboard_app(cache: ResultCache = ResultCache()) -> gr.Blocks:
         benchmark.name: all_results.select_tasks(benchmark.tasks).join_revisions()
         for benchmark in benchmarks
     }
+
     default_benchmark = mteb.get_benchmark(DEFAULT_BENCHMARK_NAME)
     default_results = all_benchmark_results[default_benchmark.name]
     logger.info("Benchmark results loaded")
@@ -257,55 +362,48 @@ def get_leaderboard_app(cache: ResultCache = ResultCache()) -> gr.Blocks:
         default_benchmark, filtered_benchmark_results
     )
 
-    lang_select = gr.Dropdown(
-        LANGUAGE,
+    lang_select = gr.CheckboxGroup(
+        sorted(default_results.languages),
         value=sorted(default_results.languages),
-        allow_custom_value=True,
-        multiselect=True,
+        show_label=True,
+        show_select_all=True,
         label="Language",
         info="Select languages to include.",
     )
-    type_select = gr.Dropdown(
-        sorted(get_args(TaskType)),
+    type_select = gr.CheckboxGroup(
+        sorted(default_results.task_types),
         value=sorted(default_results.task_types),
-        multiselect=True,
+        show_label=True,
+        show_select_all=True,
         label="Task Type",
         info="Select task types to include.",
     )
-    domain_select = gr.Dropdown(
-        sorted(get_args(TaskDomain)),
+    domain_select = gr.CheckboxGroup(
+        sorted(default_results.domains),
         value=sorted(default_results.domains),
-        multiselect=True,
+        show_label=True,
+        show_select_all=True,
         label="Domain",
         info="Select domains to include.",
     )
-    task_select = gr.Dropdown(
-        sorted(all_results.task_names),
+    task_select = gr.CheckboxGroup(
+        sorted(default_results.task_names),
         value=sorted(default_results.task_names),
-        allow_custom_value=True,
-        multiselect=True,
+        show_label=True,
+        show_select_all=True,
         label="Task",
         info="Select specific tasks to include",
     )
-    modality_select = gr.Dropdown(
-        sorted(get_args(Modalities)),
+    modality_select = gr.CheckboxGroup(
+        sorted(default_results.modalities),
         value=sorted(default_results.modalities),
-        multiselect=True,
+        show_label=True,
+        show_select_all=True,
         label="Modality",
         info="Select modalities to include.",
     )
 
-    head = """
-      <link href="https://cdn.jsdelivr.net/npm/tailwindcss@2.2.19/dist/tailwind.min.css" rel="stylesheet">
-    """
-
-    with gr.Blocks(
-        fill_width=True,
-        theme=gr.themes.Soft(
-            font=[gr.themes.GoogleFont("Roboto Mono"), "Arial", "sans-serif"],
-        ),
-        head=head,
-    ) as demo:
+    with gr.Blocks(fill_width=True) as demo:
         with gr.Sidebar(
             position="left",
             label="Benchmark Selection and Customization",
@@ -465,62 +563,25 @@ def get_leaderboard_app(cache: ResultCache = ResultCache()) -> gr.Blocks:
         # This sets the benchmark from the URL query parameters
         demo.load(_set_benchmark_on_load, inputs=[], outputs=[benchmark_select])
 
-        @cachetools.cached(
-            cache={},
-            key=lambda benchmark_name: hash(benchmark_name),
-        )
         def on_benchmark_select(benchmark_name):
-            start_time = time.time()
-            benchmark = mteb.get_benchmark(benchmark_name)
-            languages = [task.languages for task in benchmark.tasks if task.languages]
-            languages = set(itertools.chain.from_iterable(languages))
-            languages = sorted(languages)
-            domains = [
-                task.metadata.domains
-                for task in benchmark.tasks
-                if task.metadata.domains
-            ]
-            domains = set(itertools.chain.from_iterable(domains))
-            types = {
-                task.metadata.type for task in benchmark.tasks if task.metadata.type
-            }
-            modalities = set()
-            for task in benchmark.tasks:
-                modalities.update(task.metadata.modalities)
-            languages, domains, types, modalities = (
-                sorted(languages),
-                sorted(domains),
-                sorted(types),
-                sorted(modalities),
-            )
-            elapsed = time.time() - start_time
-            benchmark_results = all_benchmark_results[benchmark_name]
-            scores = benchmark_results._get_scores(format="long")
-            logger.debug(f"on_benchmark_select callback: {elapsed}s")
-            show_zero_shot = _should_show_zero_shot_filter(benchmark_name)
-
-            # Calculate initial models for this benchmark to avoid race conditions
-            benchmark_tasks = sorted([task.metadata.name for task in benchmark.tasks])
-            all_models_in_scores = list({entry["model_name"] for entry in scores})
-            initial_models = _filter_models(
-                all_models_in_scores,
-                benchmark_tasks,
-                availability=None,
-                compatibility=[],
-                instructions=None,
-                max_model_size=MAX_MODEL_SIZE,
-                zero_shot_setting="allow_all",
-            )
-            # Sort to ensure consistency with update_models
-            initial_models = sorted(initial_models)
-
-            return (
+            (
                 languages,
                 domains,
                 types,
                 modalities,
                 benchmark_tasks,
                 scores,
+                show_zero_shot,
+                initial_models,
+            ) = _cache_on_benchmark_select(benchmark_name, all_benchmark_results)
+
+            return (
+                gr.update(choices=languages, value=languages),
+                gr.update(choices=domains, value=domains),
+                gr.update(choices=types, value=types),
+                gr.update(choices=modalities, value=modalities),
+                gr.update(choices=benchmark_tasks, value=benchmark_tasks),
+                scores,
                 gr.update(visible=show_zero_shot),
                 initial_models,
             )
@@ -562,48 +623,13 @@ def get_leaderboard_app(cache: ResultCache = ResultCache()) -> gr.Blocks:
             outputs=[scores],
         )
 
-        @cachetools.cached(
-            cache={},
-            key=lambda benchmark_name,
-            type_select,
-            domain_select,
-            lang_select,
-            modality_select: hash(
-                (
-                    hash(benchmark_name),
-                    hash(tuple(type_select)),
-                    hash(tuple(domain_select)),
-                    hash(tuple(lang_select)),
-                    hash(tuple(modality_select)),
-                )
-            ),
-        )
         def update_task_list(
             benchmark_name, type_select, domain_select, lang_select, modality_select
         ):
-            if not len(lang_select):
-                return []
-            start_time = time.time()
-            tasks_to_keep = []
-            for task in mteb.get_benchmark(benchmark_name).tasks:
-                if task.metadata.type not in type_select:
-                    continue
-                if task.metadata.domains is not None and not (
-                    set(task.metadata.domains) & set(domain_select)
-                ):
-                    continue
-                if task.languages is not None and not (
-                    set(task.languages) & set(lang_select)
-                ):
-                    continue
-                if task.metadata.modalities and not (
-                    set(task.metadata.modalities) & set(modality_select)
-                ):
-                    continue
-                tasks_to_keep.append(task.metadata.name)
-            elapsed = time.time() - start_time
-            logger.debug(f"update_task_list callback: {elapsed}s")
-            return sorted(tasks_to_keep)
+            benchmark_tasks, tasks_to_keep = _cache_update_task_list(
+                benchmark_name, type_select, domain_select, lang_select, modality_select
+            )
+            return gr.update(choices=benchmark_tasks, value=tasks_to_keep)
 
         type_select.input(
             update_task_list,
@@ -913,4 +939,15 @@ if __name__ == "__main__":
     warnings.filterwarnings("ignore", message="Couldn't get scores for .* due to .*")
 
     app = get_leaderboard_app()
-    app.launch(server_name="0.0.0.0", server_port=7860)
+
+    head = """
+    <link href="https://cdn.jsdelivr.net/npm/tailwindcss@2.2.19/dist/tailwind.min.css" rel="stylesheet">
+    """
+    app.launch(
+        server_name="0.0.0.0",
+        server_port=7860,
+        theme=gr.themes.Soft(
+            font=[gr.themes.GoogleFont("Roboto Mono"), "Arial", "sans-serif"],
+        ),
+        head=head,
+    )

mteb/leaderboard/benchmark_selector.py

@@ -75,14 +75,17 @@ GP_BENCHMARK_ENTRIES = [
                         "MTEB(kor, v1)",
                         "MTEB(nld, v1)",
                         "MTEB(pol, v1)",
-                        "MTEB(rus, v1)",
+                        "MTEB(rus, v1.1)",
                         "MTEB(fas, v2)",
                         "VN-MTEB (vie, v1)",
                     ]
                 )
                 + [
                     MenuEntry(
-                        "Other", mteb.get_benchmarks(["MTEB(eng, v1)", "MTEB(fas, v1)"])
+                        "Other",
+                        mteb.get_benchmarks(
+                            ["MTEB(eng, v1)", "MTEB(fas, v1)", "MTEB(rus, v1)"]
+                        ),
                     )
                 ],
             ),

mteb/leaderboard/table.py

@@ -204,8 +204,7 @@ def _apply_summary_table_styling(joint_table: pd.DataFrame) -> gr.DataFrame:
         pinned_columns=2,
         column_widths=column_widths,
         wrap=True,
-        show_fullscreen_button=True,
-        show_copy_button=True,
+        buttons=["copy", "fullscreen"],
         show_search="filter",
     )
 
@@ -227,7 +226,6 @@ def _apply_per_task_table_styling(per_task: pd.DataFrame) -> gr.DataFrame:
         per_task_style,
         interactive=False,
         pinned_columns=1,
-        show_fullscreen_button=True,
-        show_copy_button=True,
+        buttons=["copy", "fullscreen"],
         show_search="filter",
     )

mteb/models/model_implementations/ru_sentence_models.py

@@ -43,6 +43,10 @@ GIGA_task_prompts = {
         "query": "Given a news title, retrieve relevant news article",
         "document": "",
     },
+    "RiaNewsRetrievalHardNegatives.v2": {
+        "query": "Given a news title, retrieve relevant news article",
+        "document": "",
+    },
     "MIRACLReranking": {
         "query": "Given a question, retrieve Wikipedia passages that answer the question",
         "document": "",
@@ -51,6 +55,10 @@ GIGA_task_prompts = {
         "query": "Given a question, retrieve Wikipedia passages that answer the question",
         "document": "",
     },
+    "MIRACLRetrievalHardNegatives.v2": {
+        "query": "Given a question, retrieve Wikipedia passages that answer the question",
+        "document": "",
+    },
     "ArguAna": {
         "query": "Given a search query, retrieve passages that answer the question",
         "document": "Given a search query, retrieve passages that answer the question",
@@ -755,6 +763,7 @@ frida_prompts = {
     "SensitiveTopicsClassification": "categorize_topic: ",
     "TERRa": "categorize_entailment: ",
     "RiaNewsRetrieval": "categorize: ",
+    "RiaNewsRetrievalHardNegatives.v2": "",
 }
 
 frida_training_datasets = {

{mteb-2.3.1.dist-info → mteb-2.3.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mteb
-Version: 2.3.1
+Version: 2.3.2
 Summary: Massive Text Embedding Benchmark
 Author-email: MTEB Contributors <niklas@huggingface.co>, Kenneth Enevoldsen <kenneth.enevoldsen@cas.au.dk>, Nouamane Tazi <nouamane@huggingface.co>, Nils Reimers <info@nils-reimers.de>
 Maintainer-email: Kenneth Enevoldsen <kenneth.enevoldsen@cas.au.dk>, Roman Solomatin <risolomatin@gmail.com>, Isaac Chung <chungisaac1217@gmail.com>
@@ -37,7 +37,7 @@ Requires-Dist: torchvision>0.2.1; extra == "image"
 Provides-Extra: codecarbon
 Requires-Dist: codecarbon<3.0.0,>=2.0.0; extra == "codecarbon"
 Provides-Extra: leaderboard
-Requires-Dist: gradio==5.49.1; extra == "leaderboard"
+Requires-Dist: gradio==6.0.1; extra == "leaderboard"
 Requires-Dist: plotly<6.0.0,>=5.24.0; extra == "leaderboard"
 Requires-Dist: cachetools>=5.2.0; extra == "leaderboard"
 Requires-Dist: matplotlib>=3.9.4; extra == "leaderboard"

{mteb-2.3.1.dist-info → mteb-2.3.2.dist-info}/RECORD

@@ -55,8 +55,8 @@ mteb/benchmarks/__init__.py,sha256=MQEVeli-zLaJ7Xg0z7RhXQwsdmm7Ht_W2Ln0rZo1Szc,2
 mteb/benchmarks/_create_table.py,sha256=z3iqa5dajLk0DYxEE9EeO1qpR3VJXokg8ZQ2rdUkvdM,20452
 mteb/benchmarks/benchmark.py,sha256=70RlMyyg_wkWTlU_IbfLl-KaqRWXGCKTd8fWe9X-AQE,4173
 mteb/benchmarks/get_benchmark.py,sha256=-n_O-gitRKZi48gJKNgGuI36hsP7yLVSiwulnMHN7Gw,3935
-mteb/benchmarks/benchmarks/__init__.py,sha256=UD6YjWPDVPSQdUhmD-4rho08Gs5LU9pS_C2jX5eUns0,2102
-mteb/benchmarks/benchmarks/benchmarks.py,sha256=KDJanVYs3BkFn74VHwarZ8HJ2DX6EIgcVYBrlyjbv9I,89956
+mteb/benchmarks/benchmarks/__init__.py,sha256=0ySgD14Mu3Y1nJzazR_eUir81ia3x6E23N57SzQNkF0,2150
+mteb/benchmarks/benchmarks/benchmarks.py,sha256=Ob2cHVXwFk328xbV-2ZmUibiVAMtT2RN1ygGgiP6UNQ,92662
 mteb/benchmarks/benchmarks/rteb_benchmarks.py,sha256=QnCSrTTaBfcRlAQp2Nu81tgv1idMXqiM16Fp2zKJ5Ys,10607
 mteb/cli/__init__.py,sha256=v-csUr3eUZElIvrGB6QGtaIdndDfNWEe9oZchsGsJpg,64
 mteb/cli/_display_tasks.py,sha256=7A06dT9sSoTz6shyMvskPxuc5eHY_H7PGPlROzMP0yw,2196
@@ -1424,10 +1424,10 @@ mteb/languages/language_family.json,sha256=OUGcHeOIPcZPb2FWmYLhxTS0JxjK5y3Fo6x0P
 mteb/languages/language_scripts.py,sha256=5wix9HTYolNIpTiS5oXf2pGJyL7ftdGKs_m432w81V8,3998
 mteb/languages/programming_languages.py,sha256=zxAakT3OSUnAuTnQ34VyeFIECnNXMlleZmAake6jsZE,211
 mteb/leaderboard/__init__.py,sha256=991roXmtRwEQysV-37hWEzWpkvPgMCGRqZTHR-hm2io,88
-mteb/leaderboard/app.py,sha256=EsQ_qoJ26yJbg2qExKFFAx90R8VYOO6GbLtIzFuHGpE,32642
-mteb/leaderboard/benchmark_selector.py,sha256=hnXdo_Kj4UUAruFl6nZkCxAQ88IEfbaH8EADFJMMdVo,7686
+mteb/leaderboard/app.py,sha256=rwU3sHxx8YP3kFOvFNAF8izgBd5zgv6lrvO4mZcEmfA,33255
+mteb/leaderboard/benchmark_selector.py,sha256=uH66SI0iT1J4_fnebViWa83dQwhPi7toBv7PRL_epDw,7784
 mteb/leaderboard/figures.py,sha256=Rq20LFpaUhQD4tuKp7P7ExQtAjonMLibgO3ud0ykMag,7491
-mteb/leaderboard/table.py,sha256=qs0H_Gt9FzRvzb-AL0YlqEe0YAsdYsVX3QlncfCBEqg,7828
+mteb/leaderboard/table.py,sha256=ZBCW8JDk5gLbi06FA6zuGESQ5Xri0XZIO0uK-aWb2us,7772
 mteb/leaderboard/text_segments.py,sha256=iMIkS04QQjPbT-SkU0x6fOcS8xRbUYevryu9HydipKM,6570
 mteb/models/__init__.py,sha256=ABTuoqiBjBtBWW3LYY7ItBHdylR6jWoy06HH0g6j6fU,910
 mteb/models/abs_encoder.py,sha256=m0JkRfRPMYadDgBR9eozRloI31ZSWkSzDFINpwbfLZk,16533
@@ -1531,7 +1531,7 @@ mteb/models/model_implementations/repllama_models.py,sha256=89HoqEpzkNysHeuf_-Yh
 mteb/models/model_implementations/rerankers_custom.py,sha256=ro73A9-hHudy3_qIMrhP-ja-3Xqu78r_aORm856zHQc,10651
 mteb/models/model_implementations/rerankers_monot5_based.py,sha256=rxVwzapNnHl4gCw79XVCaTXj3-wbToyj7XVL97tpAF4,34302
 mteb/models/model_implementations/richinfoai_models.py,sha256=llvYa0JUjyOOMbuTgOYoJ2qeqZ5rLHX1ZjZIYlYbdvA,989
-mteb/models/model_implementations/ru_sentence_models.py,sha256=Dstx46xFcAOC7giKPclC41GJTtFfmg4t6gLTdAnrxDk,40129
+mteb/models/model_implementations/ru_sentence_models.py,sha256=GuZFwbzaooufvSMGNjIsL0DDLrqHjhdSsAQHHZo5H08,40480
 mteb/models/model_implementations/salesforce_models.py,sha256=KslTK-IKeLvNG-vQir9k6swkaOgjk6eyozm_BOVgTpY,5160
 mteb/models/model_implementations/samilpwc_models.py,sha256=oMwKNwCxoH1jZgCy04oo2oVlBZWu253QMpnEEC6emz8,2021
 mteb/models/model_implementations/searchmap_models.py,sha256=XvVl99emIgnNUCxkTuFQXW6py2R8vgsArfpyHveCugw,1904
@@ -2567,9 +2567,9 @@ mteb/types/_metadata.py,sha256=NN-W0S6a5TDV7UkpRx1pyWtGF4TyyCyoPUfHOwdeci8,2290
 mteb/types/_result.py,sha256=CRAUc5IvqI3_9SyXDwv-PWLCXwXdZem9RePeYESRtuw,996
 mteb/types/_string_validators.py,sha256=PY-dYq4E8O50VS3bLYdldPWp400fl_WzUjfVSkNWe8U,523
 mteb/types/statistics.py,sha256=YwJsxTf1eaCI_RE-J37a-gK5wDeGAsmkeZKoZCFihSo,3755
-mteb-2.3.1.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-mteb-2.3.1.dist-info/METADATA,sha256=AeDGGuksA6YmVR7zGXWB1jbk2mUD3w5tRCgTZjTnZ4U,13798
-mteb-2.3.1.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-mteb-2.3.1.dist-info/entry_points.txt,sha256=8IJoEJFKoDHmVnNev-qJ9pp4Ln7_1-ma9QsXnzVCzGU,39
-mteb-2.3.1.dist-info/top_level.txt,sha256=OLVIjcQAlWBz0bdmutKlWHLF42FF0hp4uVAg3ZyiG4U,5
-mteb-2.3.1.dist-info/RECORD,,
+mteb-2.3.2.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+mteb-2.3.2.dist-info/METADATA,sha256=LEbGSbNtHSdIf03wLQKaayWlIbr0sGHRfUCvlO4Voe0,13797
+mteb-2.3.2.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+mteb-2.3.2.dist-info/entry_points.txt,sha256=8IJoEJFKoDHmVnNev-qJ9pp4Ln7_1-ma9QsXnzVCzGU,39
+mteb-2.3.2.dist-info/top_level.txt,sha256=OLVIjcQAlWBz0bdmutKlWHLF42FF0hp4uVAg3ZyiG4U,5
+mteb-2.3.2.dist-info/RECORD,,