Pith. sign in

REVIEW 6 cited by

A BERT Baseline for the Natural Questions

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 1901.08634 v3 pith:VSAMI45I submitted 2019-01-24 cs.CL

Chris Alberti , Kenton Lee , Michael Collins This is my paper · ORCID
classification cs.CL
keywords baselinebertmodellanguagenaturalquestionsansweranswering
verification ladder T0 review T1 audit T2 compute T3 formal

Signed reviews

No signed human review yet.

0 comments
read the original abstract

This technical note describes a new baseline for the Natural Questions. Our model is based on BERT and reduces the gap between the model F1 scores reported in the original dataset paper and the human upper bound by 30% and 50% relative for the long and short answer tasks respectively. This baseline has been submitted to the official NQ leaderboard at ai.google.com/research/NaturalQuestions. Code, preprocessed data and pretrained model are available at https://github.com/google-research/language/tree/master/language/question_answering/bert_joint.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 6 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. On Identifiability in Transformers

    cs.CL 2019-08 conditional novelty 7.0 of 10

    Self-attention weights in Transformers are non-identifiable for long sequences, and the paper offers effective attention and Hidden Token Attribution as diagnostic tools.

  2. Better Rewards Yield Better Summaries: Learning to Summarise Without References

    cs.CL 2019-09 conditional novelty 6.0 of 10

    A reward function learned from human ratings, with no reference summaries, guides RL summarizers to produce summaries rated higher by humans than ROUGE-trained systems.

  3. Multi-Granularity Representations of Dialog

    cs.CL 2019-08 conditional novelty 6.0 of 10

    A training procedure that samples negative responses by semantic distance to learn multi-granularity representations improves next-utterance retrieval and downstream transfer.

  4. Multi-passage BERT: A Globally Normalized BERT Model for Open-domain Question Answering

    cs.CL 2019-08 accept novelty 6.0 of 10

    Applying global normalization across passages, 100-word sliding windows, and a passage ranker to BERT yields state-of-the-art open-domain QA results on four benchmarks.

  5. Agentic AI-Driven Technical Troubleshooting for Enterprise Systems: A Novel Weighted Retrieval-Augmented Generation Paradigm

    cs.AI 2024-12 conditional novelty 4.0 of 10

    The paper claims a dynamically weighted RAG framework improves enterprise troubleshooting accuracy to 90.8% versus 85.2% for standard RAG, though supporting details are sparse.

  6. CFO: A Framework for Building Production NLP Systems

    cs.CL 2019-08 conditional novelty 4.0 of 10

    The paper presents CFO, a containerized orchestration framework for building production NLP systems, and GAAMA, a question answering system that combines BM25 retrieval with BERT, with empirical results on NQ and SQuAD 2.0.

Pith tools