id
stringlengths
33
34
title
stringlengths
6
218
authors
stringlengths
5
6.98k
categories
stringlengths
5
188
summary
stringlengths
34
3.26k
link_pdf
stringlengths
33
34
updated
timestamp[s]
published
timestamp[s]
embeddings
dict
http://arxiv.org/abs/2212.10843v1
Generating Multiple-Length Summaries via Reinforcement Learning for Unsupervised Sentence Summarization
Dongmin Hyun, Xiting Wang, Chanyoung Park, Xing Xie, Hwanjo Yu
cs.CL, cs.AI
Sentence summarization shortens given texts while maintaining core contents of the texts. Unsupervised approaches have been studied to summarize texts without human-written summaries. However, recent unsupervised models are extractive, which remove words from texts and thus they are less flexible than abstractive summa...
http://arxiv.org/pdf/2212.10843v1
2022-12-21T08:34:28
2022-12-21T08:34:28
{ "title": [ -0.012130737, -0.0043945312, -0.019332886, -0.04989624, -0.01826477, 0.0076026917, -0.044128418, -0.012939453, -0.03555298, 0.02911377, -0.001127243, 0.040527344, -0.034301758, -0.0045776367, -0.027297974, -0.01335907, -0.0143585205, ...
http://arxiv.org/abs/2212.13104v1
The URW-KG: a Resource for Tackling the Underrepresentation of non-Western Writers
Marco Antonio Stranisci, Giuseppe Spillo, Cataldo Musto, Viviana Patti, Rossana Damiano
cs.CL
Digital media have enabled the access to unprecedented literary knowledge. Authors, readers, and scholars are now able to discover and share an increasing amount of information about books and their authors. Notwithstanding, digital archives are still unbalanced: writers from non-Western countries are less represented,...
http://arxiv.org/pdf/2212.13104v1
2022-12-21T07:53:26
2022-12-21T07:53:26
{ "title": [ -0.04800415, 0.008857727, -0.037475586, -0.012138367, -0.019485474, -0.007411957, -0.054656982, -0.029083252, -0.035064697, 0.036590576, -0.0029468536, 0.035217285, 0.015182495, -0.048034668, 0.026855469, -0.0071907043, -0.00868988, ...
http://arxiv.org/abs/2212.10826v1
End-to-End Automatic Speech Recognition model for the Sudanese Dialect
Ayman Mansour, Wafaa F. Mukhtar
cs.CL
Designing a natural voice interface rely mostly on Speech recognition for interaction between human and their modern digital life equipment. In addition, speech recognition narrows the gap between monolingual individuals to better exchange communication. However, the field lacks wide support for several universal langu...
http://arxiv.org/pdf/2212.10826v1
2022-12-21T07:35:33
2022-12-21T07:35:33
{ "title": [ -0.006504059, 0.010818481, -0.024749756, -0.052459717, 0.0025844574, 0.0093307495, -0.052703857, 0.001996994, -0.028305054, 0.068359375, 0.042541504, -0.015449524, 0.048217773, -0.059570312, -0.03515625, 0.0362854, 0.011108398, 0.035...
http://arxiv.org/abs/2212.10823v3
Continual Contrastive Finetuning Improves Low-Resource Relation Extraction
Wenxuan Zhou, Sheng Zhang, Tristan Naumann, Muhao Chen, Hoifung Poon
cs.CL, cs.AI, cs.LG
Relation extraction (RE), which has relied on structurally annotated corpora for model training, has been particularly challenging in low-resource scenarios and domains. Recent literature has tackled low-resource RE by self-supervised learning, where the solution involves pretraining the entity pair embedding by RE-bas...
http://arxiv.org/pdf/2212.10823v3
2023-05-31T06:04:33
2022-12-21T07:30:22
{ "title": [ 0.039367676, -0.0028419495, -0.027938843, -0.097961426, 0.014678955, -0.014137268, -0.03186035, -0.008560181, 0.035095215, 0.034332275, -0.033691406, 0.01322937, -0.0065574646, -0.0042419434, -0.06359863, 0.0048713684, -0.008857727, ...
http://arxiv.org/abs/2212.10819v1
Attend to the Right Context: A Plug-and-Play Module for Content-Controllable Summarization
Wen Xiao, Lesly Miculicich, Yang Liu, Pengcheng He, Giuseppe Carenini
cs.CL
Content-Controllable Summarization generates summaries focused on the given controlling signals. Due to the lack of large-scale training corpora for the task, we propose a plug-and-play module RelAttn to adapt any general summarizers to the content-controllable summarization task. RelAttn first identifies the relevant ...
http://arxiv.org/pdf/2212.10819v1
2022-12-21T07:17:32
2022-12-21T07:17:32
{ "title": [ 0.01965332, 0.0020885468, -0.035583496, -0.09472656, 0.005458832, -0.003129959, -0.047698975, -0.0028800964, -0.024017334, 0.047668457, -0.013412476, 0.038208008, -0.0068206787, -0.006462097, -0.03366089, 0.0045433044, 0.00012624264, ...
http://arxiv.org/abs/2212.10818v2
4D ASR: Joint modeling of CTC, Attention, Transducer, and Mask-Predict decoders
Yui Sudo, Muhammad Shakeel, Brian Yan, Jiatong Shi, Shinji Watanabe
cs.SD, cs.CL, eess.AS
The network architecture of end-to-end (E2E) automatic speech recognition (ASR) can be classified into several models, including connectionist temporal classification (CTC), recurrent neural network transducer (RNN-T), attention mechanism, and non-autoregressive mask-predict models. Since each of these network architec...
http://arxiv.org/pdf/2212.10818v2
2023-05-29T23:16:56
2022-12-21T07:15:59
{ "title": [ -0.027618408, -0.009498596, -0.056854248, -0.051452637, -0.02267456, 0.005203247, -0.031341553, 0.011817932, -0.0042533875, 0.04107666, -0.0014152527, -0.023345947, 0.03164673, -0.04437256, -0.031234741, 0.005958557, -0.00793457, 0.0...
http://arxiv.org/abs/2212.10815v1
ZEROTOP: Zero-Shot Task-Oriented Semantic Parsing using Large Language Models
Dheeraj Mekala, Jason Wolfe, Subhro Roy
cs.CL
We explore the use of large language models (LLMs) for zero-shot semantic parsing. Semantic parsing involves mapping natural language utterances to task-specific meaning representations. Language models are generally trained on the publicly available text and code and cannot be expected to directly generalize to domain...
http://arxiv.org/pdf/2212.10815v1
2022-12-21T07:06:55
2022-12-21T07:06:55
{ "title": [ -0.028427124, -0.00907135, -0.039794922, -0.09020996, 0.0024166107, 0.011497498, -0.08050537, 0.01777649, 0.01626587, 0.024856567, -0.0155181885, -0.0018177032, -0.0033454895, -0.013122559, -0.0440979, 0.009017944, -0.013580322, 0.04...
http://arxiv.org/abs/2212.10792v1
Reconstruction Probing
Najoung Kim, Jatin Khilnani, Alex Warstadt, Abed Qaddoumi
cs.CL
We propose reconstruction probing, a new analysis method for contextualized representations based on reconstruction probabilities in masked language models (MLMs). This method relies on comparing the reconstruction probabilities of tokens in a given sequence when conditioned on the representation of a single token that...
http://arxiv.org/pdf/2212.10792v1
2022-12-21T06:22:03
2022-12-21T06:22:03
{ "title": [ 0.0018405914, -0.013137817, 0.013801575, -0.014808655, -0.06378174, -0.000998497, -0.051574707, -0.03189087, -0.035369873, 0.043151855, -0.032714844, 0.010643005, 0.030838013, -0.019226074, 0.03878784, -0.021392822, 0.03265381, -0.01...
http://arxiv.org/abs/2212.10791v1
OpineSum: Entailment-based self-training for abstractive opinion summarization
Annie Louis, Joshua Maynez
cs.CL, cs.AI
A typical product or place often has hundreds of reviews, and summarization of these texts is an important and challenging problem. Recent progress on abstractive summarization in domains such as news has been driven by supervised systems trained on hundreds of thousands of news articles paired with human-written summa...
http://arxiv.org/pdf/2212.10791v1
2022-12-21T06:20:28
2022-12-21T06:20:28
{ "title": [ 0.013114929, -0.00084495544, 0.0034236908, -0.0713501, -0.020767212, 0.022125244, -0.052215576, -0.013923645, 0.028442383, 0.03036499, -0.016799927, 0.013160706, -0.020263672, -0.0013837814, -0.060180664, 0.005344391, -0.0009794235, ...
http://arxiv.org/abs/2212.10789v1
Multi-modal Molecule Structure-text Model for Text-based Retrieval and Editing
Shengchao Liu, Weili Nie, Chengpeng Wang, Jiarui Lu, Zhuoran Qiao, Ling Liu, Jian Tang, Chaowei Xiao, Anima Anandkumar
cs.LG, cs.CL, q-bio.QM, stat.ML
There is increasing adoption of artificial intelligence in drug discovery. However, existing works use machine learning to mainly utilize the chemical structures of molecules yet ignore the vast textual knowledge available in chemistry. Incorporating textual knowledge enables us to realize new drug design objectives, a...
http://arxiv.org/pdf/2212.10789v1
2022-12-21T06:18:31
2022-12-21T06:18:31
{ "title": [ -0.046844482, -0.012435913, -0.04067993, -0.029220581, -0.03579712, 0.006965637, -0.059265137, -0.0017242432, 0.020019531, 0.04727173, 0.0016994476, -0.007987976, 0.046173096, -0.030258179, -0.04385376, 0.021820068, -0.011856079, 0.0...
http://arxiv.org/abs/2212.10786v2
Multi-hop Evidence Retrieval for Cross-document Relation Extraction
Keming Lu, I-Hung Hsu, Wenxuan Zhou, Mingyu Derek Ma, Muhao Chen
cs.CL, cs.IR, cs.LG
Relation Extraction (RE) has been extended to cross-document scenarios because many relations are not simply described in a single document. This inevitably brings the challenge of efficient open-space evidence retrieval to support the inference of cross-document relations, along with the challenge of multi-hop reasoni...
http://arxiv.org/pdf/2212.10786v2
2023-06-05T00:43:13
2022-12-21T06:00:22
{ "title": [ -0.006969452, -0.01612854, -0.047180176, -0.05581665, 0.0005788803, -0.0049743652, -0.029159546, 0.007160187, 0.028320312, 0.021316528, -0.031982422, 0.019439697, 0.00063467026, -0.0036334991, -0.03050232, 0.008354187, -0.018478394, ...
http://arxiv.org/abs/2212.11275v1
KL Regularized Normalization Framework for Low Resource Tasks
Neeraj Kumar, Ankur Narang, Brejesh Lall
cs.CL, cs.AI, cs.LG
Large pre-trained models, such as Bert, GPT, and Wav2Vec, have demonstrated great potential for learning representations that are transferable to a wide variety of downstream tasks . It is difficult to obtain a large quantity of supervised data due to the limited availability of resources and time. In light of this, a ...
http://arxiv.org/pdf/2212.11275v1
2022-12-21T05:59:25
2022-12-21T05:59:25
{ "title": [ -0.042541504, 0.0016422272, -0.052093506, -0.06585693, 0.007549286, 0.016952515, -0.07513428, -0.012649536, 0.0018033981, 0.03060913, 0.013908386, 0.0060691833, -0.0053901672, -0.03692627, -0.034851074, 0.026184082, 0.006877899, 0.03...
http://arxiv.org/abs/2212.10785v2
SERENGETI: Massively Multilingual Language Models for Africa
Ife Adebara, AbdelRahim Elmadany, Muhammad Abdul-Mageed, Alcides Alcoba Inciarte
cs.CL, cs.AI
Multilingual pretrained language models (mPLMs) acquire valuable, generalizable linguistic information during pretraining and have advanced the state of the art on task-specific finetuning. To date, only ~31 out of ~2,000 African languages are covered in existing language models. We ameliorate this limitation by develo...
http://arxiv.org/pdf/2212.10785v2
2023-05-26T20:20:04
2022-12-21T05:54:14
{ "title": [ -0.05117798, -0.027908325, -0.016418457, -0.052215576, -0.025177002, 0.028961182, -0.07336426, 0.0025558472, -0.007843018, 0.043273926, 0.0071487427, 0.005306244, 0.03302002, -0.029129028, -0.025482178, 0.03427124, -0.014717102, 0.03...
http://arxiv.org/abs/2212.10784v3
Can NLI Provide Proper Indirect Supervision for Low-resource Biomedical Relation Extraction?
Jiashu Xu, Mingyu Derek Ma, Muhao Chen
cs.CL, cs.AI, q-bio.QM
Two key obstacles in biomedical relation extraction (RE) are the scarcity of annotations and the prevalence of instances without explicitly pre-defined labels due to low annotation coverage. Existing approaches, which treat biomedical RE as a multi-class classification task, often result in poor generalization in low-r...
http://arxiv.org/pdf/2212.10784v3
2023-10-19T05:46:10
2022-12-21T05:49:08
{ "title": [ 0.010864258, -0.024642944, -0.0063323975, -0.03717041, -0.051239014, 0.010169983, -0.052520752, -0.013145447, -0.005077362, -0.005142212, -0.036499023, -0.02204895, -0.011039734, -0.04324341, 0.018630981, -0.0413208, -0.0075683594, 0...
http://arxiv.org/abs/2212.10773v3
MultiInstruct: Improving Multi-Modal Zero-Shot Learning via Instruction Tuning
Zhiyang Xu, Ying Shen, Lifu Huang
cs.CL
Instruction tuning, a new learning paradigm that fine-tunes pre-trained language models on tasks specified through instructions, has shown promising zero-shot performance on various natural language processing tasks. However, it has yet to be explored for vision and multimodal tasks. In this work, we introduce MUL-TIIN...
http://arxiv.org/pdf/2212.10773v3
2023-06-10T18:33:21
2022-12-21T05:17:06
{ "title": [ -0.021362305, -0.01852417, -0.028762817, -0.049560547, -0.017150879, -0.009681702, -0.044921875, 0.0037708282, 0.001964569, 0.027877808, -0.028045654, 0.020462036, -0.031982422, -0.012672424, -0.02708435, -0.003440857, 0.004463196, 0...
http://arxiv.org/abs/2212.10770v1
ImPaKT: A Dataset for Open-Schema Knowledge Base Construction
Luke Vilnis, Zach Fisher, Bhargav Kanagal, Patrick Murray, Sumit Sanghai
cs.CL
Large language models have ushered in a golden age of semantic parsing. The seq2seq paradigm allows for open-schema and abstractive attribute and relation extraction given only small amounts of finetuning data. Language model pretraining has simultaneously enabled great strides in natural language inference, reasoning ...
http://arxiv.org/pdf/2212.10770v1
2022-12-21T05:02:49
2022-12-21T05:02:49
{ "title": [ -0.03111267, -0.009742737, -0.07183838, -0.0914917, 0.019165039, 0.012962341, -0.06304932, 0.008628845, 0.024475098, 0.015037537, -0.021331787, 0.008140564, -0.0155181885, -0.019256592, -0.033447266, 0.0035877228, 0.0036392212, 0.031...
http://arxiv.org/abs/2212.10769v1
Uncontrolled Lexical Exposure Leads to Overestimation of Compositional Generalization in Pretrained Models
Najoung Kim, Tal Linzen, Paul Smolensky
cs.CL
Human linguistic capacity is often characterized by compositionality and the generalization it enables -- human learners can produce and comprehend novel complex expressions by composing known parts. Several benchmarks exploit distributional control across training and test to gauge compositional generalization, where ...
http://arxiv.org/pdf/2212.10769v1
2022-12-21T05:02:08
2022-12-21T05:02:08
{ "title": [ 0.01386261, 0.00089359283, -0.010635376, -0.066101074, -0.02154541, -0.004383087, -0.054473877, -0.009162903, -0.0070114136, 0.021621704, -0.022399902, 0.009613037, -0.01687622, -0.02848816, -0.029510498, -0.019851685, -0.01158905, 0...
http://arxiv.org/abs/2212.10767v2
How Does Beam Search improve Span-Level Confidence Estimation in Generative Sequence Labeling?
Kazuma Hashimoto, Iftekhar Naim, Karthik Raman
cs.CL
Sequence labeling is a core task in text understanding for IE/IR systems. Text generation models have increasingly become the go-to solution for such tasks (e.g., entity extraction and dialog slot filling). While most research has focused on the labeling accuracy, a key aspect -- of vital practical importance -- has sl...
http://arxiv.org/pdf/2212.10767v2
2023-10-27T04:19:50
2022-12-21T05:01:01
{ "title": [ -0.0030879974, -0.012863159, -0.032440186, -0.024963379, -0.036071777, -0.028305054, -0.040130615, -0.0012836456, -0.016815186, 0.03326416, -0.056030273, 0.015342712, -0.015106201, -0.016998291, -0.009140015, -0.03463745, -0.0061302185, ...
http://arxiv.org/abs/2212.11808v1
A Mutation-based Text Generation for Adversarial Machine Learning Applications
Jesus Guerrero, Gongbo Liang, Izzat Alsmadi
cs.CL, cs.LG
Many natural language related applications involve text generation, created by humans or machines. While in many of those applications machines support humans, yet in few others, (e.g. adversarial machine learning, social bots and trolls) machines try to impersonate humans. In this scope, we proposed and evaluated seve...
http://arxiv.org/pdf/2212.11808v1
2022-12-21T04:57:59
2022-12-21T04:57:59
{ "title": [ -0.045837402, -0.013313293, -0.024398804, -0.02848816, -0.013412476, 0.016067505, -0.04046631, 0.0017766953, -0.030944824, 0.04031372, -0.0022258759, 0.024307251, -0.019165039, -0.011329651, -0.03881836, -0.010620117, -0.0146102905, ...
http://arxiv.org/abs/2212.10764v3
Learning List-Level Domain-Invariant Representations for Ranking
Ruicheng Xian, Honglei Zhuang, Zhen Qin, Hamed Zamani, Jing Lu, Ji Ma, Kai Hui, Han Zhao, Xuanhui Wang, Michael Bendersky
cs.IR, cs.AI, cs.CL, cs.LG
Domain adaptation aims to transfer the knowledge learned on (data-rich) source domains to (low-resource) target domains, and a popular method is invariant representation learning, which matches and aligns the data distributions on the feature space. Although this method is studied extensively and applied on classificat...
http://arxiv.org/pdf/2212.10764v3
2023-10-31T16:30:32
2022-12-21T04:49:55
{ "title": [ 0.038085938, 0.019943237, -0.029388428, -0.04019165, 0.036712646, -0.028259277, -0.0026474, -0.027114868, 0.018630981, 0.029327393, -0.0061836243, 0.0074272156, -0.021484375, -0.0053749084, -0.061431885, 0.012672424, -0.0067596436, 0...
http://arxiv.org/abs/2301.05149v2
Define, Evaluate, and Improve Task-Oriented Cognitive Capabilities for Instruction Generation Models
Lingjun Zhao, Khanh Nguyen, Hal Daumé III
cs.CL, cs.AI, cs.HC, cs.LG, cs.RO
Recent work studies the cognitive capabilities of language models through psychological tests designed for humans. While these studies are helpful for understanding the general capabilities of these models, there is no guarantee that a model possessing sufficient capabilities to pass those tests would actually use thos...
http://arxiv.org/pdf/2301.05149v2
2023-05-28T14:34:54
2022-12-21T04:43:19
{ "title": [ -0.029449463, -0.007572174, -0.0340271, 0.004699707, -0.016952515, -0.038024902, -0.0063972473, -0.00466156, -0.03387451, 0.029022217, -0.0051231384, -0.017807007, 0.006668091, -0.023910522, -0.017288208, -0.011550903, 0.0058288574, ...
http://arxiv.org/abs/2212.10758v2
ORCA: A Challenging Benchmark for Arabic Language Understanding
AbdelRahim Elmadany, El Moatez Billah Nagoudi, Muhammad Abdul-Mageed
cs.CL, cs.AI
Due to their crucial role in all NLP, several benchmarks have been proposed to evaluate pretrained language models. In spite of these efforts, no public benchmark of diverse nature currently exists for evaluation of Arabic. This makes it challenging to measure progress for both Arabic and multilingual language models. ...
http://arxiv.org/pdf/2212.10758v2
2023-05-29T18:27:37
2022-12-21T04:35:43
{ "title": [ -0.035614014, -0.015617371, -0.009689331, -0.06072998, -0.0074005127, 0.012573242, -0.05307007, -0.0013990402, 0.00774765, 0.008972168, -0.0063438416, -0.0060653687, -0.023223877, -0.0317688, -0.051513672, -0.0041236877, -0.00017225742, ...
http://arxiv.org/abs/2212.10755v2
JASMINE: Arabic GPT Models for Few-Shot Learning
El Moatez Billah Nagoudi, Muhammad Abdul-Mageed, AbdelRahim Elmadany, Alcides Alcoba Inciarte, Md Tawkat Islam Khondaker
cs.CL
Scholarship on generative pretraining (GPT) remains acutely Anglocentric, leaving serious gaps in our understanding of the whole class of autoregressive models. For example, we have little knowledge about the potential of these models and their societal impacts in diverse linguistic and cultural settings. We alleviate ...
http://arxiv.org/pdf/2212.10755v2
2023-10-24T21:03:30
2022-12-21T04:21:46
{ "title": [ -0.018295288, -0.02734375, -0.011383057, -0.033721924, -0.022491455, 0.0043907166, -0.023910522, 0.0001654625, -0.010612488, -0.0079956055, -0.013298035, 0.007331848, -0.058929443, -0.043670654, -0.021469116, -0.025100708, 0.0035114288, ...
http://arxiv.org/abs/2212.10754v3
CoRRPUS: Code-based Structured Prompting for Neurosymbolic Story Understanding
Yijiang River Dong, Lara J. Martin, Chris Callison-Burch
cs.CL
Story generation and understanding -- as with all NLG/NLU tasks -- has seen a surge in neurosymbolic work. Researchers have recognized that, while large language models (LLMs) have tremendous utility, they can be augmented with symbolic means to be even better and to make up for any flaws that the neural networks might...
http://arxiv.org/pdf/2212.10754v3
2023-06-08T11:58:21
2022-12-21T04:21:35
{ "title": [ -0.02204895, -0.03302002, 0.00079107285, -0.0541687, -0.040985107, 0.0063056946, -0.04623413, 0.0071105957, 0.010887146, 0.001742363, -0.03768921, -0.015426636, 0.012626648, -0.040100098, -0.011672974, -0.014801025, 0.005428314, -0.0...
http://arxiv.org/abs/2212.10750v2
PropSegmEnt: A Large-Scale Corpus for Proposition-Level Segmentation and Entailment Recognition
Sihao Chen, Senaka Buthpitiya, Alex Fabrikant, Dan Roth, Tal Schuster
cs.CL
The widely studied task of Natural Language Inference (NLI) requires a system to recognize whether one piece of text is textually entailed by another, i.e. whether the entirety of its meaning can be inferred from the other. In current NLI datasets and models, textual entailment relations are typically defined on the se...
http://arxiv.org/pdf/2212.10750v2
2023-05-24T23:19:14
2022-12-21T04:03:33
{ "title": [ 0.0037975311, -0.010322571, -0.031311035, -0.09576416, 0.017822266, 0.018661499, -0.044952393, 0.008476257, 0.009162903, 0.01600647, -0.027114868, 0.014373779, 0.0047950745, -0.004337311, -0.031555176, 0.014862061, -0.014862061, 0.04...
http://arxiv.org/abs/2212.10740v1
ToL: A Tensor of List-Based Unified Computation Model
Hongxiao Li, Wanling Gao, Lei Wang, Jianfeng Zhan
cs.PL, cs.CC, cs.CL
Previous computation models either have equivalent abilities in representing all computations but fail to provide primitive operators for programming complex algorithms or lack generalized expression ability to represent newly-added computations. This article presents a unified computation model with generalized expres...
http://arxiv.org/pdf/2212.10740v1
2022-12-21T03:22:24
2022-12-21T03:22:24
{ "title": [ -0.016571045, 0.0016803741, -0.026321411, -0.06530762, 0.006629944, 0.005569458, -0.04031372, -0.01914978, 0.0021839142, 0.048706055, 0.021209717, 0.0076141357, 0.020339966, -0.039794922, -0.017837524, 0.040039062, -0.00075912476, -0...
http://arxiv.org/abs/2212.10728v1
Spoken Language Understanding for Conversational AI: Recent Advances and Future Direction
Soyeon Caren Han, Siqu Long, Henry Weld, Josiah Poon
cs.CL, cs.AI
When a human communicates with a machine using natural language on the web and online, how can it understand the human's intention and semantic context of their talk? This is an important AI task as it enables the machine to construct a sensible answer or perform a useful action for the human. Meaning is represented at...
http://arxiv.org/pdf/2212.10728v1
2022-12-21T02:47:52
2022-12-21T02:47:52
{ "title": [ -0.027374268, 0.000066936016, -0.028366089, -0.061309814, -0.0014257431, 0.0078086853, -0.06317139, 0.010421753, 0.007507324, 0.015380859, -0.022994995, -0.008918762, -0.0010032654, -0.01789856, -0.046325684, 0.00013828278, -0.013290405,...
http://arxiv.org/abs/2212.10726v2
Beyond Contrastive Learning: A Variational Generative Model for Multilingual Retrieval
John Wieting, Jonathan H. Clark, William W. Cohen, Graham Neubig, Taylor Berg-Kirkpatrick
cs.CL, cs.LG
Contrastive learning has been successfully used for retrieval of semantically aligned sentences, but it often requires large batch sizes or careful engineering to work well. In this paper, we instead propose a generative model for learning multilingual text embeddings which can be used to retrieve or score sentence pai...
http://arxiv.org/pdf/2212.10726v2
2023-06-04T22:42:10
2022-12-21T02:41:40
{ "title": [ 0.004371643, 0.0046577454, -0.022460938, -0.046691895, 0.0060768127, -0.0026283264, -0.036376953, -0.017684937, 0.025619507, 0.047790527, -0.004714966, 0.012023926, -0.0057411194, -0.0076446533, -0.06149292, 0.010047913, -0.009300232, ...
http://arxiv.org/abs/2212.10722v2
Contrastive Error Attribution for Finetuned Language Models
Faisal Ladhak, Esin Durmus, Tatsunori Hashimoto
cs.CL
Recent work has identified noisy and misannotated data as a core cause of hallucinations and unfaithful outputs in Natural Language Generation (NLG) tasks. Consequently, identifying and removing these examples is a key open challenge in creating reliable NLG systems. In this work, we introduce a framework to identify a...
http://arxiv.org/pdf/2212.10722v2
2023-07-11T17:06:19
2022-12-21T02:28:07
{ "title": [ 0.018539429, -0.01637268, -0.00793457, -0.054626465, 0.0024108887, -0.0041923523, -0.028579712, -0.015655518, 0.029205322, 0.034423828, -0.034423828, 0.0031337738, 0.016189575, -0.011207581, -0.05633545, 0.007156372, -0.04107666, 0.0...
http://arxiv.org/abs/2212.10720v2
MoralDial: A Framework to Train and Evaluate Moral Dialogue Systems via Moral Discussions
Hao Sun, Zhexin Zhang, Fei Mi, Yasheng Wang, Wei Liu, Jianwei Cui, Bin Wang, Qun Liu, Minlie Huang
cs.CL
Morality in dialogue systems has raised great attention in research recently. A moral dialogue system aligned with users' values could enhance conversation engagement and user connections. In this paper, we propose a framework, MoralDial to train and evaluate moral dialogue systems. In our framework, we first explore t...
http://arxiv.org/pdf/2212.10720v2
2023-05-26T14:27:14
2022-12-21T02:21:37
{ "title": [ -0.00920105, -0.001581192, -0.03994751, -0.039001465, 0.014343262, -0.018859863, -0.014671326, -0.02178955, -0.026672363, 0.016708374, -0.008880615, 0.007358551, -0.0071105957, -0.029281616, -0.025054932, -0.013206482, -0.021575928, ...
http://arxiv.org/abs/2212.10714v1
Integrating Heterogeneous Domain Information into Relation Extraction: A Case Study on Drug-Drug Interaction Extraction
Masaki Asada
cs.CL
The development of deep neural networks has improved representation learning in various domains, including textual, graph structural, and relational triple representations. This development opened the door to new relation extraction beyond the traditional text-oriented relation extraction. However, research on the effe...
http://arxiv.org/pdf/2212.10714v1
2022-12-21T01:26:07
2022-12-21T01:26:07
{ "title": [ 0.013931274, 0.0065841675, -0.076660156, -0.0892334, 0.036315918, -0.021713257, -0.0362854, 0.010429382, 0.009231567, 0.009666443, -0.023223877, 0.0008111, -0.025650024, -0.009819031, -0.038879395, -0.018875122, -0.022766113, 0.03305...
http://arxiv.org/abs/2212.10708v2
Zero-shot Triplet Extraction by Template Infilling
Bosung Kim, Hayate Iso, Nikita Bhutani, Estevam Hruschka, Ndapa Nakashole, Tom Mitchell
cs.CL
The task of triplet extraction aims to extract pairs of entities and their corresponding relations from unstructured text. Most existing methods train an extraction model on training data involving specific target relations, and are incapable of extracting new relations that were not observed at training time. Generali...
http://arxiv.org/pdf/2212.10708v2
2023-09-20T05:12:27
2022-12-21T00:57:24
{ "title": [ 0.024978638, 0.0070152283, -0.049316406, -0.091552734, 0.0077667236, -0.010940552, -0.050842285, -0.005378723, 0.029754639, 0.07287598, -0.0041236877, -0.003025055, 0.002664566, -0.012939453, -0.07299805, 0.018981934, 0.0124053955, 0...
http://arxiv.org/abs/2212.10707v1
Extractive Text Summarization Using Generalized Additive Models with Interactions for Sentence Selection
Vinícius Camargo da Silva, João Paulo Papa, Kelton Augusto Pontara da Costa
cs.CL, cs.LG
Automatic Text Summarization (ATS) is becoming relevant with the growth of textual data; however, with the popularization of public large-scale datasets, some recent machine learning approaches have focused on dense models and architectures that, despite producing notable results, usually turn out in models difficult t...
http://arxiv.org/pdf/2212.10707v1
2022-12-21T00:56:50
2022-12-21T00:56:50
{ "title": [ -0.009674072, 0.02508545, -0.01965332, -0.056854248, -0.0044517517, 0.015686035, -0.03894043, -0.024398804, 0.014503479, 0.043273926, -0.024368286, 0.0345459, -0.020111084, 0.007789612, -0.0748291, 0.00831604, -0.01878357, 0.02398681...
http://arxiv.org/abs/2212.10696v1
Analyzing Semantic Faithfulness of Language Models via Input Intervention on Conversational Question Answering
Akshay Chaturvedi, Swarnadeep Bhar, Soumadeep Saha, Utpal Garain, Nicholas Asher
cs.CL, cs.LG
Transformer-based language models have been shown to be highly effective for several NLP tasks. In this paper, we consider three transformer models, BERT, RoBERTa, and XLNet, in both small and large version, and investigate how faithful their representations are with respect to the semantic content of texts. We formali...
http://arxiv.org/pdf/2212.10696v1
2022-12-21T00:00:01
2022-12-21T00:00:01
{ "title": [ 0.0035324097, 0.00071668625, -0.009529114, -0.06298828, -0.010391235, 0.002500534, -0.06402588, -0.01626587, 0.00060129166, 0.03378296, -0.0095825195, 0.015930176, 0.022842407, -0.0071792603, -0.050201416, 0.018051147, -0.014533997, ...
http://arxiv.org/abs/2212.10692v1
Generation-Augmented Query Expansion For Code Retrieval
Dong Li, Yelong Shen, Ruoming Jin, Yi Mao, Kuan Wang, Weizhu Chen
cs.SE, cs.AI, cs.CL
Pre-trained language models have achieved promising success in code retrieval tasks, where a natural language documentation query is given to find the most relevant existing code snippet. However, existing models focus only on optimizing the documentation code pairs by embedding them into latent space, without the asso...
http://arxiv.org/pdf/2212.10692v1
2022-12-20T23:49:37
2022-12-20T23:49:37
{ "title": [ -0.008026123, -0.021011353, -0.03543091, -0.05770874, 0.01247406, -0.0115356445, -0.002696991, 0.024536133, 0.02053833, 0.050354004, -0.032287598, -0.00097227097, 0.0037670135, -0.002670288, -0.06591797, 0.0047035217, -0.004108429, 0...
http://arxiv.org/abs/2212.10690v1
METEOR Guided Divergence for Video Captioning
Daniel Lukas Rothenpieler, Shahin Amiriparian
cs.CV, cs.CL, cs.LG, I.2.10
Automatic video captioning aims for a holistic visual scene understanding. It requires a mechanism for capturing temporal context in video frames and the ability to comprehend the actions and associations of objects in a given timeframe. Such a system should additionally learn to abstract video sequences into sensible ...
http://arxiv.org/pdf/2212.10690v1
2022-12-20T23:30:47
2022-12-20T23:30:47
{ "title": [ -0.003786087, -0.009048462, -0.012329102, -0.03579712, -0.044281006, 0.0036468506, -0.037353516, -0.010421753, -0.007873535, 0.076538086, -0.06323242, -0.0076408386, 0.061645508, 0.0121536255, -0.06201172, -0.006000519, 0.0040016174, ...
http://arxiv.org/abs/2212.10678v1
Understanding Stereotypes in Language Models: Towards Robust Measurement and Zero-Shot Debiasing
Justus Mattern, Zhijing Jin, Mrinmaya Sachan, Rada Mihalcea, Bernhard Schölkopf
cs.CL, cs.LG
Generated texts from large pretrained language models have been shown to exhibit a variety of harmful, human-like biases about various demographics. These findings prompted large efforts aiming to understand and measure such effects, with the goal of providing benchmarks that can guide the development of techniques mit...
http://arxiv.org/pdf/2212.10678v1
2022-12-20T22:41:24
2022-12-20T22:41:24
{ "title": [ -0.018112183, 0.0018129349, -0.040008545, -0.042663574, 0.002779007, -0.008422852, -0.04071045, -0.019058228, 0.009475708, 0.029449463, -0.024932861, -0.009216309, 0.0060424805, -0.015129089, -0.053894043, -0.001783371, -0.03439331, ...
http://arxiv.org/abs/2212.10670v1
In-context Learning Distillation: Transferring Few-shot Learning Ability of Pre-trained Language Models
Yukun Huang, Yanda Chen, Zhou Yu, Kathleen McKeown
cs.CL, cs.LG
Given the success with in-context learning of large pre-trained language models, we introduce in-context learning distillation to transfer in-context few-shot learning ability from large models to smaller models. We propose to combine in-context learning objectives with language modeling objectives to distill both the ...
http://arxiv.org/pdf/2212.10670v1
2022-12-20T22:11:35
2022-12-20T22:11:35
{ "title": [ 0.011619568, -0.014503479, -0.034210205, -0.04034424, -0.01247406, -0.022079468, -0.01159668, 0.0078048706, -0.0016126633, 0.023483276, -0.03756714, 0.003206253, 0.008880615, -0.028549194, -0.01852417, -0.020065308, 0.0038223267, 0.0...
http://arxiv.org/abs/2212.10650v1
KronA: Parameter Efficient Tuning with Kronecker Adapter
Ali Edalati, Marzieh Tahaei, Ivan Kobyzev, Vahid Partovi Nia, James J. Clark, Mehdi Rezagholizadeh
cs.CL
Fine-tuning a Pre-trained Language Model (PLM) on a specific downstream task has been a well-known paradigm in Natural Language Processing. However, with the ever-growing size of PLMs, training the entire model on several downstream tasks becomes very expensive and resource-hungry. Recently, different Parameter Efficie...
http://arxiv.org/pdf/2212.10650v1
2022-12-20T20:56:52
2022-12-20T20:56:52
{ "title": [ -0.045196533, -0.027572632, -0.0340271, -0.08251953, 0.003004074, 0.011436462, -0.046142578, -0.005748749, -0.013214111, 0.033081055, -0.008079529, 0.012809753, 0.029586792, -0.028076172, -0.01828003, 0.020187378, -0.0020580292, 0.03...
http://arxiv.org/abs/2212.10622v1
mFACE: Multilingual Summarization with Factual Consistency Evaluation
Roee Aharoni, Shashi Narayan, Joshua Maynez, Jonathan Herzig, Elizabeth Clark, Mirella Lapata
cs.CL
Abstractive summarization has enjoyed renewed interest in recent years, thanks to pre-trained language models and the availability of large-scale datasets. Despite promising results, current models still suffer from generating factually inconsistent summaries, reducing their utility for real-world application. Several ...
http://arxiv.org/pdf/2212.10622v1
2022-12-20T19:52:41
2022-12-20T19:52:41
{ "title": [ -0.023208618, -0.0015621185, -0.037109375, -0.07800293, 0.00040841103, 0.015426636, -0.06994629, -0.008956909, -0.007129669, 0.06439209, 0.010673523, 0.008766174, 0.0206604, -0.0045776367, -0.0579834, 0.020858765, -0.01687622, 0.0522...
http://arxiv.org/abs/2212.10618v2
Ontologically Faithful Generation of Non-Player Character Dialogues
Nathaniel Weir, Ryan Thomas, Randolph D'Amore, Kellie Hill, Benjamin Van Durme, Harsh Jhamtani
cs.CL
We introduce a language generation task grounded in a popular video game environment. KNUDGE (KNowledge Constrained User-NPC Dialogue GEneration) requires models to produce trees of dialogue between video game characters that accurately reflect quest and entity specifications stated in natural language. KNUDGE is const...
http://arxiv.org/pdf/2212.10618v2
2023-05-13T19:20:17
2022-12-20T19:48:10
{ "title": [ -0.005622864, -0.0096588135, -0.019897461, -0.05596924, -0.018203735, 0.012374878, -0.050079346, -0.009399414, 0.0008568764, 0.025222778, -0.0002527237, 0.0035495758, -0.0041618347, -0.029312134, -0.02078247, 0.0064048767, -0.038116455, ...
http://arxiv.org/abs/2212.11680v2
Smooth Sailing: Improving Active Learning for Pre-trained Language Models with Representation Smoothness Analysis
Josip Jukić, Jan Šnajder
cs.LG, cs.CL
Developed to alleviate prohibitive labeling costs, active learning (AL) methods aim to reduce label complexity in supervised learning. While recent work has demonstrated the benefit of using AL in combination with large pre-trained language models (PLMs), it has often overlooked the practical challenges that hinder the...
http://arxiv.org/pdf/2212.11680v2
2023-10-23T14:05:49
2022-12-20T19:37:20
{ "title": [ 0.0012187958, 0.004825592, -0.03265381, -0.074645996, 0.024627686, -0.007385254, -0.033325195, 0.0064468384, 0.006893158, 0.040130615, -0.0008234978, 0.011650085, 0.010047913, -0.023406982, -0.029541016, 0.020462036, 0.00024151802, 0...
http://arxiv.org/abs/2212.10564v2
A Vision-free Baseline for Multimodal Grammar Induction
Boyi Li, Rodolfo Corona, Karttikeya Mangalam, Catherine Chen, Daniel Flaherty, Serge Belongie, Kilian Q. Weinberger, Jitendra Malik, Trevor Darrell, Dan Klein
cs.CL, cs.AI, cs.LG
Past work has shown that paired vision-language signals substantially improve grammar induction in multimodal datasets such as MSCOCO. We investigate whether advancements in large language models (LLMs) that are only trained with text could provide strong assistance for grammar induction in multimodal settings. We find...
http://arxiv.org/pdf/2212.10564v2
2023-10-31T17:22:17
2022-12-20T18:59:50
{ "title": [ -0.014144897, -0.0085372925, -0.04660034, -0.06817627, -0.010253906, -0.010360718, -0.051605225, 0.03050232, 0.01763916, 0.0134887695, -0.045562744, 0.01802063, -0.020248413, -0.00881958, -0.036834717, -0.0063591003, -0.019302368, 0....
http://arxiv.org/abs/2212.10563v2
BLIND: Bias Removal With No Demographics
Hadas Orgad, Yonatan Belinkov
cs.CL, 68T50, I.2.7
Models trained on real-world data tend to imitate and amplify social biases. Common methods to mitigate biases require prior information on the types of biases that should be mitigated (e.g., gender or racial bias) and the social groups associated with each data sample. In this work, we introduce BLIND, a method for bi...
http://arxiv.org/pdf/2212.10563v2
2023-06-12T03:55:12
2022-12-20T18:59:42
{ "title": [ 0.045166016, 0.032928467, -0.036621094, -0.03741455, 0.0017738342, -0.024963379, -0.018051147, -0.0030212402, 0.0057029724, 0.027496338, -0.026916504, -0.02130127, 0.010559082, -0.033050537, -0.019256592, 0.005290985, -0.0074691772, ...
http://arxiv.org/abs/2212.10561v3
Parsel: Algorithmic Reasoning with Language Models by Composing Decompositions
Eric Zelikman, Qian Huang, Gabriel Poesia, Noah D. Goodman, Nick Haber
cs.CL, cs.AI, cs.LG
Despite recent success in large language model (LLM) reasoning, LLMs struggle with hierarchical multi-step reasoning tasks like generating complex programs. For these tasks, humans often start with a high-level algorithmic design and implement each part gradually. We introduce Parsel, a framework enabling automatic imp...
http://arxiv.org/pdf/2212.10561v3
2023-05-28T21:12:31
2022-12-20T18:59:23
{ "title": [ -0.045928955, -0.034088135, -0.03338623, -0.09301758, -0.0050086975, 0.038482666, -0.083984375, 0.021636963, 0.0047721863, 0.018722534, -0.008323669, -0.004558563, 0.0004093647, -0.019897461, -0.028015137, 0.0039520264, -0.015670776, ...
http://arxiv.org/abs/2212.10562v2
Character-Aware Models Improve Visual Text Rendering
Rosanne Liu, Dan Garrette, Chitwan Saharia, William Chan, Adam Roberts, Sharan Narang, Irina Blok, RJ Mical, Mohammad Norouzi, Noah Constant
cs.CL, cs.CV
Current image generation models struggle to reliably produce well-formed visual text. In this paper, we investigate a key contributing factor: popular text-to-image models lack character-level input features, making it much harder to predict a word's visual makeup as a series of glyphs. To quantify this effect, we cond...
http://arxiv.org/pdf/2212.10562v2
2023-05-03T16:36:38
2022-12-20T18:59:23
{ "title": [ -0.0211792, -0.0015192032, -0.014862061, -0.04107666, -0.025512695, 0.008766174, -0.046813965, -0.0101623535, 0.024917603, 0.05303955, -0.029785156, -0.010856628, 0.039398193, -0.003818512, -0.060791016, 0.013092041, -0.004634857, 0....
http://arxiv.org/abs/2212.10560v2
Self-Instruct: Aligning Language Models with Self-Generated Instructions
Yizhong Wang, Yeganeh Kordi, Swaroop Mishra, Alisa Liu, Noah A. Smith, Daniel Khashabi, Hannaneh Hajishirzi
cs.CL, cs.AI
Large "instruction-tuned" language models (i.e., finetuned to respond to instructions) have demonstrated a remarkable ability to generalize zero-shot to new tasks. Nevertheless, they depend heavily on human-written instruction data that is often limited in quantity, diversity, and creativity, therefore hindering the ge...
http://arxiv.org/pdf/2212.10560v2
2023-05-25T23:50:07
2022-12-20T18:59:19
{ "title": [ -0.036712646, -0.014892578, -0.04296875, -0.06854248, -0.005355835, 0.01878357, -0.07910156, 0.0023403168, -0.0129776, 0.035858154, 0.00248909, 0.0023479462, -0.012367249, -0.016204834, -0.04849243, 0.012413025, 0.0014896393, 0.02769...
http://arxiv.org/abs/2212.10559v3
Why Can GPT Learn In-Context? Language Models Implicitly Perform Gradient Descent as Meta-Optimizers
Damai Dai, Yutao Sun, Li Dong, Yaru Hao, Shuming Ma, Zhifang Sui, Furu Wei
cs.CL
Large pretrained language models have shown surprising in-context learning (ICL) ability. With a few demonstration input-label pairs, they can predict the label for an unseen input without parameter updates. Despite the great success in performance, its working mechanism still remains an open question. In this paper, w...
http://arxiv.org/pdf/2212.10559v3
2023-05-15T11:45:12
2022-12-20T18:58:48
{ "title": [ 0.0018825531, 0.022598267, -0.0023994446, -0.021194458, -0.02583313, 0.0031986237, -0.059936523, -0.04034424, -0.011894226, 0.048553467, -0.053894043, -0.0015039444, 0.009048462, -0.00484848, -0.00274086, -0.031555176, -0.003490448, ...
http://arxiv.org/abs/2212.10558v1
On-the-fly Denoising for Data Augmentation in Natural Language Understanding
Tianqing Fang, Wenxuan Zhou, Fangyu Liu, Hongming Zhang, Yangqiu Song, Muhao Chen
cs.CL, cs.AI
Data Augmentation (DA) is frequently used to automatically provide additional training data without extra human annotation. However, data augmentation may introduce noisy data that impairs training. To guarantee the quality of augmented data, existing methods either assume no noise exists in the augmented data and adop...
http://arxiv.org/pdf/2212.10558v1
2022-12-20T18:58:33
2022-12-20T18:58:33
{ "title": [ -0.0014219284, -0.00038647652, -0.061065674, -0.08081055, 0.011680603, -0.013748169, -0.032806396, 0.021484375, 0.028869629, 0.015853882, -0.049041748, -0.0087509155, 0.006504059, -0.0017604828, -0.061767578, -0.009216309, -0.0023536682,...
http://arxiv.org/abs/2212.10557v2
DialGuide: Aligning Dialogue Model Behavior with Developer Guidelines
Prakhar Gupta, Yang Liu, Di Jin, Behnam Hedayatnia, Spandana Gella, Sijia Liu, Patrick Lange, Julia Hirschberg, Dilek Hakkani-Tur
cs.CL
Dialogue models are able to generate coherent and fluent responses, but they can still be challenging to control and may produce non-engaging, unsafe results. This unpredictability diminishes user trust and can hinder the use of the models in the real world. To address this, we introduce DialGuide, a novel framework fo...
http://arxiv.org/pdf/2212.10557v2
2023-05-21T19:38:01
2022-12-20T18:57:18
{ "title": [ -0.03161621, -0.026916504, -0.046203613, -0.06298828, -0.0010099411, -0.0012254715, -0.05331421, 0.005874634, -0.013046265, 0.032226562, 0.005279541, 0.008430481, -0.012916565, -0.030532837, -0.041290283, 0.0011577606, -0.013626099, ...
http://arxiv.org/abs/2212.10555v1
PairReranker: Pairwise Reranking for Natural Language Generation
Dongfu Jiang, Bill Yuchen Lin, Xiang Ren
cs.CL, cs.AI, cs.LG
Pre-trained language models have been successful in natural language generation (NLG) tasks. While various decoding methods have been employed, they often produce suboptimal results. We first present an empirical analysis of three NLG tasks: summarization, machine translation, and constrained text generation. We found ...
http://arxiv.org/pdf/2212.10555v1
2022-12-20T18:56:57
2022-12-20T18:56:57
{ "title": [ -0.03765869, -0.018966675, -0.02911377, -0.047302246, -0.00522995, 0.025970459, -0.06185913, -0.02104187, 0.0064849854, 0.028884888, -0.0057868958, 0.0085372925, -0.0004003048, -0.011756897, -0.03878784, 0.0066375732, -0.027709961, 0...
http://arxiv.org/abs/2212.10554v1
A Length-Extrapolatable Transformer
Yutao Sun, Li Dong, Barun Patra, Shuming Ma, Shaohan Huang, Alon Benhaim, Vishrav Chaudhary, Xia Song, Furu Wei
cs.CL
Position modeling plays a critical role in Transformers. In this paper, we focus on length extrapolation, i.e., training on short texts while evaluating longer sequences. We define attention resolution as an indicator of extrapolation. Then we propose two designs to improve the above metric of Transformers. Specificall...
http://arxiv.org/pdf/2212.10554v1
2022-12-20T18:56:20
2022-12-20T18:56:20
{ "title": [ 0.031311035, -0.009307861, -0.0031356812, -0.0039863586, -0.0569458, -0.026657104, -0.011024475, -0.025222778, -0.05404663, 0.039215088, -0.035064697, 0.015525818, -0.016067505, -0.023498535, 0.028457642, -0.003929138, 0.03353882, 0....
http://arxiv.org/abs/2212.10551v3
Lego-MT: Learning Detachable Models for Massively Multilingual Machine Translation
Fei Yuan, Yinquan Lu, WenHao Zhu, Lingpeng Kong, Lei Li, Yu Qiao, Jingjing Xu
cs.CL, cs.AI
Multilingual neural machine translation (MNMT) aims to build a unified model for many language directions. Existing monolithic models for MNMT encounter two challenges: parameter interference among languages and inefficient inference for large models. In this paper, we revisit the classic multi-way structures and devel...
http://arxiv.org/pdf/2212.10551v3
2023-07-19T05:52:32
2022-12-20T18:54:08
{ "title": [ -0.028945923, -0.015457153, -0.012321472, -0.062194824, -0.013496399, 0.034423828, -0.0546875, 0.016357422, 0.0022411346, 0.04071045, -0.011917114, 0.010284424, 0.014282227, -0.0093307495, -0.033355713, 0.020843506, -0.0077438354, 0....
http://arxiv.org/abs/2212.10549v2
Cross-modal Attention Congruence Regularization for Vision-Language Relation Alignment
Rohan Pandey, Rulin Shao, Paul Pu Liang, Ruslan Salakhutdinov, Louis-Philippe Morency
cs.CL, cs.CV, cs.LG
Despite recent progress towards scaling up multimodal vision-language models, these models are still known to struggle on compositional generalization benchmarks such as Winoground. We find that a critical component lacking from current vision-language models is relation-level alignment: the ability to match directiona...
http://arxiv.org/pdf/2212.10549v2
2023-07-04T13:18:29
2022-12-20T18:53:14
{ "title": [ -0.010879517, 0.0002117157, -0.053741455, -0.056243896, -0.0055389404, -0.015327454, -0.04748535, 0.0015182495, 0.04940796, 0.04840088, 0.00021088123, -0.016921997, 0.044677734, -0.020401001, -0.056030273, 0.023880005, 0.009605408, 0...
http://arxiv.org/abs/2212.10548v2
T-Projection: High Quality Annotation Projection for Sequence Labeling Tasks
Iker García-Ferrero, Rodrigo Agerri, German Rigau
cs.CL
In the absence of readily available labeled data for a given sequence labeling task and language, annotation projection has been proposed as one of the possible strategies to automatically generate annotated data. Annotation projection has often been formulated as the task of transporting, on parallel corpora, the labe...
http://arxiv.org/pdf/2212.10548v2
2023-10-24T10:49:58
2022-12-20T18:51:48
{ "title": [ -0.005168915, 0.0005092621, -0.030334473, -0.08892822, 0.0019044876, -0.00015330315, -0.050201416, -0.0035877228, -0.0036716461, 0.034942627, -0.012588501, 0.00504303, -0.0039787292, -0.023605347, -0.036590576, 0.0073127747, -0.000390768...
http://arxiv.org/abs/2212.10547v2
Semantically-informed Hierarchical Event Modeling
Shubhashis Roy Dipta, Mehdi Rezaee, Francis Ferraro
cs.CL
Prior work has shown that coupling sequential latent variable models with semantic ontological knowledge can improve the representational capabilities of event modeling approaches. In this work, we present a novel, doubly hierarchical, semi-supervised event modeling framework that provides structural hierarchy while al...
http://arxiv.org/pdf/2212.10547v2
2023-05-30T17:47:55
2022-12-20T18:51:23
{ "title": [ -0.0045394897, 0.010894775, -0.058685303, -0.13659668, 0.03237915, 0.012992859, -0.10003662, -0.0057411194, -0.0005412102, 0.040161133, 0.006046295, 0.0072784424, 0.016616821, -0.03338623, -0.025146484, 0.03164673, -0.0059165955, 0.0...
http://arxiv.org/abs/2212.10545v2
DimonGen: Diversified Generative Commonsense Reasoning for Explaining Concept Relationships
Chenzhengyi Liu, Jie Huang, Kerui Zhu, Kevin Chen-Chuan Chang
cs.CL, cs.AI
In this paper, we propose DimonGen, which aims to generate diverse sentences describing concept relationships in various everyday scenarios. To support this, we first create a benchmark dataset for this task by adapting the existing CommonGen dataset. We then propose a two-stage model called MoREE to generate the targe...
http://arxiv.org/pdf/2212.10545v2
2023-05-16T06:32:51
2022-12-20T18:50:29
{ "title": [ 0.008918762, -0.036315918, -0.036346436, -0.08648682, -0.0070114136, -0.003929138, -0.03640747, 0.0050086975, 0.047424316, -0.0052261353, -0.0178833, 0.0049095154, -0.0149383545, -0.035186768, -0.020233154, 0.0056915283, -0.023544312, ...
http://arxiv.org/abs/2212.10544v2
Pretraining Without Attention
Junxiong Wang, Jing Nathan Yan, Albert Gu, Alexander M. Rush
cs.CL, cs.LG
Transformers have been essential to pretraining success in NLP. While other architectures have been used, downstream accuracy is either significantly worse, or requires attention layers to match standard benchmarks such as GLUE. This work explores pretraining without attention by using recent advances in sequence routi...
http://arxiv.org/pdf/2212.10544v2
2023-05-09T01:08:20
2022-12-20T18:50:08
{ "title": [ 0.06463623, -0.0025997162, -0.014701843, -0.055145264, -0.007549286, -0.013008118, -0.00059890747, 0.005886078, -0.012084961, 0.019210815, -0.0055999756, -0.007724762, -0.042663574, -0.037872314, 0.029708862, -0.014305115, 0.032928467, ...
http://arxiv.org/abs/2212.10543v2
Detoxifying Text with MaRCo: Controllable Revision with Experts and Anti-Experts
Skyler Hallinan, Alisa Liu, Yejin Choi, Maarten Sap
cs.CL, cs.AI
Text detoxification has the potential to mitigate the harms of toxicity by rephrasing text to remove offensive meaning, but subtle toxicity remains challenging to tackle. We introduce MaRCo, a detoxification algorithm that combines controllable generation and text rewriting methods using a Product of Experts with autoe...
http://arxiv.org/pdf/2212.10543v2
2023-05-26T20:26:06
2022-12-20T18:50:00
{ "title": [ 0.01210022, 0.018844604, -0.05670166, -0.020584106, -0.001250267, -0.022521973, 0.0040245056, -0.014671326, -0.027862549, 0.03475952, -0.013679504, -0.0005207062, 0.038360596, -0.036895752, 0.0069999695, -0.0072402954, -0.00022387505, ...
http://arxiv.org/abs/2212.10539v1
Toward Human Readable Prompt Tuning: Kubrick's The Shining is a good movie, and a good prompt too?
Weijia Shi, Xiaochuang Han, Hila Gonen, Ari Holtzman, Yulia Tsvetkov, Luke Zettlemoyer
cs.CL
Large language models can perform new tasks in a zero-shot fashion, given natural language prompts that specify the desired behavior. Such prompts are typically hand engineered, but can also be learned with gradient-based methods from labeled data. However, it is underexplored what factors make the prompts effective, e...
http://arxiv.org/pdf/2212.10539v1
2022-12-20T18:47:13
2022-12-20T18:47:13
{ "title": [ 0.0048179626, -0.02357483, 0.0146865845, 0.01864624, -0.036621094, -0.01033783, 0.034301758, -0.022827148, -0.06585693, 0.017120361, -0.018447876, -0.0075645447, -0.01461792, -0.03466797, 0.019042969, -0.06518555, -0.017364502, 0.006...
http://arxiv.org/abs/2212.10537v2
Does CLIP Bind Concepts? Probing Compositionality in Large Image Models
Martha Lewis, Nihal V. Nayak, Peilin Yu, Qinan Yu, Jack Merullo, Stephen H. Bach, Ellie Pavlick
cs.CV, cs.AI, cs.CL
Large-scale neural network models combining text and images have made incredible progress in recent years. However, it remains an open question to what extent such models encode compositional representations of the concepts over which they operate, such as correctly identifying ''red cube'' by reasoning over the consti...
http://arxiv.org/pdf/2212.10537v2
2023-03-29T15:34:23
2022-12-20T18:46:28
{ "title": [ 0.0015420914, 0.016830444, -0.021453857, -0.05618286, 0.011451721, -0.0043029785, -0.04724121, -0.03555298, 0.0045700073, 0.036315918, -0.0087509155, 0.00053167343, -0.0049858093, -0.015144348, -0.04119873, -0.0034561157, 0.0054397583, ...
http://arxiv.org/abs/2212.10536v1
Measure More, Question More: Experimental Studies on Transformer-based Language Models and Complement Coercion
Yuling Gu
cs.CL
Transformer-based language models have shown strong performance on an array of natural language understanding tasks. However, the question of how these models react to implicit meaning has been largely unexplored. We investigate this using the complement coercion phenomenon, which involves sentences like "The student f...
http://arxiv.org/pdf/2212.10536v1
2022-12-20T18:46:20
2022-12-20T18:46:20
{ "title": [ -0.03491211, 0.009155273, -0.02116394, -0.0013504028, -0.031051636, -0.014312744, -0.048980713, -0.03918457, -0.02357483, 0.028564453, -0.016021729, 0.020889282, 0.0026435852, -0.023223877, -0.02671814, -0.0055389404, -0.023376465, 0...
http://arxiv.org/abs/2212.10535v2
A Survey of Deep Learning for Mathematical Reasoning
Pan Lu, Liang Qiu, Wenhao Yu, Sean Welleck, Kai-Wei Chang
cs.AI, cs.CL, cs.CV, cs.LG
Mathematical reasoning is a fundamental aspect of human intelligence and is applicable in various fields, including science, engineering, finance, and everyday life. The development of artificial intelligence (AI) systems capable of solving math problems and proving theorems has garnered significant interest in the fie...
http://arxiv.org/pdf/2212.10535v2
2023-06-22T01:37:02
2022-12-20T18:46:16
{ "title": [ -0.01448822, 0.011268616, -0.042907715, -0.051727295, 0.024154663, -0.014472961, -0.007434845, 0.00003129244, 0.021148682, 0.009407043, -0.015007019, -0.003780365, 0.0011129379, -0.026016235, -0.023025513, 0.0048332214, -0.002117157, ...
http://arxiv.org/abs/2212.10534v3
DISCO: Distilling Counterfactuals with Large Language Models
Zeming Chen, Qiyue Gao, Antoine Bosselut, Ashish Sabharwal, Kyle Richardson
cs.CL
Models trained with counterfactually augmented data learn representations of the causal structure of tasks, enabling robust generalization. However, high-quality counterfactual data is scarce for most tasks and not easily generated at scale. When crowdsourced, such data is typically limited in scale and diversity; when...
http://arxiv.org/pdf/2212.10534v3
2023-06-05T19:16:25
2022-12-20T18:46:08
{ "title": [ 0.0076522827, -0.016784668, -0.021194458, -0.05734253, -0.0007138252, 0.0014343262, -0.0446167, -0.026412964, 0.023162842, 0.017349243, -0.02861023, -0.0025234222, -0.0015830994, -0.018920898, -0.04748535, 0.0019874573, -0.020492554, ...
http://arxiv.org/abs/2212.11672v1
Trustworthy Social Bias Measurement
Rishi Bommasani, Percy Liang
cs.CL
How do we design measures of social bias that we trust? While prior work has introduced several measures, no measure has gained widespread trust: instead, mounting evidence argues we should distrust these measures. In this work, we design bias measures that warrant trust based on the cross-disciplinary theory of measur...
http://arxiv.org/pdf/2212.11672v1
2022-12-20T18:45:12
2022-12-20T18:45:12
{ "title": [ 0.0036754608, 0.0181427, -0.04638672, -0.024383545, -0.0047454834, -0.030532837, -0.00198555, -0.04031372, -0.02355957, 0.02748108, -0.04046631, -0.0067253113, -0.0152282715, -0.0357666, -0.005657196, -0.009353638, -0.01096344, 0.021...
http://arxiv.org/abs/2212.10529v2
Does GPT-3 Demonstrate Psychopathy? Evaluating Large Language Models from a Psychological Perspective
Xingxuan Li, Yutong Li, Shafiq Joty, Linlin Liu, Fei Huang, Lin Qiu, Lidong Bing
cs.CL, cs.AI, cs.CY
In this work, we determined whether large language models (LLMs) are psychologically safe. We designed unbiased prompts to systematically evaluate LLMs from a psychological perspective. First, we tested three different LLMs by using two personality tests: Short Dark Triad (SD-3) and Big Five Inventory (BFI). All models...
http://arxiv.org/pdf/2212.10529v2
2023-05-08T16:52:43
2022-12-20T18:45:07
{ "title": [ -0.023208618, 0.013908386, -0.010620117, -0.010597229, -0.008956909, -0.007045746, -0.024337769, -0.032165527, -0.015571594, 0.027420044, -0.011039734, 0.0024795532, 0.0031967163, -0.03665161, -0.019546509, 0.0020599365, -0.03010559, ...
http://arxiv.org/abs/2212.10528v1
HYRR: Hybrid Infused Reranking for Passage Retrieval
Jing Lu, Keith Hall, Ji Ma, Jianmo Ni
cs.CL, cs.IR
We present Hybrid Infused Reranking for Passages Retrieval (HYRR), a framework for training rerankers based on a hybrid of BM25 and neural retrieval models. Retrievers based on hybrid models have been shown to outperform both BM25 and neural models alone. Our approach exploits this improved performance when training a ...
http://arxiv.org/pdf/2212.10528v1
2022-12-20T18:44:21
2022-12-20T18:44:21
{ "title": [ -0.008613586, -0.005126953, -0.02949524, -0.059906006, 0.0027713776, 0.0053253174, -0.034454346, -0.029815674, 0.015411377, 0.04449463, -0.03225708, 0.026626587, 0.018814087, 0.0048103333, -0.040008545, 0.013465881, -0.0077819824, 0....
http://arxiv.org/abs/2212.10526v3
Open Domain Multi-document Summarization: A Comprehensive Study of Model Brittleness under Retrieval
John Giorgi, Luca Soldaini, Bo Wang, Gary Bader, Kyle Lo, Lucy Lu Wang, Arman Cohan
cs.CL, cs.AI
Multi-document summarization (MDS) assumes a set of topic-related documents are provided as input. In practice, this document set is not always available; it would need to be retrieved given an information need, i.e. a question or topic statement, a setting we dub "open-domain" MDS. We study this more challenging setti...
http://arxiv.org/pdf/2212.10526v3
2023-10-25T13:25:20
2022-12-20T18:41:38
{ "title": [ 0.0059127808, 0.018356323, -0.006324768, -0.068725586, -0.006500244, 0.018234253, -0.062072754, -0.034332275, -0.0019264221, 0.05593872, -0.005382538, 0.031341553, 0.025238037, -0.0020771027, -0.03640747, 0.010803223, -0.002626419, -...
http://arxiv.org/abs/2212.10525v2
SLUE Phase-2: A Benchmark Suite of Diverse Spoken Language Understanding Tasks
Suwon Shon, Siddhant Arora, Chyi-Jiunn Lin, Ankita Pasad, Felix Wu, Roshan Sharma, Wei-Lun Wu, Hung-Yi Lee, Karen Livescu, Shinji Watanabe
cs.CL, eess.AS
Spoken language understanding (SLU) tasks have been studied for many decades in the speech research community, but have not received as much attention as lower-level tasks like speech and speaker recognition. In particular, there are not nearly as many SLU task benchmarks, and many of the existing ones use data that is...
http://arxiv.org/pdf/2212.10525v2
2023-06-15T22:51:09
2022-12-20T18:39:59
{ "title": [ -0.02204895, -0.005382538, -0.033447266, -0.07183838, 0.003112793, 0.00137043, -0.07183838, 0.011238098, -0.009315491, 0.015586853, -0.006954193, 0.001698494, -0.0008263588, -0.021636963, -0.03665161, 0.00047707558, -0.006500244, 0.0...
http://arxiv.org/abs/2212.10522v1
Transformers Go for the LOLs: Generating (Humourous) Titles from Scientific Abstracts End-to-End
Yanran Chen, Steffen Eger
cs.CL
We consider the end-to-end abstract-to-title generation problem, exploring seven recent transformer based models (including ChatGPT) fine-tuned on more than 30k abstract-title pairs from NLP and machine learning venues. As an extension, we also consider the harder problem of generating humorous paper titles. For the la...
http://arxiv.org/pdf/2212.10522v1
2022-12-20T18:37:11
2022-12-20T18:37:11
{ "title": [ -0.016204834, -0.0068244934, -0.03894043, -0.040100098, -0.0065231323, -0.012794495, -0.030349731, -0.020599365, 0.0012683868, 0.017623901, 0.008399963, 0.010643005, 0.010894775, -0.026260376, -0.03363037, 0.004184723, -0.0112838745, ...
http://arxiv.org/abs/2212.10711v1
Task Ambiguity in Humans and Language Models
Alex Tamkin, Kunal Handa, Avash Shrestha, Noah Goodman
cs.CL, cs.LG
Language models have recently achieved strong performance across a wide range of NLP benchmarks. However, unlike benchmarks, real world tasks are often poorly specified, and agents must deduce the user's intended behavior from a combination of context, instructions, and examples. We investigate how both humans and mode...
http://arxiv.org/pdf/2212.10711v1
2022-12-20T18:35:33
2022-12-20T18:35:33
{ "title": [ -0.0491333, 0.009170532, -0.051208496, -0.019119263, -0.008514404, -0.01574707, -0.050231934, -0.0178833, -0.010604858, 0.033569336, 0.0024280548, -0.024993896, 0.0002951622, -0.03024292, -0.049713135, 0.00187397, -0.010093689, 0.027...
http://arxiv.org/abs/2212.10520v3
Privacy-Preserving Domain Adaptation of Semantic Parsers
Fatemehsadat Mireshghallah, Yu Su, Tatsunori Hashimoto, Jason Eisner, Richard Shin
cs.CL
Task-oriented dialogue systems often assist users with personal or confidential matters. For this reason, the developers of such a system are generally prohibited from observing actual usage. So how can they know where the system is failing and needs more training data or new functionality? In this work, we study ways ...
http://arxiv.org/pdf/2212.10520v3
2023-06-08T18:03:43
2022-12-20T18:35:21
{ "title": [ -0.031051636, -0.0024604797, -0.06951904, -0.0713501, 0.021484375, -0.01600647, -0.051208496, 0.02470398, -0.03225708, 0.014350891, -0.003873825, 0.00944519, 0.005622864, -0.017333984, -0.009239197, -0.0076942444, -0.021469116, 0.038...
http://arxiv.org/abs/2212.10515v2
CausalDialogue: Modeling Utterance-level Causality in Conversations
Yi-Lin Tuan, Alon Albalak, Wenda Xu, Michael Saxon, Connor Pryor, Lise Getoor, William Yang Wang
cs.CL
Despite their widespread adoption, neural conversation models have yet to exhibit natural chat capabilities with humans. In this research, we examine user utterances as causes and generated responses as effects, recognizing that changes in a cause should produce a different effect. To further explore this concept, we h...
http://arxiv.org/pdf/2212.10515v2
2023-07-08T21:59:43
2022-12-20T18:31:50
{ "title": [ -0.0034980774, -0.008804321, -0.040161133, -0.08935547, 0.010894775, -0.001865387, -0.06072998, 0.0030937195, 0.010925293, 0.026611328, 0.0011987686, 0.0065612793, 0.018814087, -0.020828247, -0.03878784, 0.009284973, -0.030166626, 0....
http://arxiv.org/abs/2212.10511v4
When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parametric Memories
Alex Mallen, Akari Asai, Victor Zhong, Rajarshi Das, Daniel Khashabi, Hannaneh Hajishirzi
cs.CL, cs.AI, cs.LG
Despite their impressive performance on diverse tasks, large language models (LMs) still struggle with tasks requiring rich world knowledge, implying the limitations of relying solely on their parameters to encode a wealth of world knowledge. This paper aims to understand LMs' strengths and limitations in memorizing fa...
http://arxiv.org/pdf/2212.10511v4
2023-07-02T07:21:59
2022-12-20T18:30:15
{ "title": [ -0.017318726, -0.021972656, -0.03048706, -0.035705566, -0.024108887, -0.014038086, -0.021347046, -0.017791748, -0.023986816, 0.02520752, -0.01889038, -0.00038862228, 0.006931305, -0.021102905, -0.032684326, -0.0068473816, -0.010726929, ...
http://arxiv.org/abs/2212.10509v2
Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions
Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, Ashish Sabharwal
cs.CL
Prompting-based large language models (LLMs) are surprisingly powerful at generating natural language reasoning steps or Chains-of-Thoughts (CoT) for multi-step question answering (QA). They struggle, however, when the necessary knowledge is either unavailable to the LLM or not up-to-date within its parameters. While u...
http://arxiv.org/pdf/2212.10509v2
2023-06-23T00:59:13
2022-12-20T18:26:34
{ "title": [ 0.013633728, 0.0017414093, -0.030212402, -0.064941406, 0.0021343231, -0.007987976, -0.01398468, 0.005176544, 0.02949524, 0.009239197, -0.035858154, 0.015106201, -0.025436401, -0.014709473, -0.038146973, -0.0034599304, 0.008644104, 0....
http://arxiv.org/abs/2212.10505v2
DePlot: One-shot visual language reasoning by plot-to-table translation
Fangyu Liu, Julian Martin Eisenschlos, Francesco Piccinno, Syrine Krichene, Chenxi Pang, Kenton Lee, Mandar Joshi, Wenhu Chen, Nigel Collier, Yasemin Altun
cs.CL, cs.AI, cs.CV
Visual language such as charts and plots is ubiquitous in the human world. Comprehending plots and charts requires strong reasoning skills. Prior state-of-the-art (SOTA) models require at least tens of thousands of training examples and their reasoning capabilities are still much limited, especially on complex human-wr...
http://arxiv.org/pdf/2212.10505v2
2023-05-23T18:28:39
2022-12-20T18:20:50
{ "title": [ -0.012260437, -0.02420044, -0.019592285, -0.06335449, -0.012825012, 0.017547607, -0.05340576, -0.012664795, 0.027175903, 0.011299133, -0.011894226, -0.0028362274, 0.0052108765, -0.023452759, -0.023132324, 0.0041885376, 0.0050811768, ...
http://arxiv.org/abs/2212.10504v2
Can Current Task-oriented Dialogue Models Automate Real-world Scenarios in the Wild?
Sang-Woo Lee, Sungdong Kim, Donghyeon Ko, Donghoon Ham, Youngki Hong, Shin Ah Oh, Hyunhoon Jung, Wangkyo Jung, Kyunghyun Cho, Donghyun Kwak, Hyungsuk Noh, Woomyoung Park
cs.CL
Task-oriented dialogue (TOD) systems are mainly based on the slot-filling-based TOD (SF-TOD) framework, in which dialogues are broken down into smaller, controllable units (i.e., slots) to fulfill a specific task. A series of approaches based on this framework achieved remarkable success on various TOD benchmarks. Howe...
http://arxiv.org/pdf/2212.10504v2
2023-05-24T11:46:33
2022-12-20T18:18:41
{ "title": [ 0.008522034, -0.019622803, -0.033050537, -0.016937256, -0.03366089, -0.02407837, -0.025878906, 0.0084991455, -0.033843994, 0.0259552, -0.0024299622, -0.021499634, 0.008903503, -0.033050537, -0.0073432922, -0.040740967, -0.009216309, ...
http://arxiv.org/abs/2212.10503v2
Mini-Model Adaptation: Efficiently Extending Pretrained Models to New Languages via Aligned Shallow Training
Kelly Marchisio, Patrick Lewis, Yihong Chen, Mikel Artetxe
cs.CL, cs.LG
Prior work shows that it is possible to expand pretrained Masked Language Models (MLMs) to new languages by learning a new set of embeddings, while keeping the transformer body frozen. Despite learning a small subset of parameters, this approach is not compute-efficient, as training the new embeddings requires a full f...
http://arxiv.org/pdf/2212.10503v2
2023-07-04T19:06:55
2022-12-20T18:17:28
{ "title": [ -0.017227173, -0.027893066, -0.022064209, -0.06713867, -0.011436462, 0.0014410019, -0.068847656, -0.0063438416, 0.0067253113, 0.024124146, -0.002729416, -0.0023517609, -0.0031337738, -0.03062439, -0.028274536, 0.010421753, 0.011810303, ...
http://arxiv.org/abs/2212.10502v2
A Measure-Theoretic Characterization of Tight Language Models
Li Du, Lucas Torroba Hennigen, Tiago Pimentel, Clara Meister, Jason Eisner, Ryan Cotterell
cs.CL, cs.FL
Language modeling, a central task in natural language processing, involves estimating a probability distribution over strings. In most cases, the estimated distribution sums to 1 over all finite strings. However, in some pathological cases, probability mass can ``leak'' onto the set of infinite sequences. In order to c...
http://arxiv.org/pdf/2212.10502v2
2023-08-21T18:01:57
2022-12-20T18:17:11
{ "title": [ -0.0670166, -0.007007599, -0.04135132, -0.049957275, -0.0006761551, 0.023742676, -0.07421875, 0.01826477, 0.004562378, 0.03515625, -0.03286743, 0.009239197, -0.022201538, 0.0011444092, -0.064331055, 0.004058838, -0.0055122375, 0.0173...
http://arxiv.org/abs/2212.10498v2
SimpleStyle: An Adaptable Style Transfer Approach
Elron Bandel, Yoav Katz, Noam Slonim, Liat Ein-Dor
cs.CL
Attribute-controlled text rewriting, also known as text style-transfer, has a crucial role in regulating attributes and biases of textual training data and a machine generated text. In this work we present SimpleStyle, a minimalist yet effective approach for style-transfer composed of two simple ingredients: controlled...
http://arxiv.org/pdf/2212.10498v2
2022-12-22T16:26:36
2022-12-20T18:12:49
{ "title": [ -0.022659302, -0.015945435, -0.046875, -0.051940918, -0.0015802383, 0.0031585693, -0.0519104, -0.029815674, 0.0016555786, 0.042022705, 0.015060425, -0.0016155243, 0.0067825317, -0.04284668, -0.023391724, 0.022857666, 0.031021118, -0....
http://arxiv.org/abs/2212.10496v1
Precise Zero-Shot Dense Retrieval without Relevance Labels
Luyu Gao, Xueguang Ma, Jimmy Lin, Jamie Callan
cs.IR, cs.CL
While dense retrieval has been shown effective and efficient across tasks and languages, it remains difficult to create effective fully zero-shot dense retrieval systems when no relevance label is available. In this paper, we recognize the difficulty of zero-shot learning and encoding relevance. Instead, we propose to ...
http://arxiv.org/pdf/2212.10496v1
2022-12-20T18:09:52
2022-12-20T18:09:52
{ "title": [ 0.008674622, -0.00059080124, -0.035064697, -0.049102783, 0.0129852295, -0.019088745, 0.0051193237, -0.0026378632, 0.045928955, 0.044647217, -0.02255249, 0.009750366, -0.00044846535, 0.001572609, -0.06866455, 0.024276733, 0.018035889, ...
http://arxiv.org/abs/2212.10481v2
Execution-Based Evaluation for Open-Domain Code Generation
Zhiruo Wang, Shuyan Zhou, Daniel Fried, Graham Neubig
cs.SE, cs.AI, cs.CL
To extend the scope of coding queries to more realistic settings, we propose ODEX, the first Open-Domain EXecution-based natural language (NL) to Python code generation dataset. ODEX has 945 NL-Code pairs spanning 79 diverse libraries, along with 1,707 human-written test cases for execution. Our NL-Code pairs are harve...
http://arxiv.org/pdf/2212.10481v2
2023-05-19T14:27:46
2022-12-20T17:54:37
{ "title": [ -0.034118652, 0.0005340576, -0.048980713, -0.06652832, 0.022399902, 0.0019950867, -0.046691895, 0.0029525757, -0.030181885, 0.022201538, 0.0015649796, 0.02142334, -0.02545166, -0.025375366, -0.01789856, -0.007156372, -0.015556335, 0....
http://arxiv.org/abs/2212.11119v1
A survey on text generation using generative adversarial networks
Gustavo Henrique de Rosa, João Paulo Papa
cs.CL, cs.AI, cs.LG
This work presents a thorough review concerning recent studies and text generation advancements using Generative Adversarial Networks. The usage of adversarial learning for text generation is promising as it provides alternatives to generate the so-called "natural" language. Nevertheless, adversarial text generation is...
http://arxiv.org/pdf/2212.11119v1
2022-12-20T17:54:08
2022-12-20T17:54:08
{ "title": [ -0.03414917, -0.013931274, -0.02885437, -0.03164673, -0.010429382, 0.0112838745, -0.03967285, -0.017608643, 0.0031261444, 0.047607422, 0.0026798248, 0.025131226, -0.0030384064, -0.010719299, -0.055633545, 0.011688232, -0.01802063, 0....
http://arxiv.org/abs/2212.10474v2
ByGPT5: End-to-End Style-conditioned Poetry Generation with Token-free Language Models
Jonas Belouadi, Steffen Eger
cs.CL
State-of-the-art poetry generation systems are often complex. They either consist of task-specific model pipelines, incorporate prior knowledge in the form of manually created constraints, or both. In contrast, end-to-end models would not suffer from the overhead of having to model prior knowledge and could learn the n...
http://arxiv.org/pdf/2212.10474v2
2023-05-22T21:15:06
2022-12-20T17:49:49
{ "title": [ -0.028793335, -0.009719849, -0.051208496, -0.07330322, 0.010505676, 0.0058135986, -0.059631348, 0.03262329, 0.002632141, 0.03451538, 0.0206604, 0.0053977966, 0.026687622, -0.033691406, -0.00299263, 0.025726318, -0.01586914, 0.0441894...
http://arxiv.org/abs/2212.11670v1
Evaluation for Change
Rishi Bommasani
cs.CL, cs.AI, cs.LG
Evaluation is the central means for assessing, understanding, and communicating about NLP models. In this position paper, we argue evaluation should be more than that: it is a force for driving change, carrying a sociological and political character beyond its technical dimensions. As a force, evaluation's power arises...
http://arxiv.org/pdf/2212.11670v1
2022-12-20T17:49:27
2022-12-20T17:49:27
{ "title": [ 0.013282776, 0.021255493, -0.03552246, -0.024124146, -0.023117065, -0.036254883, -0.03439331, -0.010696411, -0.014579773, 0.020721436, -0.033996582, -0.038116455, -0.026901245, -0.03869629, 0.007820129, -0.03842163, 0.0151901245, 0.0...
http://arxiv.org/abs/2212.10471v2
Little Red Riding Hood Goes Around the Globe:Crosslingual Story Planning and Generation with Large Language Models
Evgeniia Razumovskaia, Joshua Maynez, Annie Louis, Mirella Lapata, Shashi Narayan
cs.CL
Previous work has demonstrated the effectiveness of planning for story generation exclusively in a monolingual setting focusing primarily on English. We consider whether planning brings advantages to automatic story generation across languages. We propose a new task of cross-lingual story generation with planning and p...
http://arxiv.org/pdf/2212.10471v2
2023-05-22T10:24:12
2022-12-20T17:42:16
{ "title": [ -0.058685303, -0.024398804, -0.038269043, -0.0362854, -0.026107788, 0.0135650635, -0.05114746, 0.0000526309, -0.021499634, 0.03353882, -0.0053596497, 0.02078247, 0.027069092, -0.032562256, 0.014389038, 0.0037631989, -0.01600647, 0.04...
http://arxiv.org/abs/2212.10469v1
BMX: Boosting Machine Translation Metrics with Explainability
Christoph Leiter, Hoa Nguyen, Steffen Eger
cs.CL
State-of-the-art machine translation evaluation metrics are based on black-box language models. Hence, recent works consider their explainability with the goals of better understandability for humans and better metric analysis, including failure cases. In contrast, we explicitly leverage explanations to boost the metri...
http://arxiv.org/pdf/2212.10469v1
2022-12-20T17:41:18
2022-12-20T17:41:18
{ "title": [ -0.042144775, -0.01727295, -0.0023670197, -0.064208984, -0.010772705, 0.03152466, -0.052490234, 0.0006518364, -0.00086021423, 0.021270752, -0.00459671, 0.016159058, 0.015975952, -0.020065308, -0.030960083, 0.01424408, -0.01524353, 0....
http://arxiv.org/abs/2212.10467v2
Generic Temporal Reasoning with Differential Analysis and Explanation
Yu Feng, Ben Zhou, Haoyu Wang, Helen Jin, Dan Roth
cs.CL
Temporal reasoning is the task of predicting temporal relations of event pairs. While temporal reasoning models can perform reasonably well on in-domain benchmarks, we have little idea of these systems' generalizability due to existing datasets' limitations. In this work, we introduce a novel task named TODAY that brid...
http://arxiv.org/pdf/2212.10467v2
2023-05-31T17:54:08
2022-12-20T17:40:03
{ "title": [ -0.0052871704, 0.005332947, -0.047790527, -0.111816406, 0.04107666, 0.011207581, -0.028198242, 0.021942139, 0.02532959, 0.00023686886, -0.023223877, -0.01713562, 0.004081726, -0.02619934, -0.023757935, 0.004425049, 0.008293152, 0.010...
http://arxiv.org/abs/2212.10466v1
Controllable Text Generation with Language Constraints
Howard Chen, Huihan Li, Danqi Chen, Karthik Narasimhan
cs.CL
We consider the task of text generation in language models with constraints specified in natural language. To this end, we first create a challenging benchmark Cognac that provides as input to the model a topic with example text, along with a constraint on text to be avoided. Unlike prior work, our benchmark contains k...
http://arxiv.org/pdf/2212.10466v1
2022-12-20T17:39:21
2022-12-20T17:39:21
{ "title": [ -0.055114746, -0.015808105, -0.039916992, -0.049987793, -0.007896423, 0.026321411, -0.06964111, 0.003627777, 0.005680084, 0.03878784, -0.013397217, 0.010826111, 0.0016260147, -0.004055023, -0.038635254, 0.012527466, -0.013511658, 0.0...
http://arxiv.org/abs/2212.10465v3
SODA: Million-scale Dialogue Distillation with Social Commonsense Contextualization
Hyunwoo Kim, Jack Hessel, Liwei Jiang, Peter West, Ximing Lu, Youngjae Yu, Pei Zhou, Ronan Le Bras, Malihe Alikhani, Gunhee Kim, Maarten Sap, Yejin Choi
cs.CL
Data scarcity has been a long standing issue in the field of open-domain social dialogue. To quench this thirst, we present SODA: the first publicly available, million-scale high-quality social dialogue dataset. By contextualizing social commonsense knowledge from a knowledge graph, we are able to distill an exceptiona...
http://arxiv.org/pdf/2212.10465v3
2023-10-23T18:46:21
2022-12-20T17:38:47
{ "title": [ 0.021087646, 0.0010032654, -0.038848877, -0.09399414, 0.013435364, -0.0022296906, -0.063964844, -0.0010118484, 0.028030396, 0.021026611, -0.027420044, 0.01007843, -0.011917114, -0.012634277, -0.038208008, -0.0050582886, -0.018615723, ...
http://arxiv.org/abs/2212.10461v1
Go-tuning: Improving Zero-shot Learning Abilities of Smaller Language Models
Jingjing Xu, Qingxiu Dong, Hongyi Liu, Lei Li
cs.CL
With increasing scale, large language models demonstrate both quantitative improvement and new qualitative capabilities, especially as zero-shot learners, like GPT-3. However, these results rely heavily on delicate prompt design and large computation. In this work, we explore whether the strong zero-shot ability could ...
http://arxiv.org/pdf/2212.10461v1
2022-12-20T17:36:49
2022-12-20T17:36:49
{ "title": [ -0.037628174, -0.030776978, -0.026748657, -0.041046143, -0.033233643, 0.0003733635, -0.046966553, 0.01939392, -0.006450653, 0.041137695, -0.015991211, 0.023513794, -0.000088095665, -0.006713867, -0.045410156, 0.008964539, -0.0016212463, ...
http://arxiv.org/abs/2212.10455v2
MULTI3NLU++: A Multilingual, Multi-Intent, Multi-Domain Dataset for Natural Language Understanding in Task-Oriented Dialogue
Nikita Moghe, Evgeniia Razumovskaia, Liane Guillou, Ivan Vulić, Anna Korhonen, Alexandra Birch
cs.CL
Task-oriented dialogue (TOD) systems have been widely deployed in many industries as they deliver more efficient customer support. These systems are typically constructed for a single domain or language and do not generalise well beyond this. To support work on Natural Language Understanding (NLU) in TOD across multipl...
http://arxiv.org/pdf/2212.10455v2
2023-06-19T04:09:37
2022-12-20T17:34:25
{ "title": [ -0.017440796, -0.010383606, -0.062683105, -0.084228516, 0.014045715, -0.016540527, -0.061340332, 0.032806396, 0.00047516823, 0.017196655, 0.00541687, -0.0093688965, 0.00031638145, -0.027175903, -0.037017822, -0.00919342, -0.011924744, ...
http://arxiv.org/abs/2212.10450v2
Is GPT-3 a Good Data Annotator?
Bosheng Ding, Chengwei Qin, Linlin Liu, Yew Ken Chia, Shafiq Joty, Boyang Li, Lidong Bing
cs.CL
Data annotation is the process of labeling data that could be used to train machine learning models. Having high-quality annotation is crucial, as it allows the model to learn the relationship between the input data and the desired output. GPT-3, a large-scale language model developed by OpenAI, has demonstrated impres...
http://arxiv.org/pdf/2212.10450v2
2023-06-14T16:11:50
2022-12-20T17:28:41
{ "title": [ 0.023010254, 0.03665161, -0.025115967, -0.03050232, -0.008415222, -0.018417358, -0.042510986, -0.029663086, 0.0019464493, 0.029006958, -0.01939392, -0.02607727, -0.032440186, -0.026031494, -0.02116394, -0.018707275, 0.009185791, 0.01...
http://arxiv.org/abs/2212.10449v3
Socratic Pretraining: Question-Driven Pretraining for Controllable Summarization
Artidoro Pagnoni, Alexander R. Fabbri, Wojciech Kryściński, Chien-Sheng Wu
cs.CL
In long document controllable summarization, where labeled data is scarce, pretrained models struggle to adapt to the task and effectively respond to user queries. In this paper, we introduce Socratic pretraining, a question-driven, unsupervised pretraining objective specifically designed to improve controllability in ...
http://arxiv.org/pdf/2212.10449v3
2023-06-08T22:43:58
2022-12-20T17:27:10
{ "title": [ 0.015670776, 0.013648987, -0.018096924, -0.06439209, -0.014923096, -0.009635925, -0.043029785, -0.025238037, -0.019729614, 0.01953125, -0.0151901245, 0.046325684, -0.05429077, -0.0067634583, -0.032836914, -0.014709473, 0.0014591217, ...
http://arxiv.org/abs/2212.10448v1
Parameter-efficient Zero-shot Transfer for Cross-Language Dense Retrieval with Adapters
Eugene Yang, Suraj Nair, Dawn Lawrie, James Mayfield, Douglas W. Oard
cs.IR, cs.CL
A popular approach to creating a zero-shot cross-language retrieval model is to substitute a monolingual pretrained language model in the retrieval model with a multilingual pretrained language model such as Multilingual BERT. This multilingual model is fined-tuned to the retrieval task with monolingual data such as En...
http://arxiv.org/pdf/2212.10448v1
2022-12-20T17:25:04
2022-12-20T17:25:04
{ "title": [ 0.015960693, -0.008293152, -0.022613525, -0.066467285, 0.0073051453, -0.0021343231, -0.01701355, -0.0029506683, 0.050598145, 0.04928589, -0.012283325, 0.0028266907, 0.035827637, -0.009552002, -0.04788208, 0.034332275, 0.015625, 0.008...
http://arxiv.org/abs/2212.10440v1
Perplexed by Quality: A Perplexity-based Method for Adult and Harmful Content Detection in Multilingual Heterogeneous Web Data
Tim Jansen, Yangling Tong, Victoria Zevallos, Pedro Ortiz Suarez
cs.CL
As demand for large corpora increases with the size of current state-of-the-art language models, using web data as the main part of the pre-training corpus for these models has become a ubiquitous practice. This, in turn, has introduced an important challenge for NLP practitioners, as they are now confronted with the t...
http://arxiv.org/pdf/2212.10440v1
2022-12-20T17:14:45
2022-12-20T17:14:45
{ "title": [ 0.026275635, 0.015617371, -0.021316528, -0.032836914, 0.03213501, -0.032836914, -0.009933472, -0.018356323, 0.020202637, 0.017608643, -0.023513794, 0.024047852, 0.0077552795, -0.022644043, -0.0075149536, -0.006626129, -0.017150879, 0...
http://arxiv.org/abs/2212.10423v1
Fine-Grained Distillation for Long Document Retrieval
Yucheng Zhou, Tao Shen, Xiubo Geng, Chongyang Tao, Guodong Long, Can Xu, Daxin Jiang
cs.IR, cs.CL
Long document retrieval aims to fetch query-relevant documents from a large-scale collection, where knowledge distillation has become de facto to improve a retriever by mimicking a heterogeneous yet powerful cross-encoder. However, in contrast to passages or sentences, retrieval on long documents suffers from the scope...
http://arxiv.org/pdf/2212.10423v1
2022-12-20T17:00:36
2022-12-20T17:00:36
{ "title": [ 0.019424438, 0.007507324, -0.040100098, -0.077697754, 0.019729614, -0.009361267, -0.03100586, 0.0073509216, 0.040374756, 0.054840088, -0.04537964, -0.031982422, 0.017715454, -0.014549255, -0.06335449, -0.00907135, 0.0038318634, 0.013...
http://arxiv.org/abs/2212.10422v3
Localising In-Domain Adaptation of Transformer-Based Biomedical Language Models
Tommaso Mario Buonocore, Claudio Crema, Alberto Redolfi, Riccardo Bellazzi, Enea Parimbelli
cs.CL, cs.AI, cs.LG, I.2.7; J.3
In the era of digital healthcare, the huge volumes of textual information generated every day in hospitals constitute an essential but underused asset that could be exploited with task-specific, fine-tuned biomedical language representation models, improving patient care and management. For such specialized domains, pr...
http://arxiv.org/pdf/2212.10422v3
2023-06-28T08:36:20
2022-12-20T16:59:56
{ "title": [ -0.036193848, -0.0022678375, -0.036743164, -0.027908325, -0.0072631836, 0.000037789345, -0.057647705, -0.01335907, 0.008476257, 0.045654297, 0.015571594, -0.025878906, 0.043060303, -0.042663574, -0.044036865, 0.019058228, -0.008811951, ...
http://arxiv.org/abs/2212.10409v3
ClarifyDelphi: Reinforced Clarification Questions with Defeasibility Rewards for Social and Moral Situations
Valentina Pyatkin, Jena D. Hwang, Vivek Srikumar, Ximing Lu, Liwei Jiang, Yejin Choi, Chandra Bhagavatula
cs.CL
Context is everything, even in commonsense moral reasoning. Changing contexts can flip the moral judgment of an action; "Lying to a friend" is wrong in general, but may be morally acceptable if it is intended to protect their life. We present ClarifyDelphi, an interactive system that learns to ask clarification quest...
http://arxiv.org/pdf/2212.10409v3
2023-05-30T18:59:41
2022-12-20T16:33:09
{ "title": [ -0.0015106201, 0.0053482056, -0.036193848, -0.02041626, -0.016326904, -0.019317627, -0.008270264, -0.023147583, -0.02229309, -0.006942749, -0.014976501, -0.019073486, -0.004787445, -0.04815674, 0.00051641464, -0.037017822, -0.038513184, ...
http://arxiv.org/abs/2212.10408v1
Geographic and Geopolitical Biases of Language Models
Fahim Faisal, Antonios Anastasopoulos
cs.CL
Pretrained language models (PLMs) often fail to fairly represent target users from certain world regions because of the under-representation of those regions in training datasets. With recent PLMs trained on enormous data sources, quantifying their potential biases is difficult, due to their black-box nature and the sh...
http://arxiv.org/pdf/2212.10408v1
2022-12-20T16:32:54
2022-12-20T16:32:54
{ "title": [ -0.015342712, 0.0051078796, -0.022476196, -0.032073975, -0.021713257, -0.020874023, -0.030807495, -0.016921997, -0.019760132, 0.03765869, -0.022506714, -0.011062622, 0.017105103, -0.034942627, -0.019119263, -0.0022697449, -0.014251709, ...