Filters








35 Hits in 2.7 sec

A Simple Yet Strong Pipeline for HotpotQA [article]

Dirk Groeneveld, Tushar Khot, Mausam, Ashish Sabharwal
<span title="2020-04-14">2020</span> <i > arXiv </i> &nbsp; <span class="release-stage" >pre-print</span>
Our results suggest that the answer may be no, because even our simple pipeline based on BERT, named Quark, performs surprisingly well.  ...  Specifically, on HotpotQA, Quark outperforms these models on both question answering and support identification (and achieves performance very close to a RoBERTa model).  ...  We use the 90447 questions from the HotpotQA training set, shuffle them, and train for 4 epochs. Both models are trained in the distractor setting  ... 
<span class="external-identifiers"> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2004.06753v1">arXiv:2004.06753v1</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/pcz4w6ichreltje2vkrv7zjqui">fatcat:pcz4w6ichreltje2vkrv7zjqui</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20200417223643/https://arxiv.org/pdf/2004.06753v1.pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] </button> </a> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2004.06753v1" title="arxiv.org access"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> arxiv.org </button> </a>

A Simple Yet Strong Pipeline for HotpotQA

Dirk Groeneveld, Tushar Khot, Mausam, Ashish Sabharwal
<span title="">2020</span> <i title="Association for Computational Linguistics"> Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) </i> &nbsp; <span class="release-stage">unpublished</span>
Our results suggest that the answer may be no, because even our simple pipeline based on BERT, named QUARK, performs surprisingly well.  ...  The strong performance of QUARK resurfaces the importance of carefully exploring simple model designs before using popular benchmarks to justify the value of complex techniques.  ...  Acknowledgments Mausam is supported by grants from Google, Bloomberg, and 1MG, Jai Gupta Chair Fellowship, and a Visvesvaraya faculty award by the Govt. of India.  ... 
<span class="external-identifiers"> <a target="_blank" rel="external noopener noreferrer" href="https://doi.org/10.18653/v1/2020.emnlp-main.711">doi:10.18653/v1/2020.emnlp-main.711</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/iu5gq56zunaufp64erc7itn4ay">fatcat:iu5gq56zunaufp64erc7itn4ay</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20201209074422/https://www.aclweb.org/anthology/2020.emnlp-main.711.pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/db/ba/dbba416787d2d4b1c348655ac4a7d6506f0cd191.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener noreferrer" href="https://doi.org/10.18653/v1/2020.emnlp-main.711"> <button class="ui left aligned compact blue labeled icon button serp-button"> <i class="external alternate icon"></i> Publisher / doi.org </button> </a>

From Easy to Hard: Two-stage Selector and Reader for Multi-hop Question Answering [article]

Xin-Yi Li, Wei-Jun Lei, Yu-Bin Yang
<span title="2022-05-24">2022</span> <i > arXiv </i> &nbsp; <span class="release-stage" >pre-print</span>
To address the above issue, we propose a simple yet effective novel framework, From Easy to Hard (FE2H), to remove distracting information and obtain better contextual representations for the multi-hop  ...  As for the QA phase, our reader is first trained on a single-hop QA dataset and then transferred into the multi-hop QA task.  ...  Acknowledgements We thank many colleagues at Nanjing University for their help, particularly Qing-Long Zhang, for useful discussion.  ... 
<span class="external-identifiers"> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2205.11729v1">arXiv:2205.11729v1</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/7yvlb7e2wjheveqprordrbubqa">fatcat:7yvlb7e2wjheveqprordrbubqa</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20220526110015/https://arxiv.org/pdf/2205.11729v1.pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/25/da/25da873d4632b143b072a3e8390c77f267f748e0.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2205.11729v1" title="arxiv.org access"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> arxiv.org </button> </a>

Mitigating False-Negative Contexts in Multi-document Question Answering with Retrieval Marginalization [article]

Ansong Ni, Matt Gardner, Pradeep Dasigi
<span title="2021-09-08">2021</span> <i > arXiv </i> &nbsp; <span class="release-stage" >pre-print</span>
Question Answering (QA) tasks requiring information from multiple documents often rely on a retrieval model to identify relevant information for reasoning.  ...  We also show that retrieval marginalization results in 4.1 QA F1 improvement over a non-marginalized baseline on HotpotQA in the fullwiki setting.  ...  We think it is unlikely for our method to be misused for other domains.  ... 
<span class="external-identifiers"> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2103.12235v2">arXiv:2103.12235v2</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/viynqnvrbbaffdhfa6p46rzpgm">fatcat:viynqnvrbbaffdhfa6p46rzpgm</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20210916073418/https://arxiv.org/pdf/2103.12235v2.pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/44/3f/443f889a2d4da14e930586e9416068f91c93a5bb.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2103.12235v2" title="arxiv.org access"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> arxiv.org </button> </a>

A Simple Approach to Jointly Rank Passages and Select Relevant Sentences in the OBQA Context [article]

Man Luo, Shuguang Chen, Chitta Baral
<span title="2021-09-22">2021</span> <i > arXiv </i> &nbsp; <span class="release-stage" >pre-print</span>
In this work, we present a simple yet effective framework to address these problems by jointly ranking passages and selecting sentences.  ...  In the open question answering (OBQA) task, how to select the relevant information from a large corpus is a crucial problem for reasoning and inference.  ...  Groeneveld et al. (2020) proposes a pipeline based on three BERT models to solve the HotpotQA challenge.  ... 
<span class="external-identifiers"> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2109.10497v1">arXiv:2109.10497v1</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/yiny75rxqvb2xcetc3yak54yka">fatcat:yiny75rxqvb2xcetc3yak54yka</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20210924163212/https://arxiv.org/pdf/2109.10497v1.pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/b2/6e/b26ed38ff0438513f5201d0dd725e1d7ca8d5c28.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2109.10497v1" title="arxiv.org access"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> arxiv.org </button> </a>

LEPUS: Prompt-based Unsupervised Multi-hop Reranking for Open-domain QA [article]

Muhammad Khalifa, Lajanugen Logeswaran, Moontae Lee, Honglak Lee, Lu Wang
<span title="2022-05-25">2022</span> <i > arXiv </i> &nbsp; <span class="release-stage" >pre-print</span>
We study unsupervised multi-hop reranking for multi-hop QA (MQA) with open-domain questions.  ...  as the probability of generating a given question, according to a pre-trained language model.  ...  can be strong zero-shot re-rankers for MQA.  ... 
<span class="external-identifiers"> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2205.12650v1">arXiv:2205.12650v1</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/mzcozv3vifh2fb4cowpldvtgty">fatcat:mzcozv3vifh2fb4cowpldvtgty</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20220527031916/https://arxiv.org/pdf/2205.12650v1.pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/66/50/6650e44f00f74c596b3981d01493b5f75cfa6f63.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2205.12650v1" title="arxiv.org access"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> arxiv.org </button> </a>

Answering Any-hop Open-domain Questions with Iterative Document Reranking [article]

Ping Nie, Yuyu Zhang, Arun Ramamurthy, Le Song
<span title="2021-05-24">2021</span> <i > arXiv </i> &nbsp; <span class="release-stage" >pre-print</span>
Existing approaches for open-domain question answering (QA) are typically designed for questions that require either single-hop or multi-hop reasoning, which make strong assumptions of the complexity of  ...  method consistently achieves performance comparable to or better than the state-of-the-art on both single-hop and multi-hop open-domain QA datasets, including Natural Questions Open, SQuAD Open, and HotpotQA  ...  [3] , which builds a simple pipeline with a TF-IDF retriever module and a RNN-based reader module to produce answers from the top 5 retrieved documents.  ... 
<span class="external-identifiers"> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2009.07465v5">arXiv:2009.07465v5</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/m3hosnzbwvdx3k3g7x7tlm5zhi">fatcat:m3hosnzbwvdx3k3g7x7tlm5zhi</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20210521081744/https://arxiv.org/pdf/2009.07465v4.pdf" title="fulltext PDF download [not primary version]" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <span style="color: #f43e3e;">&#10033;</span> <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/23/35/23350d3c08965f6fd521a93cebab9ae44846a968.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2009.07465v5" title="arxiv.org access"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> arxiv.org </button> </a>

Break, Perturb, Build: Automatic Perturbation of Reasoning Paths Through Question Decomposition [article]

Mor Geva, Tomer Wolfson, Jonathan Berant
<span title="2021-10-18">2021</span> <i > arXiv </i> &nbsp; <span class="release-stage" >pre-print</span>
In this work, we introduce the "Break, Perturb, Build" (BPB) framework for automatic reasoning-oriented perturbation of question-answer pairs.  ...  We evaluate a range of RC models on our evaluation sets, which reveals large performance gaps on generated examples compared to the original data.  ...  This work was completed in partial fulfillment for the Ph.D degree of Mor Geva.  ... 
<span class="external-identifiers"> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2107.13935v2">arXiv:2107.13935v2</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/tiuy4yeqrzexnlnkt4nxvf62ua">fatcat:tiuy4yeqrzexnlnkt4nxvf62ua</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20211026122338/https://arxiv.org/pdf/2107.13935v2.pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/85/08/8508e3fa84e4d74c0fb9333b1b8fc8a10f30a74f.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2107.13935v2" title="arxiv.org access"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> arxiv.org </button> </a>

Selective Question Answering under Domain Shift [article]

Amita Kamath, Robin Jia, Percy Liang
<span title="2020-06-16">2020</span> <i > arXiv </i> &nbsp; <span class="release-stage" >pre-print</span>
In this work, we propose the setting of selective question answering under domain shift, in which a QA model is tested on a mixture of in-domain and out-of-domain data, and must answer (i.e., not abstain  ...  We combine this method with a SQuAD-trained QA model and evaluate on mixtures of SQuAD and five other QA datasets.  ...  We thank Ananya Kumar, John Hewitt, Dan Iter, and the anonymous reviewers for their helpful comments and insights.  ... 
<span class="external-identifiers"> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2006.09462v1">arXiv:2006.09462v1</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/uyaux4gjo5bnpomxp5dhlanqya">fatcat:uyaux4gjo5bnpomxp5dhlanqya</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20200623084559/https://arxiv.org/pdf/2006.09462v1.pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] </button> </a> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2006.09462v1" title="arxiv.org access"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> arxiv.org </button> </a>

Knowing More About Questions Can Help: Improving Calibration in Question Answering [article]

Shujian Zhang, Chengyue Gong, Eunsol Choi
<span title="2021-06-02">2021</span> <i > arXiv </i> &nbsp; <span class="release-stage" >pre-print</span>
We study calibration in question answering, estimating whether model correctly predicts answer for each question.  ...  Our simple and efficient calibrator can be easily adapted to many tasks and model architectures, showing robust gains in all settings.  ...  Acknowledgments We would like to thank UT Austin NLP group, especially Kaj Bostrom and Greg Durrett for feedback and suggestions.  ... 
<span class="external-identifiers"> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2106.01494v1">arXiv:2106.01494v1</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/ng2o6xdfevatbozasglpbke4xu">fatcat:ng2o6xdfevatbozasglpbke4xu</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20210615045124/https://arxiv.org/pdf/2106.01494v1.pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/2c/33/2c33f4aa30a4d8f708db79eef5c9e2a0ce84695b.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2106.01494v1" title="arxiv.org access"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> arxiv.org </button> </a>

KECP: Knowledge Enhanced Contrastive Prompting for Few-shot Extractive Question Answering [article]

Jianing Wang, Chengyu Wang, Minghui Qiu, Qiuhui Shi, Hongbin Wang, Jun Huang, Ming Gao
<span title="2022-05-06">2022</span> <i > arXiv </i> &nbsp; <span class="release-stage" >pre-print</span>
Instead of adding pointer heads to PLMs, we introduce a seminal paradigm for EQA that transform the task into a non-autoregressive Masked Language Modeling (MLM) generation problem.  ...  However, most existing approaches for MRC may perform poorly in the few-shot learning scenario.  ...  C Details of Negative Span Sampling In order to construct negative spans for span-level contrastive learning (SCL), we follow a simple pipeline to implement confusion span sampling.  ... 
<span class="external-identifiers"> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2205.03071v1">arXiv:2205.03071v1</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/sxa4ufwaarf3bnkyvmkop6u7xa">fatcat:sxa4ufwaarf3bnkyvmkop6u7xa</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20220510063645/https://arxiv.org/pdf/2205.03071v1.pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/37/5c/375c41566e8710c7b6cbf12c1bf6347f5aa23ab8.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2205.03071v1" title="arxiv.org access"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> arxiv.org </button> </a>

FiD-Ex: Improving Sequence-to-Sequence Models for Extractive Rationale Generation [article]

Kushal Lakhotia, Bhargavi Paranjape, Asish Ghoshal, Wen-tau Yih, Yashar Mehdad, Srinivasan Iyer
<span title="2020-12-31">2020</span> <i > arXiv </i> &nbsp; <span class="release-stage" >pre-print</span>
Natural language (NL) explanations of model predictions are gaining popularity as a means to understand and verify decisions made by large black-box pre-trained models, for NLP tasks such as Question Answering  ...  However, these models have many shortcomings; they can fabricate explanations even for incorrect predictions, they are difficult to adapt to long input documents, and their training requires a large amount  ...  SentencePiece: A simple and language independent subword tok- enizer and detokenizer for neural text processing.  ... 
<span class="external-identifiers"> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2012.15482v1">arXiv:2012.15482v1</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/4tdg4dcq5vfwvnqb6ta4fwldbu">fatcat:4tdg4dcq5vfwvnqb6ta4fwldbu</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20210102025502/https://arxiv.org/pdf/2012.15482v1.pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/17/d4/17d4681b29b79c4ee5029ae39acabfdf9946bd77.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2012.15482v1" title="arxiv.org access"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> arxiv.org </button> </a>

UnifieR: A Unified Retriever for Large-Scale Retrieval [article]

Tao Shen, Xiubo Geng, Chongyang Tao, Can Xu, Kai Zhang, Daxin Jiang
<span title="2022-05-23">2022</span> <i > arXiv </i> &nbsp; <span class="release-stage" >pre-print</span>
Large-scale retrieval is to recall relevant documents from a huge collection given a query. It relies on representation learning to embed documents and queries into a common semantic encoding space.  ...  model with a dual-representing capability.  ...  In experiments, we train our retriever on ad-hoc retrieval datasets upon static hard negative technique [64, 16] only without strong yet costly rerankers/cross-encoders for distillations [50, 66] .  ... 
<span class="external-identifiers"> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2205.11194v1">arXiv:2205.11194v1</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/ttwz7flj75dg7ezchxsos2tfum">fatcat:ttwz7flj75dg7ezchxsos2tfum</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20220529181338/https://arxiv.org/pdf/2205.11194v1.pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/4c/2b/4c2b24df128cc6409201efd8830656056731be51.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2205.11194v1" title="arxiv.org access"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> arxiv.org </button> </a>

The Unreliability of Explanations in Few-Shot In-Context Learning [article]

Xi Ye, Greg Durrett
<span title="2022-05-06">2022</span> <i > arXiv </i> &nbsp; <span class="release-stage" >pre-print</span>
Following these observations, we present a framework for calibrating model predictions based on the reliability of the explanations.  ...  Despite careful prompting, explanations generated by GPT-3 may not even be factually grounded in the input, even on simple tasks with straightforward extractive explanations.  ...  This work was partially supported by NSF Grant IIS-1814522 and a gift from Salesforce Inc.  ... 
<span class="external-identifiers"> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2205.03401v1">arXiv:2205.03401v1</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/uc6v7zi52nfi5duwdnetasxa3a">fatcat:uc6v7zi52nfi5duwdnetasxa3a</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20220510050249/https://arxiv.org/pdf/2205.03401v1.pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/e8/11/e811e771f5950d86eafe50655c0d1e5b571e19b6.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2205.03401v1" title="arxiv.org access"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> arxiv.org </button> </a>

Question Answering Infused Pre-training of General-Purpose Contextualized Representations [article]

Robin Jia, Mike Lewis, Luke Zettlemoyer
<span title="2022-03-16">2022</span> <i > arXiv </i> &nbsp; <span class="release-stage" >pre-print</span>
We propose a pre-training objective based on question answering (QA) for learning general-purpose contextual representations, motivated by the intuition that the representation of a phrase in a passage  ...  By encoding QA-relevant information, the bi-encoder's token-level representations are useful for non-QA downstream tasks without extensive (or in some cases, any) fine-tuning.  ...  Acknowledgements We thank Terra Blevins for investigating applications to word sense disambiguation, Jiaxin Huang for providing the few-shot NER splits used in their paper, and Douwe Kiela, Max Bartolo  ... 
<span class="external-identifiers"> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2106.08190v2">arXiv:2106.08190v2</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/bh63p5or5zcixh63hpedad7s7i">fatcat:bh63p5or5zcixh63hpedad7s7i</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20220516042113/https://arxiv.org/pdf/2106.08190v2.pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/78/51/78513aa199fa5d71f0ab6154fb422d7f0674795e.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/2106.08190v2" title="arxiv.org access"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> arxiv.org </button> </a>
&laquo; Previous Showing results 1 &mdash; 15 out of 35 results