metadata

language:
  - en
license: apache-2.0
tags:
  - question-answering
datasets:
  - adversarial_qa
  - mbartolo/synQA
  - squad
metrics:
  - exact_match
  - f1
model-index:
  - name: mbartolo/roberta-large-synqa-ext
    results:
      - task:
          type: question-answering
          name: Question Answering
        dataset:
          name: adversarial_qa
          type: adversarial_qa
          config: adversarialQA
          split: validation
        metrics:
          - type: exact_match
            value: 53.2
            name: Exact Match
            verified: true
            verifyToken: >-
              eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9.eyJoYXNoIjoiZDVkOTMxYjYxMTkzYzY1ZWYwZWU2ZGEwNTFkNTVlYjQwYzYyNjkwNzM3Yzc5NzBhZWI3MjYxMzQ2ZmNhNWNhNCIsInZlcnNpb24iOjF9.A0BkWcotTRPButOduP74809_EjrMv55qZxG1t5vCXgp25FHmyaLJOVhAvwYckwWDoatkffRkPg63K45nedsXDA
          - type: f1
            value: 64.6266
            name: F1
            verified: true
            verifyToken: >-
              eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9.eyJoYXNoIjoiODkyZjY3NTEwYTJmNmMyZGQzYTFkMjk5MDdlNWZmNWI4MGI5MDMxMTZkZWNhMGY1ZDdiMmM5YWRiOTAwMjgyOCIsInZlcnNpb24iOjF9.n7f0XdCOnszr4T_FnEuRhkW6_aG2IjLrkjkIaoPX1bB5UCk_veUoXqQWP728-AOSzxZwOaHv-7EjuCDip5h3Bw

Model Overview

This is a RoBERTa-Large QA Model trained from https://huggingface.co/roberta-large in two stages. First, it is trained on synthetic adversarial data generated using a BART-Large question generator on Wikipedia passages from SQuAD as well as Wikipedia passages external to SQuAD, and then it is trained on SQuAD and AdversarialQA (https://arxiv.org/abs/2002.00293) in a second stage of fine-tuning.

Data

Training data: SQuAD + AdversarialQA Evaluation data: SQuAD + AdversarialQA

Training Process

Approx. 1 training epoch on the synthetic data and 2 training epochs on the manually-curated data.

Additional Information

Please refer to https://arxiv.org/abs/2104.08678 for full details.