---
title: "Hypothetical Document Embeddings (HyDE) (KO)"
source: "https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)"
wiki: "systems-analysis.info/int"
article: "Hypothetical_Document_Embeddings_(HyDE)_(KO)"
language: "ko"
categories:
  - "Category:Korean"
  - "Category:Large language models"
  - "Category:Pages with math errors"
  - "Category:Pages with math render errors"
  - "Category:Prompt engineering"
revision_id: 3117
wiki_created_at: 2026-09-06T23:15:57Z
wiki_modified_at: 2026-09-06T23:15:57Z
downloaded_at: 2026-09-07T22:55:16Z
---

# Hypothetical Document Embeddings (HyDE) (KO)

**Hypothetical Document Expansion (HyDE)** — 벡터 검색(retrieval) 및 retrieval‑augmented generation (RAG)을 개선하는 방법으로, 대형 언어 모델(LLM)이 원래 쿼리로부터 '가상 문서(hypothetical document)'를 생성하고, 이 텍스트를 인코더로 벡터화한 뒤, 생성된 벡터와의 유사도를 기준으로 실제 문서를 검색하는 방식이다. 이 접근법은 LLM에 인코딩된 '관련성 패턴'을 활용하여 밀집 embedding을 통해 코퍼스에 '근거(grounding)'를 부여할 수 있게 한다<sup>[\[1\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-1)</sup>.

## 정의와 직관

HyDE는 검색 작업을 두 단계로 분해한다:

\(1\) LLM이 쿼리에 대한 '관련 응답 예시'(*hypothetical document*)를 생성하여 관련성 특징을 모델링하고;

\(2\) 대조적(contrastive) 인코더(예: Contriever)가 해당 텍스트를 벡터로 변환하여, 이를 기반으로 인덱스에서 실제 문서를 추출한다. 생성된 텍스트에는 사실적 오류가 포함될 수 있지만, 인코더가 포착하는 주제적·용어적 패턴이 중요하다<sup>[\[2\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-2)</sup>.

## 역사와 출처

합성 텍스트를 이용한 검색 확장 아이디어는 쿼리 확장(query expansion) 및 의사 관련성 피드백(PRF) 연구에 그 뿌리를 두고 있다: Rocchio 알고리즘과 관련성 언어 모델<sup>[\[3\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-3)[\[4\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-4)</sup>. 밀집 검색(dense retrieval)에는 대조적으로 학습된 인코더(Contriever)<sup>[\[5\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-5)</sup>와 Dense Passage Retrieval (DPR)<sup>[\[6\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-6)</sup>이 활용되었다. BEIR benchmark는 zero‑shot 평가를 표준화하였다<sup>[\[7\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-7)</sup>. 이러한 배경에서 HyDE는 인코더를 fine-tuning하지 않고도 LLM을 통해 관련성 지식을 zero-shot 방식으로 도입하는 방법으로 제안되었다<sup>[\[8\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-8)</sup>.

## 방법과 형식화

문서 코퍼스를 $\mathcal{D} = \{ d_{1},\ldots,d_{N}\}$, 텍스트 인코더 $E:\text{text} \rightarrow {\mathbb{R}}^{n}$가 문서의 벡터 표현 $\mathbf{v}_{d} = E(d)$을 정의한다고 하자. 유사도 측정에는 코사인 유사도 또는 내적이 사용된다. 중요한 점은 \*\*내적은 두 벡터의 L2 노름이 단위 노름일 때만 코사인 유사도와 일치한다\*\*는 것이다($\|\mathbf{u}\| = \|\mathbf{v}\| = 1$)<sup>[\[9\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-9)</sup>.

HyDE는 LLM이 생성한 '가상 문서'를 통해 쿼리 표현을 재정의한다. 형식적으로:

$$
\begin{matrix}
 & \text{(1) Генерация гипотетического текста:} & & {\overset{\sim}{d}\; = \; G\!\left( q;\,{inst} \right),} \\
 & \text{(2) Эмбеддинг гипотетического текста:} & & {\mathbf{v}_{h}\; = \; E(\overset{\sim}{d}),} \\
 & \text{(3) Поиск ближайших соседей:} & & {\mathcal{R}_{k}(q)\; = \;{TopK}_{\, d \in \mathcal{D}}\; S\!\left( \mathbf{v}_{h},\mathbf{v}_{d} \right),}
\end{matrix}
$$

여기서 $G$은 지시문 $inst$(예: "다음 질문에 답하는 단락을 작성하시오 …")을 갖는 LLM이고, $S$은 유사도 측정값(코사인 또는 정규화된 내적), $\mathcal{R}_{k}(q)$는 최대 유사도를 갖는 $k$개 문서의 집합이다<sup>[\[10\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-10)[\[11\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-11)</sup>.

실제 엔지니어링에서는 \*\*여러\*\* 가상 텍스트를 생성하여 그 표현을 집계하는 방식이 자주 사용되며, 이는 안정성을 높인다:

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

여기서 $\xi_{j}$는 확률적 디코딩 파라미터(예: temperature/top‑p)이다. 이러한 앙상블(ensemble)화는 적절한 지연 시간 증가로 Recall을 향상시킨다<sup>[\[12\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-12)</sup>.

### HyDE 기본 파이프라인

    # 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

### 다른 방법과의 관계 (QE, doc2query, PRF)

- **QE(쿼리 확장)**는 쿼리에 용어를 추가하는 방식이고, HyDE는 대신 전체 '준문서(quasi‑document)'를 생성하여 밀집 인코더와 더 잘 호환된다<sup>[\[13\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-13)</sup>.
- **doc2query / docTTTTTquery**는 인덱싱 전에 합성 쿼리로 **문서**를 확장한다<sup>[\[14\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-14)[\[15\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-15)</sup>. HyDE는 재인덱싱 없이 실시간으로 **쿼리**를 확장한다.
- **PRF**(Rocchio, Relevance LM)는 상위 결과를 기반으로 쿼리 벡터를 업데이트하는 반면, HyDE는 LLM에서 직접 '관련성 패턴'을 추출한 뒤 코퍼스 검색을 통해 '근거'를 부여한다<sup>[\[16\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-16)</sup>.

## RAG 및 재순위화(reranking)와의 통합

RAG에서 HyDE는 검색의 첫 번째 단계로 사용된다: 가상 문서 → embedding → k개 후보. 이후 재순위화가 적용된다: BERT 계열 크로스 인코더<sup>[\[17\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-17)</sup> 또는 ColBERT의 후기 상호작용(late interaction)<sup>[\[18\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-18)</sup>. 목록 병합(예: BM25+벡터 하이브리드)에는 일반적으로 RRF(*reciprocal rank fusion*)가 사용된다: $\operatorname{RRF}(d) = \sum\limits_{r \in \mathcal{R}}\frac{1}{k + \operatorname{rank}_{r}(d)},\qquad k \approx 60.$ RRF 방법은 결합된 순위 목록의 전반적인 품질을 안정적으로 향상시킨다<sup>[\[19\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-19)</sup>.

## 벤치마크 평가 (BEIR 등)

원저 논문은 TREC DL'19/20(웹 검색)과 BEIR 컬렉션의 하위 집합(Scifact, ArguAna, TREC‑COVID, FiQA, DBPedia, TREC‑NEWS, Climate‑FEVER)에서 HyDE를 zero-shot 방식으로 평가한다. 결과 일부 — *2023-07 기준*:

| 방법                      | DL19                   | DL20                   | 출처                                                                                                             |
|---------------------------|------------------------|------------------------|------------------------------------------------------------------------------------------------------------------|
| BM25                      | 30.1 / 50.6 / 75.0     | 28.6 / 48.0 / 78.6     | <sup>[\[20\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#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)_(KO)#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)_(KO)#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)_(KO)#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)_(KO)#cite_note-24)</sup> |

*TREC DL19/20 (웹 검색)* — mAP / nDCG@10 / Recall@1k

| 방법       | Scifact         | ArguAna         | TREC‑COVID      | FiQA        | DBPedia     | TREC‑NEWS       | Climate‑FEVER   | 출처                                                                                                             |
|------------|-----------------|-----------------|-----------------|-------------|-------------|-----------------|-----------------|------------------------------------------------------------------------------------------------------------------|
| 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)_(KO)#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)_(KO)#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)_(KO)#cite_note-27)</sup> |

*BEIR (데이터셋 모음)* — nDCG@10 / Recall@100

HyDE는 다국어 데이터셋 Mr.TyDi(sw/ko/ja/bn)에서도 mContriever 대비 MRR@100을 향상시킨다<sup>[\[28\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-28)</sup>.

## 실용적 권고사항

HyDE 적용 시점

- Zero-shot/전이(transfer) 환경(관련성 레이블이 없거나, 학습 코퍼스와 도메인이 크게 다른 경우)<sup>[\[29\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-29)</sup>.
- 적절한 정밀도를 유지하면서 Recall@k를 높여야 할 때 — HyDE는 벡터 공간에서 관련 영역을 자주 '발견'한다<sup>[\[30\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-30)</sup>.

일반적인 설정

- **LLM 및 prompt**: "다음 질문에 답하는 단락을 작성하시오 …" 지시문; 적당한 확률성(예: *temperature*≈0.7)<sup>[\[31\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-31)</sup>.
- **가상 텍스트 수**: 1–5개; embedding 평균화가 안정성을 높인다<sup>[\[32\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-32)</sup>.
- **임베더**: fine-tuning 없는 (m)Contriever; fine-tuning된 인코더도 사용 가능(HyDE 효과 유지됨)<sup>[\[33\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-33)</sup>.
- **Embedding 정규화**: L2 노름; 내적이 코사인과 동등해진다<sup>[\[34\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-34)</sup>.
- **하이브리드 검색**: BM25+벡터 후 재순위화<sup>[\[35\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-35)</sup>.
- **재순위기**: Cross-Encoder(BERT re‑ranker)<sup>[\[36\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-36)</sup> 또는 ColBERT<sup>[\[37\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-37)</sup>.
- **결과 병합**: 다양한 전략의 결과를 RRF(*k*≈60)로 통합<sup>[\[38\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-38)</sup>.

품질/비용 모니터링

- 검색: nDCG@k, Recall@k, MRR; end‑to‑end RAG: EM/F1 또는 *groundedness* 지표(RAGAS/TruLens)<sup>[\[39\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-39)[\[40\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-40)</sup>.
- 비용/지연 시간: LLM 생성과 (있을 경우) 재순위화가 지배적이며, 가상 텍스트 수와 응답 길이로 최적화할 수 있다<sup>[\[41\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-41)</sup>.

## 한계와 미해결 과제

- **가상 텍스트의 환각(hallucination)**: LLM이 사실적 오류를 도입할 수 있으며, 인코더와 코퍼스를 통한 '근거 부여'가 위험을 줄이지만 완전히 제거하지는 못한다<sup>[\[42\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-42)</sup>.
- **도메인/언어 제한**: 고도로 전문화된 도메인과 저자원 언어에서는 HyDE의 이점이 감소한다<sup>[\[43\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-43)</sup>.
- **지연 시간과 비용**: LLM 생성은 지연 및 token 비용을 추가하며, 온라인 시나리오 및 긴 가상 텍스트에서 특히 중요하다<sup>[\[44\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-44)</sup>.
- **윤리 및 편향**: 안전한 LLM과 필터링을 사용하는 것이 바람직하다<sup>[\[45\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-45)</sup>.

## 방법 비교표

| 방법                      | 분류              | 텍스트 생성 위치     | 인코더/인덱스          | 재순위기(2단계)              | 대표 지표(예시)                                    | 비용/지연 시간                       | 출처                                                                                                                                                                                                                  |
|---------------------------|-------------------|----------------------|------------------------|------------------------------|----------------------------------------------------|--------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| **HyDE**                  | 쿼리→*가상 문서*  | 쿼리 측(LLM → 단락)  | (m)Contriever; ANN     | BERT re‑rank / ColBERT / RRF | DL19 nDCG@10≈61.3; DL20≈57.9; ArguAna nDCG@10≈46.6 | \+ LLM 생성; + 재순위화(선택)        | <sup>[\[46\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-46)</sup>                                                                                                      |
| BM25                      | 어휘적(lexical)   | —                    | 역색인(inverted index) | 선택적                       | 위 표 참조                                         | 낮음(lexical)                        | <sup>[\[47\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-47)</sup>                                                                                                      |
| DPR / ANCE                | 밀집(fine-tuning) | —                    | Bi-encoder; ANN        | 선택적                       | DL19 nDCG@10≈62–65                                 | 보통(LLM 없음)                       | <sup>[\[48\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-48)[\[49\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-49)</sup> |
| doc2query / docTTTTTquery | 문서 확장         | 컬렉션 측(인덱싱 전) | BM25/sparse+확장       | 선택적                       | MS MARCO에서 BM25 개선                             | 오프라인 생성 비용 높음; 온라인 빠름 | <sup>[\[50\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-50)[\[51\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-51)</sup> |
| PRF (Rocchio, RLM)        | 피드백 기반 QE    | 쿼리(상위 결과 기반) | 임의                   | 선택적                       | Recall 향상/드리프트 위험                          | \+ 추가 검색 패스                    | <sup>[\[52\]](https://systems-analysis.info/int/Hypothetical_Document_Embeddings_(HyDE)_(KO)#cite_note-52)</sup>                                                                                                      |

HyDE와 관련 접근법 비교

## 같이 보기

- BM25
- 벡터 표현 기반 검색
- RAG
- 의사 관련성 피드백
- BEIR

## 참고 문헌

- 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.

## 외부 링크

- HyDE 저장소: github.com/texttron/hyde.
- 문서: Haystack — HyDE: docs.haystack.deepset.ai.
- 문서: LangChain — HyDE Retriever: docs.langchain.com.

## 주석

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