End-to-End Training of Multi-Document Reader and Retriever for Open-Domain Question Answering
By Devendra Singh Sachan, Siva Reddy, William L. Hamilton, Chris Dyer, Dani Yogatama. 2021
192 citations recorded by the source corpus.
We present an end-to-end differentiable training method for retrieval-augmented open-domain question answering systems that combine information from multiple retrieved documents when generating answers. We model retrieval decisions as latent variables over sets of relevant documents. Since marginalizing over sets of retrieved documents is computationally hard, we approximate this using an expectation-maximization algorithm. We iteratively estimate the value of our latent variable (the set of relevant documents for a given question) and then use this estimate to update the retriever and reader parameters. We hypothesize that such end-to-end training allows training signals to flow to the reader and then to the retriever better than staged-wise training. This results in a retriever that is able to select more relevant documents for a question and a reader that is trained on more accurate documents to generate an answer. Experiments on three benchmark datasets demonstrate that our proposed method outperforms all existing approaches of comparable size by 2-3% absolute exact match points, achieving new state-of-the-art results. Our results also demonstrate the feasibility of learning to retrieve to improve answer generation without explicit supervision of retrieval decisions.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Information Systems, Retrieval-Augmented Generation, Information Retrieval & Search
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Luke S. Zettlemoyer
- Mohit Bansal
- Jianfeng Gao
- Graham Neubig
- Percy Liang
- Hannaneh Hajishirzi
- Yoshua Bengio
- Quoc V. Le
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Trevor Darrell
- Andrew Zisserman
- Ming-Hsuan Yang
- Cordelia Schmid
- Xiaogang Wang
- Jitendra Malik
In Information Systems
- Sebastian Riedel
- Luyu Gao
- Danqi Chen
- Wen-tau Yih
- Graham Neubig
- Luke S. Zettlemoyer
- Jamie Callan
- Christopher D. Manning
In Retrieval-Augmented Generation
- Sebastian Riedel
- Danqi Chen
- Wen-tau Yih
- Fabio Petroni
- Luyu Gao
- Jamie Callan
- Luke S. Zettlemoyer
- Patrick Lewis
In Information Retrieval & Search
- Sebastian Riedel
- Wen-tau Yih
- Danqi Chen
- Luyu Gao
- Edouard Grave
- Fabio Petroni
- Gautier Izacard
- Jamie Callan
Citations in the corpus
- Few-shot Learning with Retrieval Augmented Language Models (2022, methodology): “Sachan et al. (2021) and Lee et al. (2021a) both employ salient span masking to pre-train retrievers, leveraging the perplexity and attention scores from the language model.”
- Improving language models by retrieving from trillions of tokens (2021, positive): “RAG (Lewis et al., 2020) and FiD (Izacard and Grave, 2021) (Sachan et al., 2021) extends FiD by using an expectation-maximization algorithm to train the retriever end-to-end and achieves state of the art results compared to similarly sized models.”
- Multi-Task Retrieval-Augmented Text Generation with Relevance Sampling (2022, background): “Many of the recent papers improving RAG-style models optimized end-to-end processes (f.e. EMDR2 (Singh et al., 2021)), ensembling multiple modules (f.e. R2-D2 (Fajcik et al., 2021)), or creating multiple training loops to update the indexed documents multiple times (f.e. Hindsight (Paranjape et…”
- Internet-augmented language models through few-shot prompting for open-domain question answering (2022, methodology): “When applied to question answering tasks (Lewis et al., 2020; Izacard & Grave, 2020; Sachan et al., 2021), they surpass performance of parametric-only models, as they are able to efficiently handle an increasing number of retrieved passages and ground their predictions into additional information.”
- Unsupervised Dense Information Retrieval with Contrastive Learning (2021, methodology): “On QA datasets, we compare to dense retrievers trained with ICT and the Masked Salient Spans from Sachan et al. (2021).”
- You Only Need One Model for Open-domain Question Answering (2021, methodology): “The query is encoded with ‘ query :’ prefix while passages are encoded with ‘ title :’ and ‘ context :’ prefixes following previous approaches (Izac-ard and Grave, 2021b; Singh et al., 2021).”
- Learning to Retrieve Passages without Supervision (2021, methodology): “Sachan et al. (2021a) trained their own ICT model, which shows stronger performance than Lee et al. (2019).”
- LIFT: Language-Interfaced Fine-Tuning for Non-Language Machine Learning Tasks (2022, background): “[100] Devendra Singh, Siva Reddy, Will Hamilton, Chris Dyer, and Dani Yogatama.”
- Improving Passage Retrieval with Zero-Shot Question Generation (2022, methodology): “For our experiments, we use the Fusion-in-Decoder model implementation from the open-source repository ( https: //github.com/DevSinghSachan/emdr2 ) (Sachan et al., 2021b).”
- Training Language Models with Memory Augmentation (2022, methodology): “Memory-augmented models for downstream tasks While our paper focuses on improving language models with memory augmentation, other works improve models for downstream tasks with a retrieval component, such as question answering (Kumar et al., 2016; de Masson D’Autume et al., 2019; Karpukhin et al., 2”
- Reveal: Retrieval-Augmented Visual-Language Pre-Training with Multi-Source Multimodal Knowledge Memory (2022, methodology): “Retrieval-augmented models have attracted a fair amount of attention in the fields of NLP [14, 18] and computer vision [13,25].”
- Reasoning on Graphs: Faithful and Interpretable Large Language Model Reasoning (2023, methodology): “For the retrieval-reasoning module, we follow the FiD framework (Izacard & Grave, 2021; Singh et al., 2021), which allows reasoning on multiple retrieved reasoning paths 2 , formulated as By approximating the expectation with K sampled plans Z ∗ K ⊆ Z ∗ , the objective function of reasoning…”
- A Survey on RAG Meeting LLMs: Towards Retrieval-Augmented Large Language Models (2024, unknown): “In this part, we will briefly review studies with these two types of generators: parameter-accessible (white-box) and parameter-inaccessible (black-box).”
- Improving the Domain Adaptation of Retrieval Augmented Generation (RAG) Models for Open Domain Question Answering (2022, methodology): “Singh et al. (2021) mainly focused on improving the document reading ability and answer generation rather than domain adaptation.”
- A Unified Encoder-Decoder Framework with Entity Memory (2022, methodology): “Subsequent approaches combined DPR with a generative objective to build large, powerful models on open-domain QA and generation tasks (Lewis et al., 2020b; Izacard and Grave, 2021; Sachan et al., 2021; Yu et al., 2022a).”
- Active Retrieval Augmented Generation (2023, background): “…a complete answer conditioning on the retrieved documents (Chen et al., 2017; Guu et al., 2020; Lewis et al., 2020; Izacard and Grave, 2021; Sachan et al., 2021; Lee et al., 2021; Jiang et al., 2022; Izacard et al., 2022; Nakano et al., 2021; Qian et al., 2023; Lazaridou et al., 2022; Shi…”
- Retrieval as Attention: End-to-end Learning of Retrieval and Reading within a Single Transformer (2022, methodology): “There have been several attempts to connect up reader and retriever training (Lee et al., 2019; Guu et al., 2020; Lewis et al., 2020; Sachan et al., 2021; Lee et al., 2021a; Izacard et al., 2022).”
- Generate rather than Retrieve: Large Language Models are Strong Context Generators (2022, background): “, 2022), or training the system end-to-end (Lewis et al., 2020; Singh et al., 2021).”
- StreamingQA: A Benchmark for Adaptation to New Knowledge over Time in Question Answering Models (2022, methodology): “Recent work in the domain has focused on improving performance by combining information from multiple documents efficiently (Sachan et al., 2021; Izacard & Grave, 2020) and on performance analysis of the dense retrievals, for instance, when dealing with named entities (Sciavolino et al., 2021; Liu…”