---
title: "Hypothetical Document Embeddings (HyDE) (BG)"
source: "https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)"
wiki: "systems-analysis.info/int"
article: "Hypothetical_Document_Embeddings_(HyDE)_(BG)"
language: "bg"
categories:
  - "Category:Bulgarian"
  - "Category:Large language models"
  - "Category:Pages with math errors"
  - "Category:Pages with math render errors"
  - "Category:Prompt engineering"
revision_id: 3105
wiki_created_at: 2026-09-06T23:15:46Z
wiki_modified_at: 2026-09-06T23:15:46Z
downloaded_at: 2026-09-07T22:55:10Z
---

# Hypothetical Document Embeddings (HyDE) (BG)

**Hypothetical Document Expansion (HyDE)** — метод за подобряване на векторното извличане и retrieval‑augmented generation (RAG), при който голям езиков модел (LLM) генерира „хипотетичен документ" по зададена заявка; след това този текст се векторизира от encoder, а търсенето се извършва сред реални документи по близост до получения вектор. Подходът позволява да се използват „шаблони на релевантност", кодирани от LLM, и да се „заземят" върху корпуса с помощта на плътни embedding-и<sup>[\[1\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-1)</sup>.\n\n== Определение и интуиция ==\nHyDE декомпозира задачата за търсене на два етапа:\n\n(1) LLM създава „пример на релевантен отговор" (*hypothetical document*) към заявката, като по този начин моделира признаците на релевантност;\n\n(2) контрастивен encoder (напр. Contriever) превежда този текст във вектор, по който се извличат реални документи от индекса. Генерираният текст може да съдържа фактически грешки, но важни са тематичните и терминологичните шаблони, улавяни от encoder-а<sup>[\[2\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-2)</sup>.\n\n== История и източници ==\nИдеята за разширяване на търсенето със синтетични текстове води началото си от работи по разширяване на заявки и псевдорелевантна обратна връзка (PRF): алгоритъмът на Рокио и езикови модели на релевантност<sup>[\[3\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-3)[\[4\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-4)</sup>. За плътно извличане са използвани контрастивно обучени encoder-и (Contriever)<sup>[\[5\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-5)</sup> и Dense Passage Retrieval (DPR)<sup>[\[6\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-6)</sup>. Бенчмаркът BEIR стандартизира zero‑shot оценката<sup>[\[7\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-7)</sup>. На този фон е предложен HyDE като начин да се „привнесе" в нулевия режим знание за релевантност чрез LLM, без да се дообучава encoder-ът<sup>[\[8\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-8)</sup>.\n\n== Метод и формализация ==\nНека корпусът от документи е $\mathcal{D} = \{ d_{1},\ldots,d_{N}\}$, а encoder-ът на текстове $E:\text{text} \rightarrow {\mathbb{R}}^{n}$ задава векторни представяния на документите $\mathbf{v}_{d} = E(d)$. За измерване на близостта се използва косинусово сходство или скаларно произведение; важна забележка: \*\*скаларното произведение съвпада с косинусовото сходство само при единична L2‑норма на двата вектора\*\* ($\|\mathbf{u}\| = \|\mathbf{v}\| = 1$)<sup>[\[9\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-9)</sup>.\n\nHyDE предефинира представянето на заявката чрез „хипотетичен документ", генериран от LLM. Формално:\n\n$\begin{matrix}
 & \text{(1) Генерация гипотетического текста:} & & {\overset{\sim}{d}\; = \; G\!\left( q;\,{inst} \right),} \\
 & \text{(2) Эмбеддинг гипотетического текста:} & & {\mathbf{v}_{h}\; = \; E(\overset{\sim}{d}),} \\
 & \text{(3) Поиск ближайших соседей:} & & {\mathcal{R}_{k}(q)\; = \;{TopK}_{\, d \in \mathcal{D}}\; S\!\left( \mathbf{v}_{h},\mathbf{v}_{d} \right),}
\end{matrix}$\n\nкъдето $G$ — LLM с инструкция $inst$ (например: „Напиши абзац, отговарящ на въпроса …"), $S$ — мярка за сходство (косинус или IP с нормализация), а $\mathcal{R}_{k}(q)$ — множество от $k$ документа с максимално сходство<sup>[\[10\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-10)[\[11\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-11)</sup>.\n\nВ инженерната практика често се генерират \*\*няколко\*\* хипотетични текста и техните представяния се агрегират, което повишава устойчивостта:\n\n${\overset{\sim}{d}}^{(j)} = G\!\left( q;\,{inst},\xi_{j} \right),\quad\mathbf{v}_{h}\; = \;\frac{1}{m}\sum\limits_{j = 1}^{m}E\!\left( {\overset{\sim}{d}}^{(j)} \right),$\n\nкъдето $\xi_{j}$ — стохастични параметри на декодиране (напр. temperature/top‑p). Такова ансамблиране подобрява Recall при умерен ръст на латентността<sup>[\[12\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-12)</sup>.\n\n=== Базов конвейер на HyDE ===\n

    # 1) prompt(query) -> hypothetical_doc
    # 2) embed(hypothetical_doc) -> v_h
    # 3) retrieve(index, v_h, k) -> candidates
    # 4) (optional) rerank(query, candidates) -> topN
    # 5) (для RAG) stuff / map-reduce / refine на topN

\n\n\n\n=== Връзка с други методи (QE, doc2query, PRF) ===\n\* **QE (разширяване на заявка)** добавя термини към заявката; HyDE вместо това генерира цял „квази‑документ", което се съгласува по-добре с плътните encoder-и<sup>[\[13\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-13)</sup>.\n\* **doc2query / docTTTTTquery** разширяват **документите** със синтетични заявки преди индексиране<sup>[\[14\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-14)[\[15\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-15)</sup>; HyDE разширява **заявката** в реално време, без да изисква преиндексиране.\n\* **PRF** (Rocchio, Relevance LM) актуализира вектора на заявката по топ резултатите; HyDE извлича „шаблона на релевантност" директно от LLM и след това го „заземява" чрез извличане по корпуса<sup>[\[16\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-16)</sup>.\n\n== Интеграция в RAG и преранжиране ==\nВ RAG HyDE се прилага като първи етап на извличане: хипотетичен документ → embedding → k кандидата. След това се използва преранжиране: кросс-encoder-и от клас BERT<sup>[\[17\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-17)</sup> или ColBERT с късно взаимодействие<sup>[\[18\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-18)</sup>. За сливане на списъци (напр. хибрид BM25+vector) типично се прилага RRF (*reciprocal rank fusion*):\n$\operatorname{RRF}(d) = \sum\limits_{r \in \mathcal{R}}\frac{1}{k + \operatorname{rank}_{r}(d)},\qquad k \approx 60.$\nМетодът RRF стабилно повишава съвкупното качество на обединените наредби<sup>[\[19\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-19)</sup>.\n\n== Оценка на бенчмаркове (BEIR и др.) ==\nОригиналната работа оценява HyDE в нулев режим на TREC DL'19/20 (уеб търсене) и на подмножество от колекции BEIR (Scifact, ArguAna, TREC‑COVID, FiQA, DBPedia, TREC‑NEWS, Climate‑FEVER). Фрагмент от резултатите — *към 2023‑07*:\n\n{\| class=\\wikitable\\\n\|+ *TREC DL19/20 (уеб търсене)* — mAP / nDCG@10 / Recall@1k\n\|-\n! Метод !! DL19 !! DL20 !! Източник\n\|-\n\| BM25 \|\| 30.1 / 50.6 / 75.0 \|\| 28.6 / 48.0 / 78.6 \|\| <sup>[\[20\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-20)</sup>\n\|-\n\| Contriever (unsup.) \|\| 24.0 / 44.5 / 74.6 \|\| 24.0 / 42.1 / 75.4 \|\| <sup>[\[21\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-21)</sup>\n\|-\n\| **HyDE** (Contriever+LLM) \|\| **41.8 / 61.3 / 88.0** \|\| **38.2 / 57.9 / 84.4** \|\| <sup>[\[22\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-22)</sup>\n\|-\n\| DPR (ft) \|\| 36.5 / 62.2 / 76.9 \|\| 41.8 / 65.3 / 81.4 \|\| <sup>[\[23\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-23)</sup>\n\|-\n\| ANCE (ft) \|\| 37.1 / 64.5 / 75.5 \|\| 40.8 / 64.6 / 77.6 \|\| <sup>[\[24\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-24)</sup>\n\|}\n\n{\| class=\\wikitable\\\n\|+ *BEIR (подборка от набори)* — nDCG@10 / Recall@100\n\|-\n! Метод !! Scifact !! ArguAna !! TREC‑COVID !! FiQA !! DBPedia !! TREC‑NEWS !! Climate‑FEVER !! Източник\n\|-\n\| BM25 \|\| 67.9 / 92.5 \|\| 39.7 / 93.2 \|\| **59.5 / 49.8** \|\| 23.6 / 54.0 \|\| 31.8 / 46.8 \|\| 39.5 / 44.7 \|\| 16.5 / 42.5 \|\| <sup>[\[25\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-25)</sup>\n\|-\n\| Contriever \|\| 64.9 / 92.6 \|\| 37.9 / 90.1 \|\| 27.3 / 17.2 \|\| 24.5 / 56.2 \|\| 29.2 / 45.3 \|\| 34.8 / 42.3 \|\| 15.5 / 44.1 \|\| <sup>[\[26\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-26)</sup>\n\|-\n\| **HyDE** \|\| **69.1 / 96.4** \|\| **46.6 / 97.9** \|\| 59.3 / 41.4 \|\| 27.3 / 62.1 \|\| 36.8 / 47.2 \|\| **44.0 / 50.9** \|\| **22.3 / 53.0** \|\| <sup>[\[27\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-27)</sup>\n\|}\n\nHyDE също подобрява MRR@100 на многоезичните набори Mr.TyDi (sw/ko/ja/bn) спрямо mContriever<sup>[\[28\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-28)</sup>.\n\n== Практически препоръки ==\n;Кога да се прилага HyDE\n\* Нулев/преносен режим (няма релевантни етикети; домейново „несходство" с обучаващите корпуси)<sup>[\[29\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-29)</sup>.\n\* Необходимо е повишаване на Recall@k при приемлива точност — HyDE често „открива" релевантни области на векторното пространство<sup>[\[30\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-30)</sup>.\n\n;Типични настройки\n\* **LLM и prompt**: инструкция „Напиши абзац, отговарящ на въпроса …"; умерена стохастичност (напр. *temperature*≈0.7)<sup>[\[31\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-31)</sup>.\n\* **Брой хипотетични текстове**: 1–5; усредняването на embedding-ите повишава устойчивостта<sup>[\[32\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-32)</sup>.\n\* **Embedder**: (m)Contriever без дообучаване; възможно е прилагане на дообучени encoder-и (ефектът на HyDE се запазва)<sup>[\[33\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-33)</sup>.\n\* **Нормализация на embedding-ите**: L2‑норма; вътрешното произведение е еквивалентно на косинуса<sup>[\[34\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-34)</sup>.\n\* **Хибридно извличане**: BM25+vector с последващо преранжиране<sup>[\[35\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-35)</sup>.\n\* **Преранжиращ модел**: Cross-Encoder (BERT re‑ranker)<sup>[\[36\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-36)</sup> или ColBERT<sup>[\[37\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-37)</sup>.\n\* **Сливане** на резултати от различни стратегии: RRF (*k*≈60)<sup>[\[38\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-38)</sup>.\n\n;Мониторинг на качество/разход\n\* Извличане: nDCG@k, Recall@k, MRR; end‑to‑end RAG: EM/F1 или метрики *groundedness* (RAGAS/TruLens)<sup>[\[39\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-39)[\[40\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-40)</sup>.\n\* Разход/латентност: доминира генерацията от LLM и (ако е налично) преранжирането; оптимизира се с броя „хипотетики" и дължината на отговора<sup>[\[41\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-41)</sup>.\n\n== Ограничения и отворени въпроси ==\n\* **Халюцинации** в хипотетичния текст: LLM може да въвежда фактически грешки; „заземяването" чрез encoder и корпус намалява риска, но не го елиминира напълно<sup>[\[42\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-42)</sup>.\n\* **Домейнови/езикови ограничения**: ползата от HyDE намалява в тясноспециализирани домейни и при езици с малко ресурси<sup>[\[43\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-43)</sup>.\n\* **Латентност и разход**: генерацията от LLM добавя закъснение и разход на токени; критично за онлайн сценарии и дълги „хипотетики"<sup>[\[44\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-44)</sup>.\n\* **Етика и отклонения**: препоръчително е използването на безопасни LLM и филтриране<sup>[\[45\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-45)</sup>.\n\n== Сравнителна таблица на методите ==\n{\| class=\\wikitable\\\n\|+ Съпоставяне на HyDE и сродни подходи\n\|-\n! Метод !! Клас !! Къде се генерира текстът !! Encoder/индекс !! Преранжиращ модел (2‑ри етап) !! Типични метрики (пример) !! Разход/латентност !! Източници\n\|-\n\| **HyDE** \|\| Query→*hypo‑doc* \|\| На страната на заявката (LLM → абзац) \|\| (m)Contriever; ANN \|\| BERT re‑rank / ColBERT / RRF \|\| DL19 nDCG@10≈61.3; DL20≈57.9; ArguAna nDCG@10≈46.6 \|\| + генерация от LLM; + преранжиране (опц.) \|\| <sup>[\[46\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-46)</sup>\n\|-\n\| BM25 \|\| Лексикален \|\| — \|\| Инвертиран индекс \|\| По избор \|\| вж. табл. (по-горе) \|\| Нисък (lexical) \|\| <sup>[\[47\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-47)</sup>\n\|-\n\| DPR / ANCE \|\| Плътен (ft) \|\| — \|\| Bi‑encoder; ANN \|\| По избор \|\| DL19 nDCG@10≈62–65 \|\| Среден (без LLM) \|\| <sup>[\[48\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-48)[\[49\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-49)</sup>\n\|-\n\| doc2query / docTTTTTquery \|\| Разширяване на документи \|\| На страната на колекцията (преди индексиране) \|\| BM25/sparse+expanded \|\| По избор \|\| Подобрения на BM25 при MS MARCO \|\| Висока офлайн генерация; бърз онлайн \|\| <sup>[\[50\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-50)[\[51\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-51)</sup>\n\|-\n\| PRF (Rocchio, RLM) \|\| QE по обратна връзка \|\| Заявка (по топ резултати) \|\| Произволен \|\| По избор \|\| Ръст на Recall/рискове от отклонение \|\| + допълнителен проход на извличане \|\| <sup>[\[52\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_note-52)</sup>\n\|}\n\n== Вижте също ==\n\* BM25\n\* Търсене по векторни представяния,\n\* RAG\n\* Псевдорелевантна обратна връзка\n\* BEIR\n\n== Литература ==\n\* Manning, C. D.; Raghavan, P.; Schütze, H. (2008). *Introduction to Information Retrieval*. Cambridge University Press. ISBN 978‑0521865715.\n\* Robertson, S.; Zaragoza, H. (2009). *The Probabilistic Relevance Framework: BM25 and Beyond*. Foundations and Trends in IR, 3(4), 333–389. DOI:10.1561/1500000019.\n\n== Препратки ==\n\* Хранилище на HyDE: github.com/texttron/hyde.\n\* Документация: Haystack — HyDE: docs.haystack.deepset.ai.\n\* Документация: LangChain — HyDE Retriever: docs.langchain.com.\n\n== Бележки ==\n\n\n\n<a href="https://systems-analysis.info/int/index.php?title=Template:SEOMeta%5Cn&amp;action=edit&amp;redlink=1" class="new" title="Template:SEOMeta\n (page does not exist)">Template:SEOMeta\n</a>

1.  <span id="cite_note-1">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-1) Gao, L.; Ma, X.; Lin, J.; Callan, J. (2023). ‘‘Precise Zero‑Shot Dense Retrieval without Relevance Labels (HyDE)’’. ACL 2023. pp. 1762–1777. DOI:10.18653/v1/2023.acl-long.99. <a href="https://arxiv.org/abs/2212.10496" class="external text" rel="nofollow">arXiv:2212.10496</a></span>
2.  <span id="cite_note-2">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-2) Gao, L. et al. (2023). ACL 2023, §3.2. DOI:10.18653/v1/2023.acl-long.99.</span>
3.  <span id="cite_note-3">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-3) Rocchio, J. (1971). ‘‘Relevance Feedback in Information Retrieval’’. In: Salton, G. (ed.) *The SMART Retrieval System*. Prentice‑Hall, pp. 313–323. ISBN 978‑0138145255.</span>
4.  <span id="cite_note-4">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-4) Lavrenko, V.; Croft, W. B. (2001). ‘‘Relevance‑Based Language Models’’. SIGIR. DOI:10.1145/383952.383972.</span>
5.  <span id="cite_note-5">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-5) Izacard, G. et al. (2021/2022). ‘‘Unsupervised Dense Information Retrieval with Contrastive Learning’’. <a href="https://arxiv.org/abs/2112.09118" class="external text" rel="nofollow">arXiv:2112.09118</a>.</span>
6.  <span id="cite_note-6">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-6) Karpukhin, V. et al. (2020). ‘‘Dense Passage Retrieval for Open‑Domain QA’’. EMNLP. DOI:10.18653/v1/2020.emnlp-main.550.</span>
7.  <span id="cite_note-7">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-7) Thakur, N. et al. (2021). ‘‘BEIR: A Heterogeneous Benchmark for Zero‑shot Evaluation of Information Retrieval Models’’. NeurIPS Datasets Track. <a href="https://arxiv.org/abs/2104.08663" class="external text" rel="nofollow">arXiv:2104.08663</a>.</span>
8.  <span id="cite_note-8">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-8) Gao, L. et al. (2023). DOI:10.18653/v1/2023.acl-long.99.</span>
9.  <span id="cite_note-9">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-9) Milvus Docs. ‘‘Similarity Metrics’’ — при L2‑нормализации векторов внутр. произведение эквивалентно косинусу. URL: <a href="https://milvus.io/docs/v2.2.x/metric.md" class="external free" rel="nofollow">https://milvus.io/docs/v2.2.x/metric.md</a></span>
10. <span id="cite_note-10">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-10) Gao, L.; Ma, X.; Lin, J.; Callan, J. (2023). ‘‘Precise Zero‑Shot Dense Retrieval without Relevance Labels (HyDE)’’. ACL 2023, §3–4. arXiv:2212.10496. DOI:10.18653/v1/2023.acl-long.99.</span>
11. <span id="cite_note-11">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-11) Izacard, G. et al. (2021/2022). ‘‘Unsupervised Dense Information Retrieval with Contrastive Learning (Contriever)’’. arXiv:2112.09118.</span>
12. <span id="cite_note-12">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-12) Gao, L. et al. (2023). Прил. (ablation): влияние числа гипотетических текстов и параметров генерации. arXiv:2212.10496.</span>
13. <span id="cite_note-13">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-13) Gao, L. et al. (2023). DOI:10.18653/v1/2023.acl-long.99.</span>
14. <span id="cite_note-14">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-14) Nogueira, R. et al. (2019). ‘‘Document Expansion by Query Prediction’’ (doc2query). <a href="https://arxiv.org/abs/1904.08375" class="external text" rel="nofollow">arXiv:1904.08375</a>.</span>
15. <span id="cite_note-15">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-15) Nogueira, R.; Lin, J. (2019). ‘‘From doc2query to docTTTTTquery’’ (tech report). <a href="https://cs.uwaterloo.ca/~jimmylin/publications/Nogueira_Lin_2019_docTTTTTquery-v2.pdf" class="external text" rel="nofollow">PDF</a></span>
16. <span id="cite_note-16">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-16) Rocchio, J. (1971); Lavrenko & Croft (2001), см. выше.</span>
17. <span id="cite_note-17">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-17) Nogueira, R.; Cho, K. (2019). ‘‘Passage Re‑ranking with BERT’’. <a href="https://arxiv.org/abs/1901.04085" class="external text" rel="nofollow">arXiv:1901.04085</a>.</span>
18. <span id="cite_note-18">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-18) Khattab, O.; Zaharia, M. (2020). ‘‘ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT’’. SIGIR. DOI:10.1145/3397271.3401075; <a href="https://arxiv.org/abs/2004.12832" class="external text" rel="nofollow">arXiv:2004.12832</a>.</span>
19. <span id="cite_note-19">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-19) Cormack, G. V.; Clarke, C. L. A.; Büttcher, S. (2009). ‘‘Reciprocal Rank Fusion Outperforms Condorcet and Nearly Optimally Combines Rankings’’. SIGIR. DOI:10.1145/1571941.1572114.</span>
20. <span id="cite_note-20">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-20) Gao, L. et al. (2023). Табл. 1. DOI:10.18653/v1/2023.acl-long.99.</span>
21. <span id="cite_note-21">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-21) Izacard, G. et al. (2022); сводные метрики — в Gao et al., 2023, табл. 1. arXiv:2112.09118.</span>
22. <span id="cite_note-22">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-22) Gao, L. et al. (2023). Табл. 1.</span>
23. <span id="cite_note-23">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-23) Karpukhin, V. et al. (2020); сводные — в Gao et al., 2023.</span>
24. <span id="cite_note-24">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-24) Xiong, L. et al. (2021). ICLR. <a href="https://arxiv.org/abs/2007.00808" class="external text" rel="nofollow">arXiv:2007.00808</a>.</span>
25. <span id="cite_note-25">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-25) Thakur, N. et al. (2021); сводные — в Gao et al., 2023, табл. 2. arXiv:2104.08663.</span>
26. <span id="cite_note-26">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-26) Izacard, G. et al. (2022); сводные — в Gao et al., 2023, табл. 2.</span>
27. <span id="cite_note-27">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-27) Gao, L. et al. (2023). Табл. 2.</span>
28. <span id="cite_note-28">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-28) Gao, L. et al. (2023). Табл. 3. DOI:10.18653/v1/2023.acl-long.99.</span>
29. <span id="cite_note-29">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-29) Gao, L. et al. (2023). §4–5.</span>
30. <span id="cite_note-30">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-30) Gao, L. et al. (2023). §4.2–4.3.</span>
31. <span id="cite_note-31">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-31) Gao, L. et al. (2023). §4.1.</span>
32. <span id="cite_note-32">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-32) Haystack Docs. ‘‘Hypothetical Document Embeddings (HyDE)’’ (инженерная справка). <a href="https://docs.haystack.deepset.ai/docs/hypothetical-document-embeddings-hyde" class="external text" rel="nofollow">docs.haystack.deepset.ai</a></span>
33. <span id="cite_note-33">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-33) Gao, L. et al. (2023). Табл. 6.</span>
34. <span id="cite_note-34">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-34) Milvus Docs. ‘‘Similarity Metrics’’.</span>
35. <span id="cite_note-35">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-35) Haystack × Milvus Integration (официальная док.). <a href="https://haystack.deepset.ai/integrations/milvus-document-store" class="external text" rel="nofollow">haystack.deepset.ai</a></span>
36. <span id="cite_note-36">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-36) Nogueira, R.; Cho, K. (2019). arXiv:1901.04085.</span>
37. <span id="cite_note-37">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-37) Khattab, O.; Zaharia, M. (2020). DOI:10.1145/3397271.3401075.</span>
38. <span id="cite_note-38">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-38) Cormack, G. V. et al. (2009). DOI:10.1145/1571941.1572114.</span>
39. <span id="cite_note-39">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-39) Manning, C. D.; Raghavan, P.; Schütze, H. (2008). *Introduction to Information Retrieval*. Cambridge Univ. Press. ISBN 978‑0521865715.</span>
40. <span id="cite_note-40">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-40) Es, S. et al. (2023). ‘‘RAGAS: Automated Evaluation of Retrieval‑Augmented Generation’’. <a href="https://arxiv.org/abs/2309.15217" class="external text" rel="nofollow">arXiv:2309.15217</a>.</span>
41. <span id="cite_note-41">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-41) Gao, L. et al. (2023). §5.</span>
42. <span id="cite_note-42">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-42) Gao, L. et al. (2023). §3.2; §4.1. DOI:10.18653/v1/2023.acl-long.99.</span>
43. <span id="cite_note-43">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-43) Gao, L. et al. (2023). Табл. 3; §4.4.</span>
44. <span id="cite_note-44">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-44) Gao, L. et al. (2023). §4–5.</span>
45. <span id="cite_note-45">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-45) Ouyang, L. et al. (2022). ‘‘Training language models to follow instructions with human feedback (InstructGPT)’’. NeurIPS. <a href="https://arxiv.org/abs/2203.02155" class="external text" rel="nofollow">arXiv:2203.02155</a>.</span>
46. <span id="cite_note-46">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-46) Gao, L. et al. (2023). Табл. 1–2.</span>
47. <span id="cite_note-47">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-47) Robertson, S.; Zaragoza, H. (2009). ‘‘The Probabilistic Relevance Framework: BM25 and Beyond’’. Found. Trends IR. DOI:10.1561/1500000019.</span>
48. <span id="cite_note-48">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-48) Karpukhin, V. et al. (2020). DOI:10.18653/v1/2020.emnlp-main.550.</span>
49. <span id="cite_note-49">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-49) Xiong, L. et al. (2021). <a href="https://arxiv.org/abs/2007.00808" class="external text" rel="nofollow">arXiv:2007.00808</a>.</span>
50. <span id="cite_note-50">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-50) Nogueira, R. et al. (2019). arXiv:1904.08375.</span>
51. <span id="cite_note-51">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-51) Nogueira, R.; Lin, J. (2019). tech report.</span>
52. <span id="cite_note-52">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(BG)#cite_ref-52) Rocchio, J. (1971). SMART; Lavrenko & Croft (2001) SIGIR.</span>