---
title: "Hypothetical Document Embeddings (HyDE) (PT)"
source: "https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)"
wiki: "systems-analysis.info/int"
article: "Hypothetical_Document_Embeddings_(HyDE)_(PT)"
language: "pt"
categories:
  - "Category:Large language models"
  - "Category:Portuguese"
  - "Category:Prompt engineering"
revision_id: 3120
wiki_created_at: 2026-09-06T23:15:59Z
wiki_modified_at: 2026-09-06T23:15:59Z
downloaded_at: 2026-09-07T22:55:18Z
---

# Hypothetical Document Embeddings (HyDE) (PT)

**Hypothetical Document Expansion (HyDE)** — é um método para aprimorar a recuperação vetorial (vector retrieval) e a geração aumentada por recuperação (retrieval-augmented generation, RAG), no qual um modelo de linguagem grande (LLM) gera um "documento hipotético" a partir de uma consulta original. Em seguida, esse texto é vetorizado por um codificador (encoder), e a busca é realizada entre documentos reais com base na proximidade com o vetor resultante. A abordagem permite utilizar "padrões de relevância" codificados pelo LLM e "ancorá-los" (grounding) a um corpus por meio de embeddings densos<sup>[\[1\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-1)</sup>.

## Definição e Intuição

O HyDE decompõe a tarefa de busca em duas etapas:

\(1\) O LLM cria um "exemplo de resposta relevante" (*hypothetical document*) para a consulta, modelando assim as características de relevância;

\(2\) Um codificador contrastivo (por exemplo, Contriever) converte esse texto em um vetor, com base no qual documentos reais são recuperados do índice. O texto gerado pode conter erros factuais, mas o que importa são os padrões temáticos e terminológicos capturados pelo codificador<sup>[\[2\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-2)</sup>.

## História e Fontes

A ideia de expandir a busca com textos sintéticos remonta a trabalhos sobre expansão de consulta (query expansion) e feedback de pseudo-relevância (pseudo-relevance feedback, PRF): o algoritmo de Rocchio e os modelos de linguagem de relevância<sup>[\[3\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-3)[\[4\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-4)</sup>. Para a recuperação densa (dense retrieval), foram utilizados codificadores treinados de forma contrastiva (Contriever)<sup>[\[5\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-5)</sup> e o Dense Passage Retrieval (DPR)<sup>[\[6\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-6)</sup>. O benchmark BEIR padronizou a avaliação zero-shot<sup>[\[7\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-7)</sup>. Nesse contexto, o HyDE foi proposto como uma forma de "injetar" conhecimento de relevância no modo zero-shot por meio de um LLM, sem a necessidade de treinar novamente o codificador<sup>[\[8\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-8)</sup>.

## Método e Formalização

Seja um corpus de documentos $\mathcal{D} = \{ d_{1},\ldots,d_{N}\}$ e um codificador de texto $E:\text{text} \rightarrow {\mathbb{R}}^{n}$ que define as representações vetoriais dos documentos $\mathbf{v}_{d} = E(d)$. Para medir a proximidade, utiliza-se a similaridade de cosseno ou o produto escalar. Uma observação importante: \*\*o produto escalar coincide com a similaridade de cosseno apenas quando a norma L2 de ambos os vetores é unitária\*\* ($\|\mathbf{u}\| = \|\mathbf{v}\| = 1$)<sup>[\[9\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-9)</sup>.

O HyDE redefine a representação da consulta por meio de um "documento hipotético" gerado pelo LLM. Formalmente:

$$
\begin{matrix}
 & \text{(1) Geração do texto hipotético:} & & {\left. \overset{\sim}{d}\; = \; G\!(q;\,{inst} \right),} \\
 & \text{(2) Embedding do texto hipotético:} & & {\mathbf{v}_{h}\; = \; E(\overset{\sim}{d}),} \\
 & \text{(3) Busca dos vizinhos mais próximos:} & & {\mathcal{R}_{k}(q)\; = \;{TopK}_{\, d \in \mathcal{D}}\; S\!\left( \mathbf{v}_{h},\mathbf{v}_{d} \right),}
\end{matrix}
$$

onde $G$ é o LLM com a instrução $inst$ (por exemplo: "Escreva um parágrafo que responda à pergunta..."), $S$ é a medida de similaridade (cosseno ou produto escalar com normalização), e $\mathcal{R}_{k}(q)$ é o conjunto dos $k$ documentos com a maior similaridade<sup>[\[10\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-10)[\[11\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-11)</sup>.

Na prática de engenharia, é comum gerar \*\*vários\*\* textos hipotéticos e agregar suas representações, o que aumenta a robustez:

$$
\left. {\overset{\sim}{d}}^{(j)} = G\!(q;\,{inst},\xi_{j} \right),\quad\mathbf{v}_{h}\; = \;\frac{1}{m}\sum\limits_{j = 1}^{m}E\!\left( {\overset{\sim}{d}}^{(j)} \right),
$$

onde $\xi_{j}$ são parâmetros estocásticos de decodificação (ex: temperature/top-p). Essa abordagem de ensemble melhora o Recall com um aumento moderado na latência<sup>[\[12\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-12)</sup>.

### Pipeline Básico do HyDE

    # 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) (para RAG) stuff / map-reduce / refine nos topN

### Relação com outros métodos (QE, doc2query, PRF)

- **QE (expansão de consulta)** adiciona termos à consulta; o HyDE, em vez disso, gera um "quase-documento" inteiro, o que se alinha melhor com codificadores densos<sup>[\[13\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-13)</sup>.
- **doc2query / docTTTTTquery** expandem os **documentos** com consultas sintéticas antes da indexação<sup>[\[14\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-14)[\[15\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-15)</sup>; o HyDE expande a **consulta** em tempo de execução, sem exigir reindexação.
- **PRF** (Rocchio, Relevance LM) atualiza o vetor da consulta com base nos principais resultados; o HyDE extrai o "padrão de relevância" diretamente do LLM e depois o "ancora" (grounds) através da recuperação no corpus<sup>[\[16\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-16)</sup>.

## Integração em RAG e Reclassificação (Reranking)

Em sistemas RAG, o HyDE é aplicado como a primeira etapa de recuperação: documento hipotético → embedding → k candidatos. Em seguida, utiliza-se a reclassificação (reranking): cross-encoders da classe BERT<sup>[\[17\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-17)</sup> ou interação tardia (late interaction) do ColBERT<sup>[\[18\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-18)</sup>. Para fundir listas (por exemplo, um híbrido de BM25 e vetores), tipicamente se aplica o RRF (*reciprocal rank fusion*): $\operatorname{RRF}(d) = \sum\limits_{r \in \mathcal{R}}\frac{1}{k + \operatorname{rank}_{r}(d)},\qquad k \approx 60.$ O método RRF aumenta consistentemente a qualidade agregada das classificações combinadas<sup>[\[19\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-19)</sup>.

## Avaliação em Benchmarks (BEIR e outros)

O trabalho original avalia o HyDE no modo zero-shot nos benchmarks TREC DL'19/20 (busca na web) e em um subconjunto de coleções do BEIR (Scifact, ArguAna, TREC-COVID, FiQA, DBPedia, TREC-NEWS, Climate-FEVER). Um trecho dos resultados — *atualizado até 07/2023*:

| Método                    | DL19                   | DL20                   | Fonte                                                                                                            |
|---------------------------|------------------------|------------------------|------------------------------------------------------------------------------------------------------------------|
| BM25                      | 30.1 / 50.6 / 75.0     | 28.6 / 48.0 / 78.6     | <sup>[\[20\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-20)</sup> |
| 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)_(PT)#cite_note-21)</sup> |
| **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)_(PT)#cite_note-22)</sup> |
| 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)_(PT)#cite_note-23)</sup> |
| 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)_(PT)#cite_note-24)</sup> |

*TREC DL19/20 (busca na web)* — mAP / nDCG@10 / Recall@1k

| Método     | Scifact         | ArguAna         | TREC‑COVID      | FiQA        | DBPedia     | TREC‑NEWS       | Climate‑FEVER   | Fonte                                                                                                            |
|------------|-----------------|-----------------|-----------------|-------------|-------------|-----------------|-----------------|------------------------------------------------------------------------------------------------------------------|
| 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)_(PT)#cite_note-25)</sup> |
| 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)_(PT)#cite_note-26)</sup> |
| **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)_(PT)#cite_note-27)</sup> |

*BEIR (seleção de datasets)* — nDCG@10 / Recall@100

O HyDE também melhora o MRR@100 em datasets multilíngues do Mr.TyDi (sw/ko/ja/bn) em comparação com o mContriever<sup>[\[28\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-28)</sup>.

## Recomendações Práticas

Quando aplicar o HyDE

- Modos zero-shot ou de transferência (sem rótulos de relevância; "incompatibilidade" de domínio com os corpora de treinamento)<sup>[\[29\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-29)</sup>.
- Necessidade de aumentar o Recall@k com precisão aceitável — o HyDE frequentemente "descobre" regiões relevantes do espaço vetorial<sup>[\[30\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-30)</sup>.

Configurações Típicas

- **LLM e prompt**: instrução "Escreva um parágrafo que responda à pergunta..."; estocasticidade moderada (ex: *temperature*≈0.7)<sup>[\[31\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-31)</sup>.
- **Número de textos hipotéticos**: 1–5; a média dos embeddings aumenta a robustez<sup>[\[32\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-32)</sup>.
- **Embedder**: (m)Contriever sem treinamento adicional; é possível usar codificadores treinados (o efeito do HyDE é preservado)<sup>[\[33\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-33)</sup>.
- **Normalização de embeddings**: norma L2; o produto interno é equivalente ao cosseno<sup>[\[34\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-34)</sup>.
- **Recuperação híbrida**: BM25 + vetores, seguido de reclassificação<sup>[\[35\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-35)</sup>.
- **Reclassificador (Reranker)**: Cross-Encoder (BERT re-ranker)<sup>[\[36\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-36)</sup> ou ColBERT<sup>[\[37\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-37)</sup>.
- **Fusão** de resultados de diferentes estratégias: RRF (*k*≈60)<sup>[\[38\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-38)</sup>.

Monitoramento de Qualidade/Custo

- Recuperação: nDCG@k, Recall@k, MRR; RAG de ponta a ponta: EM/F1 ou métricas de *groundedness* (RAGAS/TruLens)<sup>[\[39\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-39)[\[40\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-40)</sup>.
- Custo/Latência: dominado pela geração do LLM e (se houver) pela reclassificação; otimizado pelo número de textos "hipotéticos" e pelo comprimento da resposta<sup>[\[41\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-41)</sup>.

## Limitações e Questões Abertas

- **Alucinações** no texto hipotético: o LLM pode introduzir erros factuais; a "ancoragem" (grounding) através do codificador e do corpus reduz o risco, mas não o elimina completamente<sup>[\[42\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-42)</sup>.
- **Limitações de domínio/idioma**: o ganho do HyDE diminui em domínios altamente especializados e em idiomas com poucos recursos<sup>[\[43\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-43)</sup>.
- **Latência e Custo**: a geração do LLM adiciona atraso e custo de tokens; crítico para cenários online e textos "hipotéticos" longos<sup>[\[44\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-44)</sup>.
- **Ética e Vieses**: é preferível usar LLMs seguros e aplicar filtragem<sup>[\[45\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-45)</sup>.

## Tabela Comparativa de Métodos

| Método                    | Classe           | Onde o texto é gerado                   | Codificador/Índice   | Reclassificador (2ª etapa)   | Métricas típicas (exemplo)                         | Custo/Latência                              | Fontes                                                                                                                                                                                                                |
|---------------------------|------------------|-----------------------------------------|----------------------|------------------------------|----------------------------------------------------|---------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| **HyDE**                  | Query→*hypo‑doc* | No lado da consulta (LLM → parágrafo)   | (m)Contriever; ANN   | BERT re‑rank / ColBERT / RRF | DL19 nDCG@10≈61.3; DL20≈57.9; ArguAna nDCG@10≈46.6 | \+ geração do LLM; + reclassificação (opc.) | <sup>[\[46\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-46)</sup>                                                                                                      |
| BM25                      | Lexical          | —                                       | Índice invertido     | Opcional                     | ver tabelas (acima)                                | Baixo (lexical)                             | <sup>[\[47\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-47)</sup>                                                                                                      |
| DPR / ANCE                | Denso (ft)       | —                                       | Bi‑encoder; ANN      | Opcional                     | DL19 nDCG@10≈62–65                                 | Médio (sem LLM)                             | <sup>[\[48\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-48)[\[49\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-49)</sup> |
| doc2query / docTTTTTquery | Expansão de doc. | No lado da coleção (antes da indexação) | BM25/sparse+expanded | Opcional                     | Melhorias no BM25 em MS MARCO                      | Geração offline alta; online rápido         | <sup>[\[50\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-50)[\[51\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-51)</sup> |
| PRF (Rocchio, RLM)        | QE por feedback  | Consulta (com base nos melhores res.)   | Qualquer             | Opcional                     | Aumento do Recall/riscos de desvio                 | \+ passagem adicional de recuperação        | <sup>[\[52\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#cite_note-52)</sup>                                                                                                      |

Comparação entre HyDE e abordagens relacionadas

## Ver também

- BM25
- Busca por representações vetoriais
- RAG
- Feedback de pseudo-relevância
- BEIR

## Ligações externas

- Repositório do HyDE: <a href="https://github.com/texttron/hyde" class="external text" rel="nofollow">github.com/texttron/hyde</a>.
- Documentação: Haystack — HyDE: <a href="https://docs.haystack.deepset.ai/docs/hypothetical-document-embeddings-hyde" class="external text" rel="nofollow">docs.haystack.deepset.ai</a>.
- Documentação: LangChain — HyDE Retriever: <a href="https://docs.langchain.com/oss/javascript/integrations/retrievers/hyde" class="external text" rel="nofollow">docs.langchain.com</a>.

## Leitura adicional

- Manning, C. D.; Raghavan, P.; Schütze, H. (2008). *Introduction to Information Retrieval*. Cambridge University Press. ISBN 978‑0521865715.
- 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.

## Notas

1.  <span id="cite_note-1">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#cite_ref-34) Milvus Docs. ‘‘Similarity Metrics’’.</span>
35. <span id="cite_note-35">[↑](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#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)_(PT)#cite_ref-52) Rocchio, J. (1971). SMART; Lavrenko & Croft (2001) SIGIR.</span>