interactions: - request: body: null headers: Accept: - '*/*' Accept-Encoding: - gzip, deflate Connection: - keep-alive User-Agent: - python-requests/2.32.3 method: GET uri: https://raw.githubusercontent.com/avidale/encodechka/master/README.md response: body: string: "# encodechka\n## encodechka-eval\n\n\u042D\u0442\u043E\u0442 \u0440\u0435\u043F\u043E\u0437\u0438\u0442\u043E\u0440\u0438\u0439 - \u0440\u0430\u0437\u0432\u0438\u0442\u0438\u0435 \u043F\u043E\u0434\u0445\u043E\u0434\u0430 \u043A \u043E\u0446\u0435\u043D\u043A\u0435 \u043C\u043E\u0434\u0435\u043B\u0435\u0439 \u0438\u0437 \u043F\u043E\u0441\u0442\u0430\n[\u041C\u0430\u043B\u0435\u043D\u044C\u043A\u0438\u0439 \u0438 \u0431\u044B\u0441\u0442\u0440\u044B\u0439 BERT \u0434\u043B\u044F \u0440\u0443\u0441\u0441\u043A\u043E\u0433\u043E \u044F\u0437\u044B\u043A\u0430](https://habr.com/ru/post/562064), \n\u044D\u0432\u043E\u043B\u044E\u0446\u0438\u043E\u043D\u0438\u0440\u043E\u0432\u0430\u0432\u0448\u0435\u0433\u043E \u0432 [\u0420\u0435\u0439\u0442\u0438\u043D\u0433 \u0440\u0443\u0441\u0441\u043A\u043E\u044F\u0437\u044B\u0447\u043D\u044B\u0445 \u044D\u043D\u043A\u043E\u0434\u0435\u0440\u043E\u0432 \u043F\u0440\u0435\u0434\u043B\u043E\u0436\u0435\u043D\u0438\u0439](https://habr.com/ru/post/669674/).\n\u0418\u0434\u0435\u044F \u0432 \u0442\u043E\u043C, \u0447\u0442\u043E\u0431\u044B \u043F\u043E\u043D\u044F\u0442\u044C, \u043A\u0430\u043A \u0445\u043E\u0440\u043E\u0448\u043E \u0440\u0430\u0437\u043D\u044B\u0435 \u043C\u043E\u0434\u0435\u043B\u0438 \u043F\u0440\u0435\u0432\u0440\u0430\u0449\u0430\u044E\u0442 \u043A\u043E\u0440\u043E\u0442\u043A\u0438\u0435 \u0442\u0435\u043A\u0441\u0442\u044B\n\u0432 \u043E\u0441\u043C\u044B\u0441\u043B\u0435\u043D\u043D\u044B\u0435 \u0432\u0435\u043A\u0442\u043E\u0440\u044B.\n\n\u041F\u043E\u0445\u043E\u0436\u0438\u0435 \u043F\u0440\u043E\u0435\u043A\u0442\u044B:\n* [RussianSuperGLUE](https://russiansuperglue.com/): \u0444\u043E\u043A\u0443\u0441 \u043D\u0430 \u0434\u043E\u043E\u0431\u0443\u0447\u0430\u0435\u043C\u044B\u0445 \u043C\u043E\u0434\u0435\u043B\u044F\u0445\n* [MOROCCO](https://github.com/RussianNLP/MOROCCO/): RussianSuperGLUE + \u043E\u0446\u0435\u043D\u043A\u0430 \u043F\u0440\u043E\u0438\u0437\u0432\u043E\u0434\u0438\u0442\u0435\u043B\u044C\u043D\u043E\u0441\u0442\u0438, \u0442\u0440\u0443\u0434\u043D\u043E\u0432\u043E\u0441\u043F\u0440\u043E\u0438\u0437\u0432\u043E\u0434\u0438\u043C\n* [RuSentEval](https://github.com/RussianNLP/RuSentEval): \u0431\u043E\u043B\u0435\u0435 \u0430\u043A\u0430\u0434\u0435\u043C\u0438\u0447\u0435\u0441\u043A\u0438\u0435/\u043B\u0438\u043D\u0433\u0432\u0438\u0441\u0442\u0438\u0447\u0435\u0441\u043A\u0438\u0435 \u0437\u0430\u0434\u0430\u0447\u0438\n* \u0421\u0442\u0430\u0442\u044C\u044F \u043E\u0442 \u0412\u044B\u0448\u043A\u0438 [Popov et al, 2019](https://arxiv.org/abs/1910.13291): \u043F\u0435\u0440\u0432\u0430\u044F \u043D\u0430\u0443\u0447\u043D\u0430\u044F \u0441\u0442\u0430\u0442\u044C\u044F \u043D\u0430 \u044D\u0442\u0443 \u0442\u0435\u043C\u0443, \u043D\u043E \u043C\u0430\u043B\u043E\u0432\u0430\u0442\u043E \u043C\u043E\u0434\u0435\u043B\u0435\u0439 \u0438 \u0437\u0430\u0434\u0430\u0447\n* [SentEvalRu](https://github.com/comptechml/SentEvalRu) \u0438 [deepPavlovEval](https://github.com/deepmipt/deepPavlovEval): \u0434\u0432\u0430 \u0445\u043E\u0440\u043E\u0448\u0438\u0445, \u043D\u043E \u0434\u0430\u0432\u043D\u043E \u043D\u0435 \u043E\u0431\u043D\u043E\u0432\u043B\u044F\u0432\u0448\u0438\u0445\u0441\u044F \u0431\u0435\u043D\u0447\u043C\u0430\u0440\u043A\u0430. \n\n\u041F\u0440\u0438\u043C\u0435\u0440 \u0437\u0430\u043F\u0443\u0441\u043A\u0430 \u043C\u0435\u0442\u0440\u0438\u043A \u2013 \u0432 \u0431\u043B\u043E\u043A\u043D\u043E\u0442\u0435 [evaluation example](https://github.com/avidale/encodechka/blob/master/evaluation%20example.ipynb). \n\n\u0411\u043B\u043E\u043A\u043D\u043E\u0442 \u0434\u043B\u044F \u0432\u043E\u0441\u043F\u0440\u043E\u0438\u0437\u0432\u0435\u0434\u0435\u043D\u0438\u044F \u043B\u0438\u0434\u0435\u0440\u0431\u043E\u0440\u0434\u0430: [v2021](https://colab.research.google.com/drive/1fu2i7A-Yr-85Ex_NvIyeCIO7lN2R7P-k?usp=sharing), \n[v2023](https://colab.research.google.com/drive/1t956aJsp5qPnst3379vI8NNRqiqJUFMn?usp=sharing).\n\n### \u041B\u0438\u0434\u0435\u0440\u0431\u043E\u0440\u0434\n\n\u0420\u0430\u043D\u0436\u0438\u0440\u043E\u0432\u0430\u043D\u0438\u0435 \u043C\u043E\u0434\u0435\u043B\u0435\u0439 \u0432 \u043F\u043E \u0441\u0440\u0435\u0434\u043D\u0435\u043C\u0443 \u043A\u0430\u0447\u0435\u0441\u0442\u0432\u0443 \u0438 \u043F\u0440\u043E\u0438\u0437\u0432\u043E\u0434\u0438\u0442\u0435\u043B\u044C\u043D\u043E\u0441\u0442\u0438. \n\u041F\u043E\u0434\u0441\u0432\u0435\u0447\u0435\u043D\u044B \u041F\u0430\u0440\u0435\u0442\u043E-\u043E\u043F\u0442\u0438\u043C\u0430\u043B\u044C\u043D\u044B\u0435 \u043C\u043E\u0434\u0435\u043B\u0438 \u043F\u043E \u043A\u0430\u0436\u0434\u043E\u043C\u0443 \u0438\u0437 \u043A\u0440\u0438\u0442\u0435\u0440\u0438\u0435\u0432. \n\n| model | CPU | GPU | size | Mean S | Mean S+W | dim |\n|:------------------------------------------------------------|:----------|:---------|:--------------|---------:|:-----------|------:|\n| BAAI/bge-m3 | 523.4 | 22.5 | **2166.0** | 0.787 | 0.696 | 1024 |\n| intfloat/multilingual-e5-large-instruct \ | 501.5 | 25.71 | **2136.0** | 0.784 | 0.684 \ | 1024 |\n| intfloat/multilingual-e5-large | **506.8** | **30.8** | **2135.9389** | 0.78 | 0.686 | 1024 |\n| sentence-transformers/paraphrase-multilingual-mpnet-base-v2 | **20.5** | **19.9** | **1081.8485** | 0.762 | | 768 |\n| intfloat/multilingual-e5-base \ | 130.61 | 14.39 | **1061.0** | 0.761 | 0.669 | 768 |\n| intfloat/multilingual-e5-small | 40.86 | 12.09 | **449.0** | 0.742 | 0.645 | 384 |\n| symanto/sn-xlm-roberta-base-snli-mnli-anli-xnli | **20.2** | **16.5** | **1081.8474** | 0.739 | | 768 |\n| cointegrated/LaBSE-en-ru \ | 133.4 | **15.3** | **489.6621** \ | 0.739 | 0.668 | 768 |\n| sentence-transformers/LaBSE | 135.1 | **13.3** | 1796.5078 | 0.739 | 0.667 | 768 |\n| MUSE-3 | 200.1 | 30.7 | **303.0** | 0.736 | | 512 |\n| text-embedding-ada-002 \ | ? | | ? | \ 0.734 | | 1536 |\n| sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2 | **18.2** | 14.9 | 479.2547 | 0.734 | | 384 |\n| sentence-transformers/distiluse-base-multilingual-cased-v1 | **11.8** | **8.0** | 517.7452 | 0.722 | | 512 |\n| SONAR | ? | ? | 3060.0 | 0.721 | | 1024 |\n| facebook/nllb-200-distilled-600M | 252.3 | 15.9 | 1577.4828 | 0.709 | 0.64 | 1024 |\n| sentence-transformers/distiluse-base-multilingual-cased-v2 \ | **11.2** | 9.2 | 517.7453 | 0.708 | | 512 |\n| cointegrated/rubert-tiny2 | **6.2** | **4.6** | **111.3823** | 0.704 | 0.638 | 312 |\n| ai-forever/sbert_large_mt_nlu_ru \ | 504.5 | 29.7 | 1628.6539 | 0.703 | 0.626 | 1024 |\n| laser | 192.5 | 13.5 | 200.0 | 0.699 | | 1024 |\n| laser2 | 163.4 | 8.6 | 175.0 | 0.694 | | 1024 |\n| ai-forever/sbert_large_nlu_ru \ | 497.7 | 29.9 | 1628.6539 | 0.688 | 0.626 | 1024 |\n| clips/mfaq | 18.1 | 18.2 | 1081.8576 | 0.687 | | 768 |\n| cointegrated/rut5-base-paraphraser | 137.0 | 15.6 | 412.0015 | 0.685 | 0.634 | 768 |\n| DeepPavlov/rubert-base-cased-sentence \ | 128.4 | 13.2 | 678.5215 | 0.678 | 0.612 | 768 |\n| DeepPavlov/distilrubert-base-cased-conversational \ | 64.2 | 10.4 | 514.002 | 0.676 | 0.624 | \ 768 |\n| DeepPavlov/distilrubert-tiny-cased-conversational | 21.2 | **3.3** | 405.8292 | 0.67 | 0.616 | 768 |\n| cointegrated/rut5-base-multitask | 136.9 | 12.7 | 412.0015 | 0.668 | 0.623 | 768 |\n| ai-forever/ruRoberta-large \ | 512.3 | 25.5 | 1355.7162 | \ 0.666 | 0.609 | 1024 |\n| DeepPavlov/rubert-base-cased-conversational \ | 127.5 | 16.3 | 678.5215 | 0.653 | 0.606 \ | 768 |\n| deepvk/deberta-v1-base | 128.6 | 19.0 | 473.2402 | 0.653 | 0.591 | 768 |\n| cointegrated/rubert-tiny | 7.5 | 5.9 | **44.97** | 0.645 | 0.575 | 312 |\n| ai-forever/FRED-T5-large \ | 479.4 | 23.3 | 1372.9988 | \ 0.639 | 0.551 | 1024 |\n| inkoziev/sbert_synonymy | 6.9 | 4.2 | 111.3823 | 0.637 | 0.566 | 312 |\n| numind/NuNER-multilingual-v0.1 | 186.9 | 10 | 678.0 | 0.633 | 0.572 | 768 |\n| cointegrated/rubert-tiny-toxicity \ | 10 | 5.5 | 47.2 | 0.621 | 0.553 | 312 |\n| ft_geowac_full | **0.3** | | 1910.0 | 0.617 | 0.55 | 300 |\n| bert-base-multilingual-cased | 141.4 | 13.7 | 678.5215 | 0.614 | 0.565 | 768 |\n| ai-forever/ruT5-large \ | 489.6 | 20.2 | 1277.7571 \ | 0.61 | 0.578 | 1024 |\n| cointegrated/rut5-small | 37.6 | 8.6 | 111.3162 | 0.602 | 0.564 | 512 |\n| ft_geowac_21mb | 1.2 | \ | **21.0** | 0.597 | 0.531 | 300 |\n| inkoziev/sbert_pq \ | 7.4 | 4.2 | 111.3823 \ | 0.596 | 0.526 | 312 |\n| ai-forever/ruT5-base | 126.3 | 12.8 | 418.2325 | 0.571 | 0.544 | 768 |\n| hashing_1000_char | 0.5 | \ | **1.0** | 0.557 | 0.464 | 1000 |\n| cointegrated/rut5-base \ | 127.8 | 15.5 | 412.0014 | \ 0.554 | 0.53 | 768 |\n| hashing_300_char | 0.8 | | 1.0 | 0.529 | 0.433 | 300 |\n| hashing_1000 | **0.2** | \ | 1.0 | 0.513 | 0.416 | 1000 |\n| hashing_300 \ | 0.3 | | 1.0 | 0.491 | 0.397 | 300 |\n\n\u0420\u0430\u043D\u0436\u0438\u0440\u043E\u0432\u0430\u043D\u0438\u0435 \u043C\u043E\u0434\u0435\u043B\u0435\u0439 \u043F\u043E \u0437\u0430\u0434\u0430\u0447\u0430\u043C.\n\u041F\u043E\u0434\u0441\u0432\u0435\u0447\u0435\u043D\u044B \u043D\u0430\u0438\u043B\u0443\u0447\u0448\u0438\u0435 \u043C\u043E\u0434\u0435\u043B\u0438 \u043F\u043E \u043A\u0430\u0436\u0434\u043E\u0439 \u0438\u0437 \u0437\u0430\u0434\u0430\u0447. \n\n| model | STS | PI | NLI | SA | TI | IA | IC | ICX | NE1 | NE2 |\n|:------------------------------------------------------------|:---------|:---------|:---------|:---------|:---------|:---------|:---------|:---------|:---------|:---------|\n| BAAI/bge-m3 | **0.86** | **0.75** | 0.51 | **0.82** | 0.97 | 0.79 | 0.81 | **0.78** | 0.24 | 0.42 |\n| intfloat/multilingual-e5-large-instruct | 0.86 | 0.74 | 0.47 | 0.81 | 0.98 | 0.8 | **0.82** | 0.77 | 0.21 | 0.35 |\n| intfloat/multilingual-e5-large | 0.86 | 0.73 | 0.47 | 0.81 | 0.98 | 0.8 | 0.82 | 0.77 | 0.24 | 0.37 |\n| sentence-transformers/paraphrase-multilingual-mpnet-base-v2 | 0.85 | 0.66 | 0.54 | 0.79 | 0.95 | 0.78 | 0.79 | 0.74 | | |\n| intfloat/multilingual-e5-base | 0.83 | 0.7 | 0.46 | 0.8 | 0.96 | 0.78 | 0.8 | 0.74 | 0.23 | 0.38 |\n| intfloat/multilingual-e5-small | 0.82 | 0.71 | 0.46 | 0.76 | 0.96 | 0.76 | 0.78 | 0.69 | 0.23 | 0.27 |\n| symanto/sn-xlm-roberta-base-snli-mnli-anli-xnli \ | 0.76 | 0.6 | **0.86** | 0.76 | 0.91 | 0.72 \ | 0.71 | 0.6 | | |\n| cointegrated/LaBSE-en-ru \ | 0.79 | 0.66 | 0.43 | 0.76 \ | 0.95 | 0.77 | 0.79 | 0.77 | 0.35 | 0.42 |\n| sentence-transformers/LaBSE | 0.79 | 0.66 \ | 0.43 | 0.76 | 0.95 | 0.77 | 0.79 | 0.76 | 0.35 \ | 0.41 |\n| MUSE-3 | 0.81 | 0.61 | 0.42 | 0.77 | 0.96 | 0.79 | 0.77 | 0.75 | | |\n| text-embedding-ada-002 | 0.78 | 0.66 | 0.44 | 0.77 | 0.96 | 0.77 | 0.75 | 0.73 | | |\n| sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2 | 0.84 | 0.62 | 0.5 | 0.76 | 0.92 | 0.74 | 0.77 | 0.72 | | |\n| sentence-transformers/distiluse-base-multilingual-cased-v1 \ | 0.8 | 0.6 | 0.43 | 0.75 | 0.94 | 0.76 | 0.76 \ | 0.74 | | |\n| SONAR | 0.71 | 0.58 | 0.41 | 0.77 | 0.98 | 0.79 | 0.78 | 0.74 | | |\n| facebook/nllb-200-distilled-600M | 0.71 | 0.54 | 0.41 | 0.76 | 0.95 | 0.76 | 0.8 | 0.75 | 0.31 | 0.42 |\n| sentence-transformers/distiluse-base-multilingual-cased-v2 \ | 0.79 | 0.55 | 0.42 | 0.75 | 0.91 | 0.75 | 0.76 \ | 0.73 | | |\n| cointegrated/rubert-tiny2 | 0.75 | 0.65 | 0.42 | 0.74 | 0.94 | 0.75 | 0.76 | 0.64 | 0.36 | 0.39 |\n| ai-forever/sbert_large_mt_nlu_ru | 0.78 | 0.65 | 0.4 | 0.8 | 0.98 | 0.8 | 0.76 | 0.45 | 0.3 | 0.34 |\n| laser | 0.75 | 0.6 | 0.41 | 0.73 | 0.96 | 0.72 | 0.72 | 0.7 | | |\n| laser2 | 0.74 | 0.6 | 0.41 | 0.73 | 0.95 | 0.72 | 0.72 | 0.69 | | |\n| ai-forever/sbert_large_nlu_ru | 0.68 | 0.62 | 0.39 | 0.78 | 0.98 | 0.8 | 0.78 | 0.48 | 0.36 | 0.4 |\n| clips/mfaq | 0.63 | 0.59 | 0.35 | 0.79 | 0.95 | 0.74 | 0.76 | 0.69 | | |\n| cointegrated/rut5-base-paraphraser | 0.65 | 0.53 | 0.4 | 0.78 | 0.95 | 0.75 | 0.75 | 0.67 | 0.45 | 0.41 |\n| DeepPavlov/rubert-base-cased-sentence \ | 0.74 | 0.66 | 0.49 | 0.75 | 0.92 \ | 0.75 | 0.72 | 0.39 | 0.36 | 0.34 |\n| DeepPavlov/distilrubert-base-cased-conversational \ | 0.7 | 0.56 | 0.39 | 0.76 | 0.98 | 0.78 | 0.76 | 0.48 | 0.4 | 0.43 |\n| DeepPavlov/distilrubert-tiny-cased-conversational \ | 0.7 | 0.55 | 0.4 | 0.74 | 0.98 | 0.78 | 0.76 | 0.45 | 0.35 | 0.44 |\n| cointegrated/rut5-base-multitask \ | 0.65 | 0.54 | 0.38 | 0.76 | 0.95 | 0.75 | 0.72 | 0.59 | 0.47 | 0.41 |\n| ai-forever/ruRoberta-large \ | 0.7 | 0.6 | 0.35 | 0.78 | 0.98 | 0.8 | 0.78 | 0.32 | 0.3 | **0.46** |\n| DeepPavlov/rubert-base-cased-conversational \ | 0.68 | 0.52 | 0.38 | 0.73 | 0.98 | 0.78 | 0.75 | 0.42 | 0.41 | 0.43 |\n| deepvk/deberta-v1-base \ | 0.68 | 0.54 | 0.38 | 0.76 \ | 0.98 | 0.8 | 0.78 | 0.29 | 0.29 | 0.4 |\n| cointegrated/rubert-tiny | 0.66 | 0.53 \ | 0.4 | 0.71 | 0.89 | 0.68 | 0.7 | 0.58 | 0.24 \ | 0.34 |\n| ai-forever/FRED-T5-large | 0.62 | 0.44 | 0.37 | 0.78 | 0.98 | **0.81** | 0.67 | 0.45 | 0.25 | 0.15 |\n| inkoziev/sbert_synonymy | 0.69 | 0.49 | 0.41 | 0.71 | 0.91 | 0.72 | 0.69 | 0.47 | 0.32 | 0.24 |\n| numind/NuNER-multilingual-v0.1 | 0.67 | 0.53 | 0.4 | 0.71 | 0.89 | 0.72 | 0.7 | 0.46 | 0.32 | 0.34 |\n| cointegrated/rubert-tiny-toxicity | 0.57 | 0.44 | 0.37 | 0.68 | **1.0** | 0.78 | 0.7 | 0.43 | 0.24 | 0.32 |\n| ft_geowac_full | 0.69 | 0.53 | 0.37 | 0.72 | 0.97 | 0.76 | 0.66 | 0.26 | 0.22 | 0.34 |\n| bert-base-multilingual-cased | 0.66 | 0.53 | 0.37 | 0.7 | 0.89 | 0.7 | 0.69 | 0.38 | 0.36 | 0.38 |\n| ai-forever/ruT5-large | 0.51 | 0.39 | 0.35 | 0.77 | 0.97 | 0.79 | 0.72 | 0.38 | 0.46 | 0.44 |\n| cointegrated/rut5-small | 0.61 | 0.53 | 0.34 | 0.73 | 0.92 | 0.71 | 0.7 | 0.27 | 0.44 | 0.38 |\n| ft_geowac_21mb | 0.68 | 0.52 | 0.36 | 0.72 | 0.96 | 0.74 | 0.65 | 0.15 | 0.21 | 0.32 |\n| inkoziev/sbert_pq | 0.57 | 0.41 | 0.38 | 0.7 | 0.92 | 0.69 | 0.68 | 0.43 | 0.26 | 0.24 |\n| ai-forever/ruT5-base | 0.5 | 0.28 | 0.34 | 0.73 | 0.97 | 0.76 | 0.7 | 0.29 | 0.45 | 0.41 |\n| hashing_1000_char | 0.7 | 0.53 | 0.4 | 0.7 | 0.84 | 0.59 | 0.63 | 0.05 | 0.05 | 0.14 |\n| cointegrated/rut5-base | 0.44 | 0.28 | 0.33 | 0.74 | 0.92 | 0.75 | 0.58 | 0.39 | **0.48** | 0.39 |\n| hashing_300_char | 0.69 | 0.51 | 0.39 | 0.67 | 0.75 | 0.57 | 0.61 | 0.04 | 0.03 | 0.08 |\n| hashing_1000 | 0.63 | 0.49 | 0.39 | 0.66 | 0.77 | 0.55 | 0.57 | 0.05 | 0.02 | 0.04 |\n| hashing_300 | 0.61 | 0.48 | 0.4 | 0.64 | 0.71 | 0.54 | 0.5 | 0.05 | 0.02 | 0.02 |\n\n#### \u0417\u0430\u0434\u0430\u0447\u0438\n- Semantic text similarity (**STS**) \u043D\u0430 \u043E\u0441\u043D\u043E\u0432\u0435 \u043F\u0435\u0440\u0435\u0432\u0435\u0434\u0451\u043D\u043D\u043E\u0433\u043E \u0434\u0430\u0442\u0430\u0441\u0435\u0442\u0430 [STS-B](https://huggingface.co/datasets/stsb_multi_mt);\n- Paraphrase identification (**PI**) \u043D\u0430 \u043E\u0441\u043D\u043E\u0432\u0435 \u0434\u0430\u0442\u0430\u0441\u0435\u0442\u0430 paraphraser.ru;\n- Natural language inference (**NLI**) \u043D\u0430 \u0434\u0430\u0442\u0430\u0441\u0435\u0442\u0435 [XNLI](https://github.com/facebookresearch/XNLI);\n- Sentiment analysis (**SA**) \u043D\u0430 \u0434\u0430\u043D\u043D\u044B\u0445 [SentiRuEval2016](http://www.dialog-21.ru/evaluation/2016/sentiment/).\n- Toxicity identification (**TI**) \u043D\u0430 \u0434\u0430\u0442\u0430\u0441\u0435\u0442\u0435 \u0442\u043E\u043A\u0441\u0438\u0447\u043D\u044B\u0445 \u043A\u043E\u043C\u043C\u0435\u043D\u0442\u0430\u0440\u0438\u0435\u0432 \u0438\u0437 [OKMLCup](https://cups.mail.ru/ru/contests/okmlcup2020);\n- Inappropriateness identification (**II**) \u043D\u0430 [\u0434\u0430\u0442\u0430\u0441\u0435\u0442\u0435 \u0421\u043A\u043E\u043B\u0442\u0435\u0445\u0430](https://github.com/skoltech-nlp/inappropriate-sensitive-topics);\n- Intent classification (**IC**) \u0438 \u0435\u0451 \u043A\u0440\u043E\u0441\u0441-\u044F\u0437\u044B\u0447\u043D\u0430\u044F \u0432\u0435\u0440\u0441\u0438\u044F **ICX** \u043D\u0430 \u0434\u0430\u0442\u0430\u0441\u0435\u0442\u0435 [NLU-evaluation-data](https://github.com/xliuhw/NLU-Evaluation-Data), \u043A\u043E\u0442\u043E\u0440\u044B\u0439 \u044F \u0430\u0432\u0442\u043E\u043C\u0430\u0442\u0438\u0447\u0435\u0441\u043A\u0438 \u043F\u0435\u0440\u0435\u0432\u0451\u043B \u043D\u0430 \u0440\u0443\u0441\u0441\u043A\u0438\u0439. \u0412 IC \u043A\u043B\u0430\u0441\u0441\u0438\u0444\u0438\u043A\u0430\u0442\u043E\u0440 \u043E\u0431\u0443\u0447\u0430\u0435\u0442\u0441\u044F \u043D\u0430 \u0440\u0443\u0441\u0441\u043A\u0438\u0445 \u0434\u0430\u043D\u043D\u044B\u0445, \u0430 \u0432 ICX \u2013 \u043D\u0430 \u0430\u043D\u0433\u043B\u0438\u0439\u0441\u043A\u0438\u0445, \u0430 \u0442\u0435\u0441\u0442\u0438\u0440\u0443\u0435\u0442\u0441\u044F \u0432 \u043E\u0431\u043E\u0438\u0445 \u0441\u043B\u0443\u0447\u0430\u044F\u0445 \u043D\u0430 \u0440\u0443\u0441\u0441\u043A\u0438\u0445.\n- \u0420\u0430\u0441\u043F\u043E\u0437\u043D\u0430\u0432\u0430\u043D\u0438\u0435 \u0438\u043C\u0435\u043D\u043E\u0432\u0430\u043D\u043D\u044B\u0445 \u0441\u0443\u0449\u043D\u043E\u0441\u0442\u0435\u0439 \u043D\u0430 \u0434\u0430\u0442\u0430\u0441\u0435\u0442\u0430\u0445 [factRuEval-2016](https://github.com/dialogue-evaluation/factRuEval-2016) (**NE1**) \u0438 [RuDReC](https://github.com/cimm-kzn/RuDReC) (**NE2**). \u042D\u0442\u0438 \u0434\u0432\u0435 \u0437\u0430\u0434\u0430\u0447\u0438 \u0442\u0440\u0435\u0431\u0443\u044E\u0442 \u043F\u043E\u043B\u0443\u0447\u0430\u0442\u044C \u044D\u043C\u0431\u0435\u0434\u0434\u0438\u043D\u0433\u0438 \u043E\u0442\u0434\u0435\u043B\u044C\u043D\u044B\u0445 \u0442\u043E\u043A\u0435\u043D\u043E\u0432, \u0430 \u043D\u0435 \u0446\u0435\u043B\u044B\u0445 \u043F\u0440\u0435\u0434\u043B\u043E\u0436\u0435\u043D\u0438\u0439; \u043F\u043E\u044D\u0442\u043E\u043C\u0443 \u0442\u0430\u043C \u0443\u0447\u0430\u0441\u0442\u0432\u0443\u044E\u0442 \u043D\u0435 \u0432\u0441\u0435 \u043C\u043E\u0434\u0435\u043B\u0438.\n\n### Changelog\n* \u0410\u0432\u0433\u0443\u0441\u0442 2023 - \u043E\u0431\u043D\u043E\u0432\u0438\u043B \u0440\u0435\u0439\u0442\u0438\u043D\u0433:\n * \u043F\u043E\u043F\u0440\u0430\u0432\u0438\u0432 \u043E\u0448\u0438\u0431\u043A\u0443 \u0432 \u0432\u044B\u0447\u0438\u0441\u043B\u0435\u043D\u0438\u0438 mean token embeddings\n * \u0434\u043E\u0431\u0430\u0432\u0438\u043B \u043D\u0435\u0441\u043A\u043E\u043B\u044C\u043A\u043E \u043C\u043E\u0434\u0435\u043B\u0435\u0439, \u0432\u043A\u043B\u044E\u0447\u0430\u044F \u043D\u043E\u0432\u043E\u0433\u043E \u043B\u0438\u0434\u0435\u0440\u0430 - `intfloat/multilingual-e5-large`\n * \u043F\u043E \u043F\u0440\u043E\u0441\u044C\u0431\u0430\u043C \u0442\u0440\u0443\u0434\u044F\u0449\u0438\u0445\u0441\u044F, \u0434\u043E\u0431\u0430\u0432\u0438\u043B `text-embedding-ada-002` (\u0440\u0430\u0437\u043C\u0435\u0440 \u0438 \u043F\u0440\u043E\u0438\u0437\u0432\u043E\u0434\u0438\u0442\u0435\u043B\u044C\u043D\u043E\u0441\u0442\u044C \u0443\u043A\u0430\u0437\u0430\u043D\u044B \u043E\u0442 \u0431\u0430\u043B\u0434\u044B)\n* \u041B\u0435\u0442\u043E 2022 - \u043E\u043F\u0443\u0431\u043B\u0438\u043A\u043E\u0432\u0430\u043B \u043F\u0435\u0440\u0432\u044B\u0439 \u0440\u0435\u0439\u0442\u0438\u043D\u0433\n" headers: Accept-Ranges: - bytes Access-Control-Allow-Origin: - '*' Cache-Control: - max-age=300 Connection: - keep-alive Content-Encoding: - gzip Content-Length: - '4972' Content-Security-Policy: - default-src 'none'; style-src 'unsafe-inline'; sandbox Content-Type: - text/plain; charset=utf-8 Cross-Origin-Resource-Policy: - cross-origin Date: - Thu, 13 Jun 2024 17:29:26 GMT ETag: - W/"6ef42cd6939559c9e297cd85ab8b8a44b6ce19809ce92e1efcf39d06809cd99a" Expires: - Thu, 13 Jun 2024 17:34:26 GMT Source-Age: - '245' Strict-Transport-Security: - max-age=31536000 Vary: - Authorization,Accept-Encoding,Origin Via: - 1.1 varnish X-Cache: - HIT X-Cache-Hits: - '0' X-Content-Type-Options: - nosniff X-Fastly-Request-ID: - 0b5812cb6e8627abe030f2ff2764205ee7247b21 X-Frame-Options: - deny X-GitHub-Request-Id: - 3467:253C76:A903D8:B1E9A7:666B25FA X-Served-By: - cache-ams21038-AMS X-Timer: - S1718299767.633243,VS0,VE2 X-XSS-Protection: - 1; mode=block status: code: 200 message: OK version: 1