Matches in SemOpenAlex for { <https://semopenalex.org/work/W4384264626> ?p ?o ?g. }
Showing items 1 to 67 of
67
with 100 items per page.
- W4384264626 abstract "We propose a framework - Prompt, Generate, Train (PGT) - to efficiently develop a generative question-answering model for open-book question-answering over a proprietary collection of text documents. The framework adapts a retriever augmented generation (RAG) model to the target domain using supervised fine-tuning and reinforcement learning with synthetic feedback in a few-shot setting. This, we hypothesize, will yield an aligned, uncertainty calibrated model that is competitive with GPT-4 based in-context retrieval augmented generation in generating relevant answers at lower serving costs. The framework's synthetic generation pipeline will generate synthetic training data comprising <passage, question, answer> tuples using an open-source LLM and a novel consistency filtering scheme. The pipeline will be designed to generate both abstractive and extractive questions that span the entire corpus. The framework proposes to fine-tune a smaller RAG model comprising a dense retriever (ColBERTv2) and a smaller sized LLM on the synthetic dataset. In parallel, the framework will train a Reward model to score domain grounded answers higher than hallucinated answers using an a priori relevance ordering of synthetically assembled samples. In the next phase, the framework will align the RAG model with the target domain using reinforcement learning (Proximal Policy Optimization). This step may improve the RAG model's ability to generate grounded answers and ignore out of domain questions. In the final phase, the framework will calibrate the model's uncertainty for extractive question-answers." @default.
- W4384264626 created "2023-07-14" @default.
- W4384264626 creator A5029852787 @default.
- W4384264626 date "2023-07-12" @default.
- W4384264626 modified "2023-10-14" @default.
- W4384264626 title "Prompt Generate Train (PGT): Few-shot Domain Adaption of Retrieval Augmented Generation Models for Open Book Question-Answering" @default.
- W4384264626 doi "https://doi.org/10.48550/arxiv.2307.05915" @default.
- W4384264626 hasPublicationYear "2023" @default.
- W4384264626 type Work @default.
- W4384264626 citedByCount "0" @default.
- W4384264626 crossrefType "posted-content" @default.
- W4384264626 hasAuthorship W4384264626A5029852787 @default.
- W4384264626 hasBestOaLocation W43842646261 @default.
- W4384264626 hasConcept C118615104 @default.
- W4384264626 hasConcept C118930307 @default.
- W4384264626 hasConcept C119857082 @default.
- W4384264626 hasConcept C134306372 @default.
- W4384264626 hasConcept C153083717 @default.
- W4384264626 hasConcept C154945302 @default.
- W4384264626 hasConcept C158154518 @default.
- W4384264626 hasConcept C160920958 @default.
- W4384264626 hasConcept C17744445 @default.
- W4384264626 hasConcept C199360897 @default.
- W4384264626 hasConcept C199539241 @default.
- W4384264626 hasConcept C23123220 @default.
- W4384264626 hasConcept C2776436953 @default.
- W4384264626 hasConcept C33923547 @default.
- W4384264626 hasConcept C36503486 @default.
- W4384264626 hasConcept C41008148 @default.
- W4384264626 hasConcept C43521106 @default.
- W4384264626 hasConcept C44291984 @default.
- W4384264626 hasConcept C97541855 @default.
- W4384264626 hasConceptScore W4384264626C118615104 @default.
- W4384264626 hasConceptScore W4384264626C118930307 @default.
- W4384264626 hasConceptScore W4384264626C119857082 @default.
- W4384264626 hasConceptScore W4384264626C134306372 @default.
- W4384264626 hasConceptScore W4384264626C153083717 @default.
- W4384264626 hasConceptScore W4384264626C154945302 @default.
- W4384264626 hasConceptScore W4384264626C158154518 @default.
- W4384264626 hasConceptScore W4384264626C160920958 @default.
- W4384264626 hasConceptScore W4384264626C17744445 @default.
- W4384264626 hasConceptScore W4384264626C199360897 @default.
- W4384264626 hasConceptScore W4384264626C199539241 @default.
- W4384264626 hasConceptScore W4384264626C23123220 @default.
- W4384264626 hasConceptScore W4384264626C2776436953 @default.
- W4384264626 hasConceptScore W4384264626C33923547 @default.
- W4384264626 hasConceptScore W4384264626C36503486 @default.
- W4384264626 hasConceptScore W4384264626C41008148 @default.
- W4384264626 hasConceptScore W4384264626C43521106 @default.
- W4384264626 hasConceptScore W4384264626C44291984 @default.
- W4384264626 hasConceptScore W4384264626C97541855 @default.
- W4384264626 hasLocation W43842646261 @default.
- W4384264626 hasOpenAccess W4384264626 @default.
- W4384264626 hasPrimaryLocation W43842646261 @default.
- W4384264626 hasRelatedWork W2031812225 @default.
- W4384264626 hasRelatedWork W2921507169 @default.
- W4384264626 hasRelatedWork W3048366122 @default.
- W4384264626 hasRelatedWork W3094502663 @default.
- W4384264626 hasRelatedWork W3170628611 @default.
- W4384264626 hasRelatedWork W3210635025 @default.
- W4384264626 hasRelatedWork W4287690154 @default.
- W4384264626 hasRelatedWork W4310560699 @default.
- W4384264626 hasRelatedWork W4319083788 @default.
- W4384264626 hasRelatedWork W4321473427 @default.
- W4384264626 isParatext "false" @default.
- W4384264626 isRetracted "false" @default.
- W4384264626 workType "article" @default.