English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

We propose a framework - Prompt, Generate, Train (PGT) - to efficientlydevelop a generative question-answering model for open-book question-answeringover a proprietary collection of text documents. The framework adapts aretriever augmented generation (RAG) model to the target domain usingsupervised fine-tuning and reinforcement learning with synthetic feedback in afew-shot setting. This, we hypothesize, will yield an aligned, uncertaintycalibrated model that is competitive with GPT-4 based in-context retrievalaugmented generation in generating relevant answers at lower serving costs. Theframework's synthetic generation pipeline will generate synthetic training datacomprising <passage, question, answer> tuples using an open-source LLM and anovel consistency filtering scheme. The pipeline will be designed to generateboth abstractive and extractive questions that span the entire corpus. Theframework proposes to fine-tune a smaller RAG model comprising a denseretriever (ColBERTv2) and a smaller sized LLM on the synthetic dataset. Inparallel, the framework will train a Reward model to score domain groundedanswers higher than hallucinated answers using an a priori relevance orderingof synthetically assembled samples. In the next phase, the framework will alignthe RAG model with the target domain using reinforcement learning (ProximalPolicy Optimization). This step may improve the RAG model's ability to generategrounded answers and ignore out of domain questions. In the final phase, theframework will calibrate the model's uncertainty for extractivequestion-answers.

Prompt Generate Train (PGT): Few-shot Domain Adaption of Retrieval Augmented Generation Models for Open Book Question-Answering