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/2305.14658v2
Evaluate What You Can't Evaluate: Unassessable Quality for Generated Response
Yongkang Liu, Shi Feng, Daling Wang, Yifei Zhang, Hinrich Schütze
cs.CL
LLMs (large language models) such as ChatGPT have shown remarkable language understanding and generation capabilities. Although reference-free evaluators based on LLMs show better human alignment than traditional reference-based evaluators, there are many challenges in using reference-free evaluators based on LLMs. Ref...
http://arxiv.org/pdf/2305.14658v2
2023-11-14T19:16:35
2023-05-24T02:52:48
{ "title": [ 0.026153564, 0.015129089, -0.017959595, -0.03768921, -0.00793457, -0.01864624, -0.015930176, -0.036499023, 0.0045280457, 0.013580322, -0.01411438, -0.0055007935, -0.048187256, -0.033843994, -0.064331055, -0.022598267, 0.0016727448, 0...
http://arxiv.org/abs/2305.14652v3
Denoising Bottleneck with Mutual Information Maximization for Video Multimodal Fusion
Shaoxiang Wu, Damai Dai, Ziwei Qin, Tianyu Liu, Binghuai Lin, Yunbo Cao, Zhifang Sui
cs.CL
Video multimodal fusion aims to integrate multimodal signals in videos, such as visual, audio and text, to make a complementary prediction with multiple modalities contents. However, unlike other image-text multimodal tasks, video has longer multimodal sequences with more redundancy and noise in both visual and audio m...
http://arxiv.org/pdf/2305.14652v3
2023-05-31T08:20:33
2023-05-24T02:39:43
{ "title": [ -0.032592773, 0.052368164, -0.06036377, -0.02381897, 0.024612427, -0.011726379, -0.014915466, -0.039215088, -0.0041046143, 0.049072266, -0.017623901, 0.018707275, 0.0065727234, -0.019348145, -0.038757324, -0.003648758, -0.0129852295, ...
http://arxiv.org/abs/2305.14651v1
Revisit and Outstrip Entity Alignment: A Perspective of Generative Models
Lingbing Guo, Zhuo Chen, Jiaoyan Chen, Huajun Chen
cs.CL, cs.IR, cs.LG
Recent embedding-based methods have achieved great successes on exploiting entity alignment from knowledge graph (KG) embeddings of multiple modals. In this paper, we study embedding-based entity alignment (EEA) from a perspective of generative models. We show that EEA is a special problem where the main objective is a...
http://arxiv.org/pdf/2305.14651v1
2023-05-24T02:39:20
2023-05-24T02:39:20
{ "title": [ -0.013206482, -0.008399963, -0.05053711, -0.07244873, 0.012336731, -0.006706238, -0.057678223, -0.005126953, 0.02633667, 0.043029785, -0.0030269623, 0.0037288666, 0.01751709, -0.018249512, -0.053497314, 0.023101807, -0.02760315, 0.03...
http://arxiv.org/abs/2305.14647v2
Scientific Opinion Summarization: Meta-review Generation with Checklist-guided Iterative Introspection
Qi Zeng, Mankeerat Sidhu, Hou Pong Chan, Lu Wang, Heng Ji
cs.CL
Opinions in the scientific domain can be divergent, leading to controversy or consensus among reviewers. However, current opinion summarization datasets mostly focus on product review domains, which do not account for this variability under the assumption that the input opinions are non-controversial. To address this g...
http://arxiv.org/pdf/2305.14647v2
2023-11-13T19:47:35
2023-05-24T02:33:35
{ "title": [ -0.014442444, -0.00223732, -0.027572632, -0.040771484, -0.011123657, -0.00054073334, -0.060028076, -0.03829956, -0.00712204, 0.04940796, 0.0009255409, 0.030700684, -0.020233154, -0.00868988, -0.06384277, 0.015686035, -0.015167236, 0....
http://arxiv.org/abs/2305.14645v1
Iteratively Improving Biomedical Entity Linking and Event Extraction via Hard Expectation-Maximization
Xiaochu Li, Minqian Liu, Zhiyang Xu, Lifu Huang
cs.CL
Biomedical entity linking and event extraction are two crucial tasks to support text understanding and retrieval in the biomedical domain. These two tasks intrinsically benefit each other: entity linking disambiguates the biomedical concepts by referring to external knowledge bases and the domain knowledge further prov...
http://arxiv.org/pdf/2305.14645v1
2023-05-24T02:30:31
2023-05-24T02:30:31
{ "title": [ -0.0029468536, 0.012046814, -0.037597656, -0.08843994, 0.0284729, 0.00844574, -0.053894043, -0.009300232, 0.00995636, 0.018676758, 0.02508545, 0.012916565, -0.0029277802, -0.03375244, -0.018005371, 0.022384644, -0.011428833, 0.028305...
http://arxiv.org/abs/2305.14635v2
CMOT: Cross-modal Mixup via Optimal Transport for Speech Translation
Yan Zhou, Qingkai Fang, Yang Feng
cs.CL, cs.AI, cs.SD, eess.AS
End-to-end speech translation (ST) is the task of translating speech signals in the source language into text in the target language. As a cross-modal task, end-to-end ST is difficult to train with limited data. Existing methods often try to transfer knowledge from machine translation (MT), but their performances are r...
http://arxiv.org/pdf/2305.14635v2
2023-05-25T08:55:41
2023-05-24T02:13:48
{ "title": [ -0.011520386, 0.00018310547, -0.013900757, -0.046722412, 0.010124207, 0.007144928, -0.037200928, -0.02142334, -0.0016231537, 0.053466797, -0.003829956, 0.029205322, 0.025817871, -0.01927185, -0.03567505, 0.018173218, -0.013648987, 0....
http://arxiv.org/abs/2305.16339v2
Don't Trust ChatGPT when Your Question is not in English: A Study of Multilingual Abilities and Types of LLMs
Xiang Zhang, Senyu Li, Bradley Hauer, Ning Shi, Grzegorz Kondrak
cs.CL, cs.AI
Large Language Models (LLMs) have demonstrated exceptional natural language understanding abilities and have excelled in a variety of natural language processing (NLP)tasks in recent years. Despite the fact that most LLMs are trained predominantly in English, multiple studies have demonstrated their comparative perform...
http://arxiv.org/pdf/2305.16339v2
2023-10-24T04:38:52
2023-05-24T02:05:03
{ "title": [ -0.059417725, -0.029251099, -0.04837036, 0.0014667511, -0.021209717, -0.012275696, -0.046539307, -0.010757446, -0.02670288, 0.030776978, -0.0053138733, 0.02305603, -0.012214661, -0.032226562, -0.007774353, 0.0025253296, -0.021850586, ...
http://arxiv.org/abs/2305.14630v1
Testing Causal Models of Word Meaning in GPT-3 and -4
Sam Musker, Ellie Pavlick
cs.CL, I.2.7
Large Language Models (LLMs) have driven extraordinary improvements in NLP. However, it is unclear how such models represent lexical concepts-i.e., the meanings of the words they use. This paper evaluates the lexical representations of GPT-3 and GPT-4 through the lens of HIPE theory, a theory of concept representations...
http://arxiv.org/pdf/2305.14630v1
2023-05-24T02:03:23
2023-05-24T02:03:23
{ "title": [ -0.00749588, -0.0064468384, -0.030838013, -0.03515625, -0.017333984, -0.012924194, -0.04208374, -0.01878357, -0.024963379, 0.008155823, -0.011711121, 0.016052246, -0.02015686, -0.03778076, -0.012954712, -0.0034065247, -0.028457642, 0...
http://arxiv.org/abs/2305.14628v2
Getting MoRE out of Mixture of Language Model Reasoning Experts
Chenglei Si, Weijia Shi, Chen Zhao, Luke Zettlemoyer, Jordan Boyd-Graber
cs.CL, cs.AI
While recent large language models (LLMs) improve on various question answering (QA) datasets, it remains difficult for a single model to generalize across question types that require distinct reasoning abilities. We provide empirical evidence that state-of-the-art LLMs suffer from poor generalizability on reasoning ty...
http://arxiv.org/pdf/2305.14628v2
2023-10-20T05:16:29
2023-05-24T02:00:51
{ "title": [ 0.008392334, -0.008834839, -0.043273926, -0.095458984, -0.026367188, -0.0095825195, -0.060424805, 0.030914307, -0.036865234, 0.037963867, -0.014732361, 0.0038490295, 0.024459839, -0.044281006, 0.029708862, -0.0020771027, 0.014831543, ...
http://arxiv.org/abs/2305.14627v2
Enabling Large Language Models to Generate Text with Citations
Tianyu Gao, Howard Yen, Jiatong Yu, Danqi Chen
cs.CL, cs.IR, cs.LG
Large language models (LLMs) have emerged as a widely-used tool for information seeking, but their generated outputs are prone to hallucination. In this work, our aim is to allow LLMs to generate text with citations, improving their factual correctness and verifiability. Existing work mainly relies on commercial search...
http://arxiv.org/pdf/2305.14627v2
2023-10-31T15:04:35
2023-05-24T01:53:49
{ "title": [ -0.022247314, -0.0056419373, -0.041809082, -0.06109619, -0.001947403, 0.0014915466, -0.051452637, -0.006580353, 0.0022087097, 0.039398193, -0.014122009, -0.00034093857, 0.026123047, -0.010063171, -0.044921875, 0.014282227, -0.017868042, ...
http://arxiv.org/abs/2305.14625v1
KNN-LM Does Not Improve Open-ended Text Generation
Shufan Wang, Yixiao Song, Andrew Drozdov, Aparna Garimella, Varun Manjunatha, Mohit Iyyer
cs.CL
In this paper, we study the generation quality of interpolation-based retrieval-augmented language models (LMs). These methods, best exemplified by the KNN-LM, interpolate the LM's predicted distribution of the next word with a distribution formed from the most relevant retrievals for a given prefix. While the KNN-LM a...
http://arxiv.org/pdf/2305.14625v1
2023-05-24T01:48:33
2023-05-24T01:48:33
{ "title": [ -0.06555176, -0.006134033, 0.013694763, 0.032684326, -0.081726074, 0.04348755, -0.048980713, -0.022018433, -0.026733398, 0.07507324, -0.011199951, -0.031036377, 0.02809143, -0.02178955, -0.010940552, -0.0022964478, -0.0069007874, 0.0...
http://arxiv.org/abs/2305.14623v1
Self-Checker: Plug-and-Play Modules for Fact-Checking with Large Language Models
Miaoran Li, Baolin Peng, Zhu Zhang
cs.CL
Fact-checking is an essential task in NLP that is commonly utilized for validating the factual accuracy of claims. Prior work has mainly focused on fine-tuning pre-trained languages models on specific datasets, which can be computationally intensive and time-consuming. With the rapid development of large language model...
http://arxiv.org/pdf/2305.14623v1
2023-05-24T01:46:07
2023-05-24T01:46:07
{ "title": [ -0.008163452, -0.0074539185, -0.04425049, -0.10046387, 0.014984131, 0.010398865, -0.06567383, -0.0029907227, -0.025527954, 0.020828247, -0.0023117065, 0.009552002, 0.0047912598, -0.027267456, -0.029312134, -0.0005273819, -0.030807495, ...
http://arxiv.org/abs/2305.14622v1
EXnet: Efficient In-context Learning for Data-less Text classification
Debaditya Shome, Kuldeep Yadav
cs.CL, cs.LG
Large pre-trained language models (PLMs) have made significant progress in encoding world knowledge and spawned a new set of learning paradigms including zero-shot, few-shot, and in-context learning. Many language tasks can be modeled as a set of prompts (for example, is this text about geography?) and language models ...
http://arxiv.org/pdf/2305.14622v1
2023-05-24T01:40:57
2023-05-24T01:40:57
{ "title": [ 0.0026664734, -0.016784668, -0.04626465, -0.05316162, 0.0211792, -0.017196655, 0.0034732819, 0.0016698837, 0.024627686, 0.021087646, -0.03881836, 0.01966858, -0.007911682, -0.0037593842, -0.03857422, -0.0061073303, -0.010856628, 0.05...
http://arxiv.org/abs/2305.14618v1
Abductive Commonsense Reasoning Exploiting Mutually Exclusive Explanations
Wenting Zhao, Justin T. Chiu, Claire Cardie, Alexander M. Rush
cs.CL, cs.AI
Abductive reasoning aims to find plausible explanations for an event. This style of reasoning is critical for commonsense tasks where there are often multiple plausible explanations. Existing approaches for abductive reasoning in natural language processing (NLP) often rely on manually generated annotations for supervi...
http://arxiv.org/pdf/2305.14618v1
2023-05-24T01:35:10
2023-05-24T01:35:10
{ "title": [ 0.0069618225, -0.019439697, -0.014129639, -0.03894043, -0.005710602, -0.0118255615, 0.006832123, -0.02482605, -0.018234253, -0.0009784698, -0.020706177, 0.010353088, -0.045288086, -0.031402588, 0.0013933182, -0.014839172, -0.009155273, ...
http://arxiv.org/abs/2305.14617v2
COMET-M: Reasoning about Multiple Events in Complex Sentences
Sahithya Ravi, Raymond Ng, Vered Shwartz
cs.CL, cs.AI
Understanding the speaker's intended meaning often involves drawing commonsense inferences to reason about what is not stated explicitly. In multi-event sentences, it requires understanding the relationships between events based on contextual knowledge. We propose COMET-M (Multi-Event), an event-centric commonsense mod...
http://arxiv.org/pdf/2305.14617v2
2023-10-23T06:49:57
2023-05-24T01:35:01
{ "title": [ 0.011100769, -0.030532837, -0.017166138, -0.121276855, -0.0015821457, 0.018692017, -0.055755615, 0.016189575, 0.017852783, 0.008171082, -0.0024223328, 0.010803223, 0.0053977966, -0.030151367, -0.02470398, 0.011138916, -0.022583008, 0...
http://arxiv.org/abs/2305.14616v2
Exploring Affordance and Situated Meaning in Image Captions: A Multimodal Analysis
Pin-Er Chen, Po-Ya Angela Wang, Hsin-Yu Chou, Yu-Hsiang Tseng, Shu-Kai Hsieh
cs.CL, cs.CV
This paper explores the grounding issue regarding multimodal semantic representation from a computational cognitive-linguistic view. We annotate images from the Flickr30k dataset with five perceptual properties: Affordance, Perceptual Salience, Object Number, Gaze Cueing, and Ecological Niche Association (ENA), and exa...
http://arxiv.org/pdf/2305.14616v2
2023-10-24T11:30:07
2023-05-24T01:30:50
{ "title": [ 0.005935669, 0.0061531067, -0.0076179504, -0.024017334, 0.0030670166, -0.026382446, 0.0013980865, -0.030349731, 0.021438599, 0.044128418, -0.010871887, 0.021469116, -0.006958008, -0.008369446, -0.04888916, 0.0053253174, -0.022720337, ...
http://arxiv.org/abs/2305.14613v2
Selectively Answering Ambiguous Questions
Jeremy R. Cole, Michael J. Q. Zhang, Daniel Gillick, Julian Martin Eisenschlos, Bhuwan Dhingra, Jacob Eisenstein
cs.CL, cs.AI
Trustworthy language models should abstain from answering questions when they do not know the answer. However, the answer to a question can be unknown for a variety of reasons. Prior research has focused on the case in which the question is clear and the answer is unambiguous but possibly unknown, but the answer to a q...
http://arxiv.org/pdf/2305.14613v2
2023-11-15T02:15:02
2023-05-24T01:25:38
{ "title": [ 0.023071289, 0.010025024, -0.010787964, -0.02708435, -0.033966064, -0.005256653, -0.006832123, -0.032684326, -0.008964539, -0.004009247, -0.03451538, -0.035339355, -0.04220581, -0.04397583, -0.006389618, -0.045532227, 0.023956299, 0....
http://arxiv.org/abs/2305.16338v1
Think Before You Act: Decision Transformers with Internal Working Memory
Jikun Kang, Romain Laroche, Xindi Yuan, Adam Trischler, Xue Liu, Jie Fu
cs.LG, cs.AI, cs.CL
Large language model (LLM)-based decision-making agents have shown the ability to generalize across multiple tasks. However, their performance relies on massive data and compute. We argue that this inefficiency stems from the forgetting phenomenon, in which a model memorizes its behaviors in parameters throughout train...
http://arxiv.org/pdf/2305.16338v1
2023-05-24T01:20:22
2023-05-24T01:20:22
{ "title": [ -0.00029277802, -0.0044403076, -0.014274597, -0.028137207, 0.008178711, -0.025482178, 0.016082764, -0.027359009, -0.017318726, 0.011405945, -0.03729248, 0.011680603, -0.01889038, -0.015380859, -0.02319336, -0.0110321045, -0.0032444, ...
http://arxiv.org/abs/2305.14610v2
This Land is {Your, My} Land: Evaluating Geopolitical Biases in Language Models
Bryan Li, Chris Callison-Burch
cs.CL
Do the Spratly Islands belong to China, the Philippines, or Vietnam? A pretrained large language model (LLM) may answer differently if asked in the languages of each claimant country: Chinese, Tagalog, or Vietnamese. This contrasts with a multilingual human, who would likely answer consistently. In this work, we show t...
http://arxiv.org/pdf/2305.14610v2
2023-10-18T22:02:43
2023-05-24T01:16:17
{ "title": [ -0.010070801, 0.0012407303, -0.015205383, -0.027832031, 0.000053048134, -0.020217896, -0.027374268, -0.02810669, -0.027389526, 0.021255493, -0.025466919, 0.016525269, -0.01348114, -0.01828003, -0.027938843, -0.022232056, -0.03060913, ...
http://arxiv.org/abs/2305.14603v1
OpenPI2.0: An Improved Dataset for Entity Tracking in Texts
Li Zhang, Hainiu Xu, Abhinav Kommula, Niket Tandon, Chris Callison-Burch
cs.CL
Representing texts as information about entities has long been deemed effective in event reasoning. We propose OpenPI2.0, an improved dataset for tracking entity states in procedural texts. OpenPI2.0 features not only canonicalized entities that facilitate evaluation, but also salience annotations including both manual...
http://arxiv.org/pdf/2305.14603v1
2023-05-24T00:57:35
2023-05-24T00:57:35
{ "title": [ -0.019042969, -0.005355835, -0.037017822, -0.064697266, 0.018661499, 0.00041556358, -0.047973633, -0.024291992, 0.0015926361, 0.036071777, -0.0049819946, 0.027801514, 0.038146973, -0.013847351, -0.02722168, 0.024230957, -0.0140686035, ...
http://arxiv.org/abs/2305.14600v3
Learning Semantic Role Labeling from Compatible Label Sequences
Tao Li, Ghazaleh Kazeminejad, Susan W. Brown, Martha Palmer, Vivek Srikumar
cs.CL, cs.LG
Semantic role labeling (SRL) has multiple disjoint label sets, e.g., VerbNet and PropBank. Creating these datasets is challenging, therefore a natural question is how to use each one to help the other. Prior work has shown that cross-task interaction helps, but only explored multitask learning so far. A common issue wi...
http://arxiv.org/pdf/2305.14600v3
2023-10-19T22:34:41
2023-05-24T00:46:02
{ "title": [ -0.02204895, -0.008728027, -0.052581787, -0.08123779, 0.00055503845, 0.013786316, -0.061920166, 0.017684937, -0.010566711, 0.02609253, -0.014419556, 0.010269165, -0.013343811, -0.013908386, -0.031677246, 0.0011177063, -0.017242432, 0...
http://arxiv.org/abs/2305.14599v2
Bridging Continuous and Discrete Spaces: Interpretable Sentence Representation Learning via Compositional Operations
James Y. Huang, Wenlin Yao, Kaiqiang Song, Hongming Zhang, Muhao Chen, Dong Yu
cs.CL
Traditional sentence embedding models encode sentences into vector representations to capture useful properties such as the semantic similarity between sentences. However, in addition to similarity, sentence semantics can also be interpreted via compositional operations such as sentence fusion or difference. It is uncl...
http://arxiv.org/pdf/2305.14599v2
2023-11-05T06:03:59
2023-05-24T00:44:49
{ "title": [ -0.004020691, 0.01197052, -0.025314331, -0.07110596, 0.009223938, -0.0023880005, -0.068481445, -0.007369995, 0.0036392212, 0.023498535, -0.027145386, 0.0058555603, -0.010299683, -0.0014829636, -0.06085205, -0.022277832, -0.020263672, ...
http://arxiv.org/abs/2305.14597v1
Voices of Her: Analyzing Gender Differences in the AI Publication World
Yiwen Ding, Jiarui Liu, Zhiheng Lyu, Kun Zhang, Bernhard Schoelkopf, Zhijing Jin, Rada Mihalcea
cs.CL, cs.AI, cs.LG
While several previous studies have analyzed gender bias in research, we are still missing a comprehensive analysis of gender differences in the AI community, covering diverse topics and different development trends. Using the AI Scholar dataset of 78K researchers in the field of AI, we identify several gender differen...
http://arxiv.org/pdf/2305.14597v1
2023-05-24T00:40:49
2023-05-24T00:40:49
{ "title": [ -0.06964111, -0.021453857, -0.07086182, 0.00010961294, -0.021270752, -0.033996582, -0.0071372986, 0.016799927, -0.0024375916, 0.023986816, -0.039093018, 0.013893127, 0.009468079, -0.011810303, -0.030944824, -0.022872925, -0.012390137, ...
http://arxiv.org/abs/2305.14596v2
Increasing Probability Mass on Answer Choices Does Not Always Improve Accuracy
Sarah Wiegreffe, Matthew Finlayson, Oyvind Tafjord, Peter Clark, Ashish Sabharwal
cs.CL, cs.LG
When pretrained language models (LMs) are applied to discriminative tasks such as multiple-choice questions, they place probability mass on vocabulary tokens that aren't among the given answer choices. Spreading probability mass across multiple surface forms with identical meaning (such as "bath" and "bathtub") is thou...
http://arxiv.org/pdf/2305.14596v2
2023-10-31T22:07:10
2023-05-24T00:27:00
{ "title": [ -0.0072669983, 0.019454956, -0.011695862, 0.038970947, -0.016220093, -0.022644043, 0.05444336, -0.03945923, 0.015258789, 0.032836914, -0.0574646, -0.02633667, -0.017669678, -0.024841309, -0.03463745, -0.020523071, 0.01928711, -0.0006...
http://arxiv.org/abs/2305.14592v1
Instruction Tuning with Lexicons for Zero-Shot Style Classification
Ruohao Guo, Wei Xu, Alan Ritter
cs.CL, cs.LG
Style is used to convey authors' intentions and attitudes. Despite the success of large pre-trained language models on style classification, prior work relies on fine-tuning with labeled examples. Prompting large language models to classify style without fine-tuning is challenging because language styles can be difficu...
http://arxiv.org/pdf/2305.14592v1
2023-05-24T00:17:36
2023-05-24T00:17:36
{ "title": [ -0.058807373, -0.032165527, -0.032836914, -0.019760132, -0.045166016, -0.010726929, -0.05441284, 0.018035889, 0.014656067, 0.041809082, -0.023345947, 0.0069618225, -0.008613586, -0.010658264, -0.028366089, -0.00042676926, 0.004173279, ...
http://arxiv.org/abs/2305.14591v2
ALGO: Synthesizing Algorithmic Programs with Generated Oracle Verifiers
Kexun Zhang, Danqing Wang, Jingtao Xia, William Yang Wang, Lei Li
cs.CL, cs.SE
Large language models (LLMs) excel at implementing code from functionality descriptions but struggle with algorithmic problems that require not only implementation but also identification of the suitable algorithm. Moreover, LLM-generated programs lack guaranteed correctness and require human verification. To address t...
http://arxiv.org/pdf/2305.14591v2
2023-10-29T21:37:07
2023-05-24T00:10:15
{ "title": [ -0.033843994, -0.009468079, -0.050445557, -0.08215332, -0.010017395, 0.011177063, -0.06225586, 0.017807007, -0.006576538, 0.019943237, -0.036315918, 0.012046814, -0.026168823, -0.014022827, -0.04296875, -0.017227173, -0.0013399124, -...
http://arxiv.org/abs/2305.14590v1
RE$^2$: Region-Aware Relation Extraction from Visually Rich Documents
Pritika Ramu, Sijia Wang, Lalla Mouatadid, Joy Rimchala, Lifu Huang
cs.CL, cs.AI
Current research in form understanding predominantly relies on large pre-trained language models, necessitating extensive data for pre-training. However, the importance of layout structure (i.e., the spatial relationship between the entity blocks in the visually rich document) to relation extraction has been overlooked...
http://arxiv.org/pdf/2305.14590v1
2023-05-24T00:07:40
2023-05-24T00:07:40
{ "title": [ -0.017227173, -0.015335083, -0.042266846, -0.06317139, -0.001241684, -0.0027046204, -0.037597656, -0.003129959, 0.037200928, 0.0149383545, -0.018447876, 0.0027561188, -0.009002686, -0.007007599, -0.06829834, -0.0071792603, -0.0059928894,...
http://arxiv.org/abs/2305.14588v1
Evaluating end-to-end entity linking on domain-specific knowledge bases: Learning about ancient technologies from museum collections
Sebastian Cadavid-Sanchez, Khalil Kacem, Rafael Aparecido Martins Frade, Johannes Boehm, Thomas Chaney, Danial Lashkari, Daniel Simig
cs.CL, cs.LG
To study social, economic, and historical questions, researchers in the social sciences and humanities have started to use increasingly large unstructured textual datasets. While recent advances in NLP provide many tools to efficiently process such data, most existing approaches rely on generic solutions whose performa...
http://arxiv.org/pdf/2305.14588v1
2023-05-23T23:53:58
2023-05-23T23:53:58
{ "title": [ 0.013435364, 0.037139893, -0.0357666, -0.041015625, 0.052825928, -0.019134521, 0.013244629, -0.018600464, 0.0059890747, 0.003293991, -0.014221191, 0.012535095, -0.0045776367, -0.033081055, -0.007762909, 0.012512207, -0.0074539185, 0....
http://arxiv.org/abs/2305.14587v1
Contextualized Topic Coherence Metrics
Hamed Rahimi, Jacob Louis Hoover, David Mimno, Hubert Naacke, Camelia Constantin, Bernd Amann
cs.CL, cs.IR
The recent explosion in work on neural topic modeling has been criticized for optimizing automated topic evaluation metrics at the expense of actual meaningful topic identification. But human annotation remains expensive and time-consuming. We propose LLM-based methods inspired by standard human topic evaluations, in a...
http://arxiv.org/pdf/2305.14587v1
2023-05-23T23:53:29
2023-05-23T23:53:29
{ "title": [ -0.0076560974, 0.019943237, -0.05117798, -0.058776855, 0.0135650635, -0.011054993, -0.046783447, -0.03173828, 0.01222229, 0.064086914, 0.0036964417, 0.012763977, 0.03652954, -0.020706177, -0.023040771, 0.040008545, -0.0034217834, 0.0...
http://arxiv.org/abs/2305.14583v2
Natural Language Decompositions of Implicit Content Enable Better Text Representations
Alexander Hoyle, Rupak Sarkar, Pranav Goel, Philip Resnik
cs.CL
When people interpret text, they rely on inferences that go beyond the observed language itself. Inspired by this observation, we introduce a method for the analysis of text that takes implicitly communicated content explicitly into account. We use a large language model to produce sets of propositions that are inferen...
http://arxiv.org/pdf/2305.14583v2
2023-10-25T00:08:17
2023-05-23T23:45:20
{ "title": [ -0.0011472702, -0.010955811, -0.015991211, -0.070373535, -0.010314941, 0.007789612, -0.057891846, -0.005519867, 0.013450623, 0.03668213, -0.020904541, 0.015136719, 0.0055885315, -0.000038206577, -0.03201294, 0.015464783, -0.0054016113, ...
http://arxiv.org/abs/2305.14580v2
Evaluating OpenAI's Whisper ASR for Punctuation Prediction and Topic Modeling of life histories of the Museum of the Person
Lucas Rafael Stefanel Gris, Ricardo Marcacini, Arnaldo Candido Junior, Edresson Casanova, Anderson Soares, Sandra Maria Aluísio
cs.CL, cs.AI
Automatic speech recognition (ASR) systems play a key role in applications involving human-machine interactions. Despite their importance, ASR models for the Portuguese language proposed in the last decade have limitations in relation to the correct identification of punctuation marks in automatic transcriptions, which...
http://arxiv.org/pdf/2305.14580v2
2023-05-26T12:09:55
2023-05-23T23:37:29
{ "title": [ 0.014091492, 0.004383087, -0.008529663, -0.020431519, -0.03137207, 0.0008635521, -0.03591919, -0.033355713, -0.020645142, 0.041015625, -0.004951477, 0.01902771, 0.011802673, -0.02848816, 0.0041656494, -0.011260986, -0.0008907318, 0.0...
http://arxiv.org/abs/2305.14578v1
Connecting the Dots: What Graph-Based Text Representations Work Best for Text Classification using Graph Neural Networks?
Margarita Bugueño, Gerard de Melo
cs.CL
Given the success of Graph Neural Networks (GNNs) for structure-aware machine learning, numerous studies have explored their application to text classification, as an alternative to traditional feature representation models. However, most studies considered just a specific domain and validated on data with particular c...
http://arxiv.org/pdf/2305.14578v1
2023-05-23T23:31:24
2023-05-23T23:31:24
{ "title": [ -0.041656494, -0.0026416779, -0.028518677, -0.014961243, -0.016082764, -0.0052871704, -0.028076172, -0.013328552, -0.011528015, 0.034423828, -0.0037670135, 0.017333984, -0.017089844, -0.020751953, -0.03756714, -0.007575989, -0.021316528,...
http://arxiv.org/abs/2305.14577v2
Difference-Masking: Choosing What to Mask in Continued Pretraining
Alex Wilf, Syeda Nahida Akter, Leena Mathur, Paul Pu Liang, Sheryl Mathew, Mengrou Shou, Eric Nyberg, Louis-Philippe Morency
cs.LG, cs.CL
The self-supervised objective of masking-and-predicting has led to promising performance gains on a variety of downstream tasks. However, while most approaches randomly mask tokens, there is strong intuition that deciding what to mask can substantially improve learning outcomes. We investigate this in continued pretrai...
http://arxiv.org/pdf/2305.14577v2
2023-10-17T21:03:10
2023-05-23T23:31:02
{ "title": [ 0.01361084, 0.010650635, -0.057250977, -0.07342529, 0.015991211, -0.024230957, -0.026733398, -0.011634827, -0.035858154, 0.0026550293, -0.016189575, 0.033996582, -0.0103302, -0.04788208, 0.008430481, 0.0022659302, 0.0019159317, 0.034...
http://arxiv.org/abs/2305.14576v2
Parameter-Efficient Language Model Tuning with Active Learning in Low-Resource Settings
Josip Jukić, Jan Šnajder
cs.CL
Pre-trained language models (PLMs) have ignited a surge in demand for effective fine-tuning techniques, particularly in low-resource domains and languages. Active learning (AL), a set of algorithms designed to decrease labeling costs by minimizing label complexity, has shown promise in confronting the labeling bottlene...
http://arxiv.org/pdf/2305.14576v2
2023-10-23T13:58:26
2023-05-23T23:27:20
{ "title": [ -0.030395508, -0.008384705, -0.013809204, -0.06536865, -0.013282776, 0.00945282, -0.06530762, 0.008087158, -0.0075416565, 0.03842163, 0.00466156, 0.009399414, -0.002494812, -0.029464722, -0.032287598, 0.015014648, -0.0029029846, 0.03...
http://arxiv.org/abs/2305.14574v1
Detecting and Mitigating Indirect Stereotypes in Word Embeddings
Erin George, Joyce Chew, Deanna Needell
cs.CL, cs.LG
Societal biases in the usage of words, including harmful stereotypes, are frequently learned by common word embedding methods. These biases manifest not only between a word and an explicit marker of its stereotype, but also between words that share related stereotypes. This latter phenomenon, sometimes called "indirect...
http://arxiv.org/pdf/2305.14574v1
2023-05-23T23:23:49
2023-05-23T23:23:49
{ "title": [ 0.01638794, 0.0007529259, -0.025344849, -0.042663574, 0.017684937, -0.020385742, -0.009933472, -0.024627686, 0.00040864944, 0.024429321, -0.026428223, 0.017669678, 0.003435135, -0.007587433, -0.020965576, -0.006225586, -0.030807495, ...
http://arxiv.org/abs/2305.14571v2
From Characters to Words: Hierarchical Pre-trained Language Model for Open-vocabulary Language Understanding
Li Sun, Florian Luisier, Kayhan Batmanghelich, Dinei Florencio, Cha Zhang
cs.CL
Current state-of-the-art models for natural language understanding require a preprocessing step to convert raw text into discrete tokens. This process known as tokenization relies on a pre-built vocabulary of words or sub-word morphemes. This fixed vocabulary limits the model's robustness to spelling errors and its cap...
http://arxiv.org/pdf/2305.14571v2
2023-05-30T03:36:13
2023-05-23T23:22:20
{ "title": [ -0.021743774, -0.0067253113, -0.012954712, -0.085632324, -0.006931305, 0.024993896, -0.07678223, -0.0020713806, 0.0015239716, 0.018722534, -0.0158844, 0.0059013367, 0.012893677, -0.023239136, -0.030776978, 0.0018281937, -0.0050849915, ...
http://arxiv.org/abs/2305.14569v1
Few-shot Unified Question Answering: Tuning Models or Prompts?
Srijan Bansal, Semih Yavuz, Bo Pang, Meghana Bhat, Yingbo Zhou
cs.CL
Question-answering (QA) tasks often investigate specific question types, knowledge domains, or reasoning skills, leading to specialized models catering to specific categories of QA tasks. While recent research has explored the idea of unified QA models, such models are usually explored for high-resource scenarios and r...
http://arxiv.org/pdf/2305.14569v1
2023-05-23T23:14:38
2023-05-23T23:14:38
{ "title": [ 0.0006341934, -0.01727295, -0.009666443, -0.05456543, -0.017837524, 0.013046265, -0.031341553, 0.006801605, 0.008392334, 0.03274536, -0.0289917, 0.012306213, -0.0052490234, 0.0017528534, -0.061187744, -0.0033016205, 0.005760193, 0.01...
http://arxiv.org/abs/2305.14564v1
PEARL: Prompting Large Language Models to Plan and Execute Actions Over Long Documents
Simeng Sun, Yang Liu, Shuohang Wang, Chenguang Zhu, Mohit Iyyer
cs.CL
Strategies such as chain-of-thought prompting improve the performance of large language models (LLMs) on complex reasoning tasks by decomposing input examples into intermediate steps. However, it remains unclear how to apply such methods to reason over long input documents, in which both the decomposition and the outpu...
http://arxiv.org/pdf/2305.14564v1
2023-05-23T23:06:04
2023-05-23T23:06:04
{ "title": [ -0.031951904, -0.011413574, -0.054595947, -0.06958008, -0.019729614, 0.0037631989, -0.075683594, 0.0056114197, 0.002822876, 0.009742737, -0.04248047, 0.013221741, 0.018463135, -0.017822266, -0.007030487, -0.018875122, -0.00093460083, ...
http://arxiv.org/abs/2305.14556v1
Unraveling ChatGPT: A Critical Analysis of AI-Generated Goal-Oriented Dialogues and Annotations
Tiziano Labruna, Sofia Brenna, Andrea Zaninello, Bernardo Magnini
cs.CL, cs.AI
Large pre-trained language models have exhibited unprecedented capabilities in producing high-quality text via prompting techniques. This fact introduces new possibilities for data collection and annotation, particularly in situations where such data is scarce, complex to gather, expensive, or even sensitive. In this p...
http://arxiv.org/pdf/2305.14556v1
2023-05-23T22:31:01
2023-05-23T22:31:01
{ "title": [ -0.019561768, -0.013587952, -0.01939392, -0.019836426, -0.0046463013, -0.009300232, 0.004310608, -0.013786316, -0.033996582, 0.01966858, 0.0061531067, 0.020721436, -0.010383606, -0.025222778, -0.0178833, -0.001285553, -0.024047852, 0...
http://arxiv.org/abs/2305.14555v1
All Roads Lead to Rome? Exploring the Invariance of Transformers' Representations
Yuxin Ren, Qipeng Guo, Zhijing Jin, Shauli Ravfogel, Mrinmaya Sachan, Bernhard Schölkopf, Ryan Cotterell
cs.CL, cs.AI, cs.LG
Transformer models bring propelling advances in various NLP tasks, thus inducing lots of interpretability research on the learned representations of the models. However, we raise a fundamental question regarding the reliability of the representations. Specifically, we investigate whether transformers learn essentially ...
http://arxiv.org/pdf/2305.14555v1
2023-05-23T22:30:43
2023-05-23T22:30:43
{ "title": [ 0.00844574, 0.016815186, -0.02796936, -0.023529053, 0.024856567, -0.025665283, 0.0066184998, -0.03414917, -0.002796173, 0.03515625, -0.021133423, 0.020202637, -0.0005040169, -0.01953125, -0.030975342, 0.007827759, 0.0007338524, 0.035...
http://arxiv.org/abs/2305.14552v2
Sources of Hallucination by Large Language Models on Inference Tasks
Nick McKenna, Tianyi Li, Liang Cheng, Mohammad Javad Hosseini, Mark Johnson, Mark Steedman
cs.CL, cs.AI
Large Language Models (LLMs) are claimed to be capable of Natural Language Inference (NLI), necessary for applied tasks like question answering and summarization. We present a series of behavioral studies on several LLM families (LLaMA, GPT-3.5, and PaLM) which probe their behavior using controlled experiments. We esta...
http://arxiv.org/pdf/2305.14552v2
2023-10-22T21:22:29
2023-05-23T22:24:44
{ "title": [ -0.024368286, 0.0013093948, -0.02696228, -0.034942627, -0.01676941, -0.008178711, -0.030395508, -0.023025513, 0.022064209, 0.025054932, -0.014091492, -0.010887146, 0.012542725, -0.02407837, -0.050598145, 0.01374054, -0.009239197, 0.0...
http://arxiv.org/abs/2305.14548v1
Interpretable Automatic Fine-grained Inconsistency Detection in Text Summarization
Hou Pong Chan, Qi Zeng, Heng Ji
cs.CL
Existing factual consistency evaluation approaches for text summarization provide binary predictions and limited insights into the weakness of summarization systems. Therefore, we propose the task of fine-grained inconsistency detection, the goal of which is to predict the fine-grained types of factual errors in a summ...
http://arxiv.org/pdf/2305.14548v1
2023-05-23T22:11:47
2023-05-23T22:11:47
{ "title": [ -0.0058517456, 0.02267456, -0.025985718, -0.062164307, 0.008079529, 0.016647339, -0.04257202, -0.025665283, 0.0037384033, 0.041015625, -0.034362793, 0.023803711, 0.006965637, 0.00932312, -0.059753418, -0.007820129, -0.018478394, 0.03...
http://arxiv.org/abs/2305.14546v1
On the Transferability of Whisper-based Representations for "In-the-Wild" Cross-Task Downstream Speech Applications
Vamsikrishna Chemudupati, Marzieh Tahaei, Heitor Guimaraes, Arthur Pimentel, Anderson Avila, Mehdi Rezagholizadeh, Boxing Chen, Tiago Falk
eess.AS, cs.CL, cs.LG, cs.SD
Large self-supervised pre-trained speech models have achieved remarkable success across various speech-processing tasks. The self-supervised training of these models leads to universal speech representations that can be used for different downstream tasks, ranging from automatic speech recognition (ASR) to speaker iden...
http://arxiv.org/pdf/2305.14546v1
2023-05-23T22:02:55
2023-05-23T22:02:55
{ "title": [ 0.0067863464, -0.011795044, -0.014984131, -0.062683105, -0.01209259, -0.0013856888, -0.034484863, 0.0101623535, 0.006855011, 0.049194336, -0.023910522, 0.0011730194, 0.01374054, -0.013450623, -0.033721924, 0.010421753, -0.00011283159, ...
http://arxiv.org/abs/2305.14540v1
LLMs as Factual Reasoners: Insights from Existing Benchmarks and Beyond
Philippe Laban, Wojciech Kryściński, Divyansh Agarwal, Alexander R. Fabbri, Caiming Xiong, Shafiq Joty, Chien-Sheng Wu
cs.CL
With the recent appearance of LLMs in practical settings, having methods that can effectively detect factual inconsistencies is crucial to reduce the propagation of misinformation and improve trust in model outputs. When testing on existing factual consistency benchmarks, we find that a few large language models (LLMs)...
http://arxiv.org/pdf/2305.14540v1
2023-05-23T21:50:06
2023-05-23T21:50:06
{ "title": [ -0.030654907, -0.01739502, -0.03668213, -0.053955078, -0.034942627, -0.0015974045, -0.062194824, 0.0029029846, -0.02204895, 0.019241333, -0.010566711, 0.011734009, -0.0090789795, -0.03265381, -0.012870789, -0.0027275085, -0.018447876, ...
http://arxiv.org/abs/2305.14538v1
Cascaded Beam Search: Plug-and-Play Terminology-Forcing For Neural Machine Translation
Frédéric Odermatt, Béni Egressy, Roger Wattenhofer
cs.CL, cs.AI
This paper presents a plug-and-play approach for translation with terminology constraints. Terminology constraints are an important aspect of many modern translation pipelines. In both specialized domains and newly emerging domains (such as the COVID-19 pandemic), accurate translation of technical terms is crucial. Rec...
http://arxiv.org/pdf/2305.14538v1
2023-05-23T21:48:02
2023-05-23T21:48:02
{ "title": [ -0.022766113, -0.0052108765, -0.009025574, -0.041656494, -0.01348114, 0.01374054, -0.03466797, -0.008323669, -0.011039734, 0.0541687, -0.00762558, 0.011306763, 0.038848877, -0.013175964, -0.04058838, 0.022781372, -0.015037537, 0.0222...
http://arxiv.org/abs/2305.14536v2
MathDial: A Dialogue Tutoring Dataset with Rich Pedagogical Properties Grounded in Math Reasoning Problems
Jakub Macina, Nico Daheim, Sankalan Pal Chowdhury, Tanmay Sinha, Manu Kapur, Iryna Gurevych, Mrinmaya Sachan
cs.CL
While automatic dialogue tutors hold great potential in making education personalized and more accessible, research on such systems has been hampered by a lack of sufficiently large and high-quality datasets. Collecting such datasets remains challenging, as recording tutoring sessions raises privacy concerns and crowds...
http://arxiv.org/pdf/2305.14536v2
2023-10-23T12:00:01
2023-05-23T21:44:56
{ "title": [ -0.0031814575, -0.016799927, -0.034454346, -0.052734375, -0.010673523, -0.025482178, -0.025390625, 0.0077438354, -0.007965088, 0.008979797, -0.0052948, 0.0061950684, -0.0020885468, -0.025268555, -0.009780884, -0.014671326, -0.012069702, ...
http://arxiv.org/abs/2305.14534v1
Detecting Propaganda Techniques in Code-Switched Social Media Text
Muhammad Umar Salman, Asif Hanif, Shady Shehata, Preslav Nakov
cs.CL, cs.AI
Propaganda is a form of communication intended to influence the opinions and the mindset of the public to promote a particular agenda. With the rise of social media, propaganda has spread rapidly, leading to the need for automatic propaganda detection systems. Most work on propaganda detection has focused on high-resou...
http://arxiv.org/pdf/2305.14534v1
2023-05-23T21:37:26
2023-05-23T21:37:26
{ "title": [ 0.019500732, 0.015083313, 0.0015745163, -0.009773254, 0.023223877, -0.010261536, 0.02659607, -0.037597656, -0.05947876, 0.02684021, -0.0056114197, 0.026000977, -0.0028038025, -0.018203735, -0.010917664, -0.02104187, -0.0435791, 0.073...
http://arxiv.org/abs/2305.14533v1
How to Choose How to Choose Your Chatbot: A Massively Multi-System MultiReference Data Set for Dialog Metric Evaluation
Huda Khayrallah, Zuhaib Akhtar, Edward Cohen, João Sedoc
cs.CL
We release MMSMR, a Massively Multi-System MultiReference dataset to enable future work on metrics and evaluation for dialog. Automatic metrics for dialogue evaluation should be robust proxies for human judgments; however, the verification of robustness is currently far from satisfactory. To quantify the robustness cor...
http://arxiv.org/pdf/2305.14533v1
2023-05-23T21:33:43
2023-05-23T21:33:43
{ "title": [ -0.006713867, -0.019760132, -0.04348755, -0.028381348, -0.018188477, -0.02406311, -0.028030396, 0.0069389343, -0.009895325, 0.03414917, -0.025604248, 0.019454956, -0.018432617, -0.007827759, -0.03543091, -0.018432617, -0.007926941, 0...
http://arxiv.org/abs/2305.14521v1
Eliminating Spurious Correlations from Pre-trained Models via Data Mixing
Yihao Xue, Ali Payani, Yu Yang, Baharan Mirzasoleiman
cs.LG, cs.CL, cs.CV
Machine learning models pre-trained on large datasets have achieved remarkable convergence and robustness properties. However, these models often exploit spurious correlations between certain attributes and labels, which are prevalent in the majority of examples within specific categories but are not predictive of thes...
http://arxiv.org/pdf/2305.14521v1
2023-05-23T20:49:45
2023-05-23T20:49:45
{ "title": [ 0.0029067993, 0.017532349, -0.060577393, -0.048736572, 0.0309906, -0.018493652, -0.0256958, -0.02468872, -0.009468079, 0.035736084, -0.0019350052, 0.0072135925, -0.01184845, -0.023330688, -0.03692627, 0.001660347, 0.004310608, 0.0481...
http://arxiv.org/abs/2305.14507v1
Deduction under Perturbed Evidence: Probing Student Simulation Capabilities of Large Language Models
Shashank Sonkar, Richard G. Baraniuk
cs.CL
We explore whether Large Language Models (LLMs) are capable of logical reasoning with distorted facts, which we call Deduction under Perturbed Evidence (DUPE). DUPE presents a unique challenge to LLMs since they typically rely on their parameters, which encode mostly accurate information, to reason and make inferences....
http://arxiv.org/pdf/2305.14507v1
2023-05-23T20:26:03
2023-05-23T20:26:03
{ "title": [ -0.0015516281, 0.008956909, -0.007259369, -0.04827881, -0.0065307617, 0.0036029816, -0.031982422, -0.030075073, -0.0027446747, -0.017318726, -0.032196045, -0.003993988, -0.02079773, -0.027938843, -0.0065574646, -0.033416748, -0.032989502...
http://arxiv.org/abs/2305.14502v1
RetICL: Sequential Retrieval of In-Context Examples with Reinforcement Learning
Alexander Scarlatos, Andrew Lan
cs.CL, cs.AI, cs.LG
Many recent developments in large language models focus on prompting them to perform specific tasks. One effective prompting method is in-context learning, where the model performs a (possibly new) generation/prediction task given one (or more) examples. Past work has shown that the choice of examples can make a large ...
http://arxiv.org/pdf/2305.14502v1
2023-05-23T20:15:56
2023-05-23T20:15:56
{ "title": [ 0.028244019, 0.00044250488, -0.0027008057, -0.053710938, 0.014785767, -0.0093307495, -0.0042419434, -0.045013428, 0.00095272064, 0.029556274, -0.026565552, 0.003921509, -0.01777649, -0.025299072, -0.03741455, -0.009391785, 0.0033931732, ...
http://arxiv.org/abs/2305.14499v2
NAIL: Lexical Retrieval Indices with Efficient Non-Autoregressive Decoders
Livio Baldini Soares, Daniel Gillick, Jeremy R. Cole, Tom Kwiatkowski
cs.CL, cs.IR
Neural document rerankers are extremely effective in terms of accuracy. However, the best models require dedicated hardware for serving, which is costly and often not feasible. To avoid this serving-time requirement, we present a method of capturing up to 86% of the gains of a Transformer cross-attention model with a l...
http://arxiv.org/pdf/2305.14499v2
2023-10-23T14:46:34
2023-05-23T20:09:52
{ "title": [ -0.031585693, -0.007572174, -0.040771484, -0.043304443, -0.0018949509, 0.0007286072, -0.03540039, -0.011428833, 0.031036377, 0.052642822, -0.033721924, 0.00029921532, 0.036193848, 0.0039978027, -0.07458496, 0.018157959, -0.010635376, ...
http://arxiv.org/abs/2305.14497v1
Self-Polish: Enhance Reasoning in Large Language Models via Problem Refinement
Zhiheng Xi, Senjie Jin, Yuhao Zhou, Rui Zheng, Songyang Gao, Tao Gui, Qi Zhang, Xuanjing Huang
cs.CL, cs.AI
Prompting methods such as Chain-of-Thought (CoT) have shed new light on enhancing the reasoning capabilities of large language models, and researchers have extensively explored the generation process of rationales and answers. However, they have overlooked the potential challenges posed by the poor quality of reasoning...
http://arxiv.org/pdf/2305.14497v1
2023-05-23T19:58:30
2023-05-23T19:58:30
{ "title": [ -0.024719238, -0.024520874, -0.03137207, -0.09338379, -0.0030345917, 0.023544312, -0.07965088, 0.013511658, 0.008811951, 0.016998291, 0.0036792755, 0.00034689903, -0.013572693, -0.02407837, -0.04333496, 0.009300232, -0.011886597, 0.0...
http://arxiv.org/abs/2305.14493v3
Do prompt positions really matter?
Junyu Mao, Stuart E. Middleton, Mahesan Niranjan
cs.CL
Prompt-based models have gathered a lot of attention from researchers due to their remarkable advancements in the fields of zero-shot and few-shot learning. Developing an effective prompt template plays a critical role. However, prior studies have mainly focused on prompt vocabulary selection or embedding initializatio...
http://arxiv.org/pdf/2305.14493v3
2023-11-15T16:44:04
2023-05-23T19:45:45
{ "title": [ 0.01309967, -0.004207611, -0.013137817, 0.021850586, -0.048339844, -0.0178833, -0.014465332, -0.042297363, -0.015930176, 0.064086914, -0.0016374588, -0.06555176, 0.014732361, -0.022766113, -0.045684814, -0.029220581, 0.007041931, 0.0...
http://arxiv.org/abs/2305.14492v2
Sociocultural Norm Similarities and Differences via Situational Alignment and Explainable Textual Entailment
Sky CH-Wang, Arkadiy Saakyan, Oliver Li, Zhou Yu, Smaranda Muresan
cs.CL
Designing systems that can reason across cultures requires that they are grounded in the norms of the contexts in which they operate. However, current research on developing computational models of social norms has primarily focused on American society. Here, we propose a novel approach to discover and compare descript...
http://arxiv.org/pdf/2305.14492v2
2023-10-23T00:53:01
2023-05-23T19:43:47
{ "title": [ -0.0005469322, 0.01663208, -0.038116455, -0.061798096, 0.0052757263, -0.012245178, -0.06085205, -0.02418518, -0.009513855, 0.02935791, -0.036102295, 0.023422241, 0.04522705, -0.022506714, 0.0181427, 0.019058228, -0.0017490387, 0.0275...
http://arxiv.org/abs/2305.14489v2
Are Large Language Models Robust Coreference Resolvers?
Nghia T. Le, Alan Ritter
cs.CL
Recent work on extending coreference resolution across domains and languages relies on annotated data in both the target domain and language. At the same time, pre-trained large language models (LMs) have been reported to exhibit strong zero- and few-shot learning abilities across a wide range of NLP tasks. However, pr...
http://arxiv.org/pdf/2305.14489v2
2023-11-15T04:51:27
2023-05-23T19:38:28
{ "title": [ 0.008522034, 0.009552002, -0.022094727, -0.07800293, 0.0046920776, 0.0049057007, -0.05630493, -0.00026273727, -0.011604309, 0.060028076, -0.025268555, 0.0090789795, 0.03503418, 0.0016841888, -0.039367676, 0.023788452, -0.009841919, 0...
http://arxiv.org/abs/2305.14483v1
Language Model Self-improvement by Reinforcement Learning Contemplation
Jing-Cheng Pang, Pengyuan Wang, Kaiyuan Li, Xiong-Hui Chen, Jiacheng Xu, Zongzhang Zhang, Yang Yu
cs.CL, cs.LG
Large Language Models (LLMs) have exhibited remarkable performance across various natural language processing (NLP) tasks. However, fine-tuning these models often necessitates substantial supervision, which can be expensive and time-consuming to obtain. This paper introduces a novel unsupervised method called LanguageM...
http://arxiv.org/pdf/2305.14483v1
2023-05-23T19:25:52
2023-05-23T19:25:52
{ "title": [ -0.0020275116, -0.007965088, -0.045410156, 0.014572144, -0.02330017, -0.051513672, 0.0036239624, 0.008506775, -0.051086426, 0.031463623, 0.011726379, -0.0046691895, 0.012992859, -0.050598145, 0.013809204, -0.015235901, 0.013748169, 0...
http://arxiv.org/abs/2305.14482v2
Is a Prestigious Job the same as a Prestigious Country? A Case Study on Multilingual Sentence Embeddings and European Countries
Jindřich Libovický
cs.CL
We study how multilingual sentence representations capture European countries and occupations and how this differs across European languages. We prompt the models with templated sentences that we machine-translate into 12 European languages and analyze the most prominent dimensions in the embeddings.Our analysis reveal...
http://arxiv.org/pdf/2305.14482v2
2023-10-25T09:13:34
2023-05-23T19:24:42
{ "title": [ 0.030731201, -0.0071105957, -0.008628845, -0.01625061, 0.0003142357, -0.03189087, 0.002002716, -0.0071754456, -0.04748535, 0.03366089, -0.0074691772, 0.02281189, 0.017333984, -0.020401001, -0.011871338, -0.013343811, -0.020751953, 0....
http://arxiv.org/abs/2305.14481v2
FOCUS: Effective Embedding Initialization for Monolingual Specialization of Multilingual Models
Konstantin Dobler, Gerard de Melo
cs.CL
Using model weights pretrained on a high-resource language as a warm start can reduce the need for data and compute to obtain high-quality language models for other, especially low-resource, languages. However, if we want to use a new tokenizer specialized for the target language, we cannot transfer the source model's ...
http://arxiv.org/pdf/2305.14481v2
2023-11-06T17:47:47
2023-05-23T19:21:53
{ "title": [ -0.021499634, -0.011329651, -0.0129852295, -0.0826416, -0.0063591003, 0.024353027, -0.08001709, -0.0107421875, -0.003917694, 0.041137695, 0.008651733, 0.011398315, 0.0007929802, -0.026779175, -0.019515991, 0.012573242, -0.025360107, ...
http://arxiv.org/abs/2305.14480v2
BAND: Biomedical Alert News Dataset
Zihao Fu, Meiru Zhang, Zaiqiao Meng, Yannan Shen, David Buckeridge, Nigel Collier
cs.CL, cs.AI
Infectious disease outbreaks continue to pose a significant threat to human health and well-being. To improve disease surveillance and understanding of disease spread, several surveillance systems have been developed to monitor daily news alerts and social media. However, existing systems lack thorough epidemiological ...
http://arxiv.org/pdf/2305.14480v2
2023-10-15T15:09:24
2023-05-23T19:21:00
{ "title": [ 0.006336212, 0.00579834, -0.041625977, -0.05090332, -0.014389038, -0.0062179565, -0.051483154, 0.01687622, -0.045410156, 0.04800415, -0.012748718, 0.005744934, 0.047180176, -0.038146973, 0.017608643, -0.0025577545, 0.04336548, -0.021...
http://arxiv.org/abs/2305.14471v1
CGCE: A Chinese Generative Chat Evaluation Benchmark for General and Financial Domains
Xuanyu Zhang, Bingbing Li, Qing Yang
cs.CL
Generative chat models, such as ChatGPT and GPT-4, have revolutionized natural language generation (NLG) by incorporating instructions and human feedback to achieve significant performance improvements. However, the lack of standardized evaluation benchmarks for chat models, particularly for Chinese and domain-specific...
http://arxiv.org/pdf/2305.14471v1
2023-05-23T18:54:15
2023-05-23T18:54:15
{ "title": [ -0.0074882507, -0.0129470825, -0.021377563, -0.041503906, -0.013000488, -0.015136719, -0.035888672, -0.0090408325, 0.034851074, 0.052642822, -0.011497498, 0.0013475418, 0.02154541, -0.024276733, -0.06530762, 0.0026168823, -0.029129028, ...
http://arxiv.org/abs/2305.14468v1
Run Like a Girl! Sports-Related Gender Bias in Language and Vision
Sophia Harrison, Eleonora Gualdoni, Gemma Boleda
cs.CV, cs.CL
Gender bias in Language and Vision datasets and models has the potential to perpetuate harmful stereotypes and discrimination. We analyze gender bias in two Language and Vision datasets. Consistent with prior work, we find that both datasets underrepresent women, which promotes their invisibilization. Moreover, we hypo...
http://arxiv.org/pdf/2305.14468v1
2023-05-23T18:52:11
2023-05-23T18:52:11
{ "title": [ -0.017822266, -0.020889282, -0.03982544, 0.011894226, -0.024780273, -0.04916382, 0.014823914, -0.008720398, -0.03540039, 0.04724121, -0.04547119, 0.005027771, 0.0158844, -0.008232117, -0.026779175, -0.029846191, -0.0010814667, 0.0353...
http://arxiv.org/abs/2305.14463v2
ReadMe++: Benchmarking Multilingual Language Models for Multi-Domain Readability Assessment
Tarek Naous, Michael J. Ryan, Anton Lavrouk, Mohit Chandra, Wei Xu
cs.CL, cs.AI, cs.LG
We present a systematic study and comprehensive evaluation of large language models for automatic multilingual readability assessment. In particular, we construct ReadMe++, a multilingual multi-domain dataset with human annotations of 9757 sentences in Arabic, English, French, Hindi, and Russian collected from 112 diff...
http://arxiv.org/pdf/2305.14463v2
2023-11-15T15:50:31
2023-05-23T18:37:30
{ "title": [ -0.048736572, 0.0014705658, -0.0209198, -0.04159546, -0.005104065, 0.014633179, -0.064208984, -0.016693115, 0.00349617, 0.038116455, 0.0007572174, 0.0021324158, 0.016738892, -0.013511658, -0.05630493, 0.010650635, -0.03277588, 0.0592...
http://arxiv.org/abs/2305.14459v2
Enhancing Generation through Summarization Duality and Explicit Outline Control
Yunzhe Li, Qian Chen, Weixiang Yan, Wen Wang, Qinglin Zhang, Hari Sundaram
cs.CL, cs.AI
Automatically open-ended long text generation poses significant challenges due to semantic incoherence and plot implausibility. Previous works usually alleviate this problem through outlines in the form of short phrases or abstractive signals by designing unsupervised tasks, which tend to be unstable and weakly interpr...
http://arxiv.org/pdf/2305.14459v2
2023-09-16T05:21:07
2023-05-23T18:33:52
{ "title": [ 0.009002686, 0.013656616, -0.03869629, -0.08630371, 0.00004643202, 0.008270264, -0.07122803, -0.020858765, 0.0047836304, 0.04888916, 0.021713257, 0.022781372, 0.007106781, -0.035125732, -0.012680054, 0.03050232, -0.008956909, 0.00906...
http://arxiv.org/abs/2305.14458v2
Dancing Between Success and Failure: Edit-level Simplification Evaluation using SALSA
David Heineman, Yao Dou, Mounica Maddela, Wei Xu
cs.CL
Large language models (e.g., GPT-4) are uniquely capable of producing highly rated text simplification, yet current human evaluation methods fail to provide a clear understanding of systems' specific strengths and weaknesses. To address this limitation, we introduce SALSA, an edit-based human annotation framework that ...
http://arxiv.org/pdf/2305.14458v2
2023-10-22T18:25:46
2023-05-23T18:30:49
{ "title": [ -0.021224976, -0.012786865, -0.042510986, -0.06945801, -0.0052719116, 0.000027179718, -0.06463623, -0.0003950596, 0.0062217712, 0.058624268, 0.0032291412, -0.007484436, 0.0016155243, -0.028030396, -0.055267334, -0.012802124, -0.002162933...
http://arxiv.org/abs/2305.16337v2
Handling Realistic Label Noise in BERT Text Classification
Maha Tufail Agro, Hanan Aldarmaki
cs.CL, cs.AI
Labels noise refers to errors in training labels caused by cheap data annotation methods, such as web scraping or crowd-sourcing, which can be detrimental to the performance of supervised classifiers. Several methods have been proposed to counteract the effect of random label noise in supervised classification, and som...
http://arxiv.org/pdf/2305.16337v2
2023-10-20T11:26:43
2023-05-23T18:30:31
{ "title": [ -0.017501831, -0.0031757355, -0.01828003, -0.039031982, -0.010444641, -0.0037250519, -0.02130127, -0.008392334, -0.0009126663, 0.02671814, -0.029525757, 0.025665283, -0.0030117035, -0.009963989, -0.043548584, 0.0055999756, -0.023483276, ...
http://arxiv.org/abs/2305.14457v2
Pre-training Language Models for Comparative Reasoning
Mengxia Yu, Zhihan Zhang, Wenhao Yu, Meng Jiang
cs.CL
Comparative reasoning is a process of comparing objects, concepts, or entities to draw conclusions, which constitutes a fundamental cognitive ability. In this paper, we propose a novel framework to pre-train language models for enhancing their abilities of comparative reasoning over texts. While there have been approac...
http://arxiv.org/pdf/2305.14457v2
2023-10-30T20:11:05
2023-05-23T18:28:42
{ "title": [ -0.026229858, -0.031951904, -0.0064086914, -0.06713867, -0.01826477, 0.02003479, -0.05215454, -0.0024375916, -0.010215759, 0.029571533, -0.008377075, -0.00096559525, -0.01109314, -0.020980835, -0.035980225, -0.0055274963, -0.008033752, ...
http://arxiv.org/abs/2305.14456v1
Having Beer after Prayer? Measuring Cultural Bias in Large Language Models
Tarek Naous, Michael J. Ryan, Wei Xu
cs.CL, cs.AI, cs.LG
Are language models culturally biased? It is important that language models conform to the cultural aspects of the communities they serve. However, we show in this paper that language models suffer from a significant bias towards Western culture when handling and generating text in Arabic, often preferring, and produci...
http://arxiv.org/pdf/2305.14456v1
2023-05-23T18:27:51
2023-05-23T18:27:51
{ "title": [ -0.0052375793, 0.008964539, -0.026931763, -0.021957397, 0.013664246, -0.035125732, -0.0231781, -0.035827637, -0.004863739, 0.035461426, -0.056518555, 0.010353088, 0.019897461, 0.0021381378, -0.03074646, 0.0023384094, -0.012191772, 0....
http://arxiv.org/abs/2305.14453v2
On Robustness of Finetuned Transformer-based NLP Models
Pavan Kalyan Reddy Neerudu, Subba Reddy Oota, Mounika Marreddy, Venkateswara Rao Kagita, Manish Gupta
cs.CL
Transformer-based pretrained models like BERT, GPT-2 and T5 have been finetuned for a large number of natural language processing (NLP) tasks, and have been shown to be very effective. However, while finetuning, what changes across layers in these models with respect to pretrained checkpoints is under-studied. Further,...
http://arxiv.org/pdf/2305.14453v2
2023-11-08T16:46:20
2023-05-23T18:25:18
{ "title": [ 0.005809784, 0.012863159, -0.015136719, -0.055480957, 0.017791748, 0.009109497, -0.048797607, -0.029693604, 0.011169434, 0.029190063, -0.03451538, -0.013442993, 0.014274597, 0.0026664734, -0.06286621, -0.0010814667, -0.02243042, 0.04...
http://arxiv.org/abs/2305.14450v1
Is Information Extraction Solved by ChatGPT? An Analysis of Performance, Evaluation Criteria, Robustness and Errors
Ridong Han, Tao Peng, Chaohao Yang, Benyou Wang, Lu Liu, Xiang Wan
cs.CL
ChatGPT has stimulated the research boom in the field of large language models. In this paper, we assess the capabilities of ChatGPT from four perspectives including Performance, Evaluation Criteria, Robustness and Error Types. Specifically, we first evaluate ChatGPT's performance on 17 datasets with 14 IE sub-tasks un...
http://arxiv.org/pdf/2305.14450v1
2023-05-23T18:17:43
2023-05-23T18:17:43
{ "title": [ -0.030593872, -0.015411377, -0.027862549, -0.022949219, -0.009521484, 0.01322937, -0.026977539, -0.023803711, -0.007873535, 0.018676758, -0.022064209, 0.036865234, -0.026855469, -0.003540039, -0.03543091, 0.0072021484, -0.030059814, ...
http://arxiv.org/abs/2305.14441v1
Exploring Contrast Consistency of Open-Domain Question Answering Systems on Minimally Edited Questions
Zhihan Zhang, Wenhao Yu, Zheng Ning, Mingxuan Ju, Meng Jiang
cs.CL
Contrast consistency, the ability of a model to make consistently correct predictions in the presence of perturbations, is an essential aspect in NLP. While studied in tasks such as sentiment analysis and reading comprehension, it remains unexplored in open-domain question answering (OpenQA) due to the difficulty of co...
http://arxiv.org/pdf/2305.14441v1
2023-05-23T18:07:04
2023-05-23T18:07:04
{ "title": [ 0.008003235, 0.015472412, -0.0335083, -0.044311523, 0.013870239, -0.009918213, 0.00040364265, -0.010627747, -0.008392334, 0.038330078, -0.017440796, 0.0014123917, 0.004295349, -0.0043029785, -0.059020996, -0.017684937, -0.020736694, ...
http://arxiv.org/abs/2305.14434v1
Domain-Expanded ASTE: Rethinking Generalization in Aspect Sentiment Triplet Extraction
Yew Ken Chia, Hui Chen, Wei Han, Guizhen Chen, Sharifah Mahani Aljunied, Soujanya Poria, Lidong Bing
cs.CL
Aspect Sentiment Triplet Extraction (ASTE) is a subtask of Aspect-Based Sentiment Analysis (ABSA) that considers each opinion term, their expressed sentiment, and the corresponding aspect targets. However, existing methods are limited to the in-domain setting with two domains. Hence, we propose a domain-expanded benchm...
http://arxiv.org/pdf/2305.14434v1
2023-05-23T18:01:49
2023-05-23T18:01:49
{ "title": [ 0.016983032, 0.007434845, -0.022033691, -0.07635498, 0.013450623, -0.00349617, -0.028045654, -0.0056114197, 0.035583496, 0.06088257, -0.009925842, -0.004737854, 0.015159607, 0.006828308, -0.091674805, 0.015510559, -0.010360718, 0.052...
http://arxiv.org/abs/2305.14342v3
Sophia: A Scalable Stochastic Second-order Optimizer for Language Model Pre-training
Hong Liu, Zhiyuan Li, David Hall, Percy Liang, Tengyu Ma
cs.LG, cs.CL, math.OC
Given the massive cost of language model pre-training, a non-trivial improvement of the optimization algorithm would lead to a material reduction on the time and cost of training. Adam and its variants have been state-of-the-art for years, and more sophisticated second-order (Hessian-based) optimizers often incur too m...
http://arxiv.org/pdf/2305.14342v3
2023-10-17T07:44:16
2023-05-23T17:59:21
{ "title": [ -0.036499023, -0.007381439, -0.026031494, -0.05984497, -0.01499939, 0.013961792, -0.08905029, -0.008972168, -0.03390503, 0.03491211, 0.0018568039, 0.021453857, 0.009124756, -0.02508545, -0.018493652, 0.00041890144, -0.006690979, 0.03...
http://arxiv.org/abs/2305.14341v1
APPLS: A Meta-evaluation Testbed for Plain Language Summarization
Yue Guo, Tal August, Gondy Leroy, Trevor Cohen, Lucy Lu Wang
cs.CL
While there has been significant development of models for Plain Language Summarization (PLS), evaluation remains a challenge. This is in part because PLS involves multiple, interrelated language transformations (e.g., adding background explanations, removing specialized terminology). No metrics are explicitly engineer...
http://arxiv.org/pdf/2305.14341v1
2023-05-23T17:59:19
2023-05-23T17:59:19
{ "title": [ -0.01184082, -0.009963989, -0.01537323, -0.08770752, -0.00919342, 0.026306152, -0.07684326, -0.011474609, 0.014060974, 0.047973633, 0.00484848, 0.031799316, -0.0093688965, -0.0049858093, -0.06347656, 0.027160645, -0.01096344, 0.03579...
http://arxiv.org/abs/2305.14410v2
Image Manipulation via Multi-Hop Instructions -- A New Dataset and Weakly-Supervised Neuro-Symbolic Approach
Harman Singh, Poorva Garg, Mohit Gupta, Kevin Shah, Ashish Goswami, Satyam Modi, Arnab Kumar Mondal, Dinesh Khandelwal, Dinesh Garg, Parag Singla
cs.CV, cs.AI, cs.CL
We are interested in image manipulation via natural language text -- a task that is useful for multiple AI applications but requires complex reasoning over multi-modal spaces. We extend recently proposed Neuro Symbolic Concept Learning (NSCL), which has been quite effective for the task of Visual Question Answering (VQ...
http://arxiv.org/pdf/2305.14410v2
2023-10-24T20:44:09
2023-05-23T17:59:10
{ "title": [ -0.004550934, 0.0013198853, 0.0050621033, 0.01159668, -0.0031375885, -0.01499939, 0.009254456, -0.024871826, -0.009002686, 0.056243896, 0.00032281876, 0.0042686462, 0.02255249, -0.029067993, -0.02520752, 0.0052986145, 0.015289307, 0....
http://arxiv.org/abs/2305.14337v2
Anchor Prediction: Automatic Refinement of Internet Links
Nelson F. Liu, Kenton Lee, Kristina Toutanova
cs.CL, cs.IR
Internet links enable users to deepen their understanding of a topic by providing convenient access to related information. However, the majority of links are unanchored -- they link to a target webpage as a whole, and readers may expend considerable effort localizing the specific parts of the target webpage that enric...
http://arxiv.org/pdf/2305.14337v2
2023-05-24T07:12:33
2023-05-23T17:58:21
{ "title": [ 0.0014686584, -0.0017147064, -0.026245117, -0.049804688, 0.012046814, -0.010017395, -0.024734497, -0.030883789, -0.014625549, 0.04949951, -0.005050659, -0.0051994324, 0.02218628, -0.017074585, -0.03314209, 0.0039405823, -0.008178711, ...
http://arxiv.org/abs/2305.14336v2
Schema-Driven Information Extraction from Heterogeneous Tables
Fan Bai, Junmo Kang, Gabriel Stanovsky, Dayne Freitag, Alan Ritter
cs.CL
In this paper, we explore the question of whether large language models can support cost-efficient information extraction from tables. We introduce schema-driven information extraction, a new task that transforms tabular data into structured records following a human-authored schema. To assess various LLM's capabilitie...
http://arxiv.org/pdf/2305.14336v2
2023-11-15T18:56:34
2023-05-23T17:58:10
{ "title": [ -0.026229858, 0.009231567, -0.08105469, -0.07537842, 0.028793335, -0.0038318634, -0.040008545, 0.0064086914, 0.014968872, 0.035369873, -0.03945923, 0.0077667236, -0.017623901, 0.005065918, -0.04925537, -0.011711121, -0.004470825, 0.0...
http://arxiv.org/abs/2305.14333v2
Automatic Model Selection with Large Language Models for Reasoning
James Xu Zhao, Yuxi Xie, Kenji Kawaguchi, Junxian He, Michael Qizhe Xie
cs.CL, cs.AI
Chain-of-Thought (CoT) and Program-Aided Language Models (PAL) represent two distinct reasoning methods, each with its own strengths. CoT employs natural language, offering flexibility and interpretability, while PAL utilizes programming language, yielding more structured and rigorous logic. We introduce a model select...
http://arxiv.org/pdf/2305.14333v2
2023-10-23T14:46:07
2023-05-23T17:57:59
{ "title": [ -0.042541504, -0.016281128, -0.050994873, -0.074035645, -0.0016307831, 0.021621704, -0.0869751, -0.0043563843, 0.010444641, 0.0104904175, -0.021499634, -0.012145996, -0.0102005005, -0.024597168, -0.022415161, 0.001958847, -0.0077285767, ...
http://arxiv.org/abs/2305.14332v2
Evaluating and Modeling Attribution for Cross-Lingual Question Answering
Benjamin Muller, John Wieting, Jonathan H. Clark, Tom Kwiatkowski, Sebastian Ruder, Livio Baldini Soares, Roee Aharoni, Jonathan Herzig, Xinyi Wang
cs.CL
Trustworthy answer content is abundant in many high-resource languages and is instantly accessible through question answering systems, yet this content can be hard to access for those that do not speak these languages. The leap forward in cross-lingual modeling quality offered by generative language models offers much ...
http://arxiv.org/pdf/2305.14332v2
2023-11-15T17:15:31
2023-05-23T17:57:46
{ "title": [ -0.00207901, -0.0077667236, -0.012504578, -0.03378296, -0.0020313263, 0.004234314, -0.036376953, -0.030578613, 0.0051231384, 0.025634766, -0.0143966675, 0.0019187927, 0.0018854141, -0.0073547363, -0.044921875, 0.005088806, -0.020767212, ...
http://arxiv.org/abs/2305.14331v2
What Else Do I Need to Know? The Effect of Background Information on Users' Reliance on QA Systems
Navita Goyal, Eleftheria Briakou, Amanda Liu, Connor Baumler, Claire Bonial, Jeffrey Micher, Clare R. Voss, Marine Carpuat, Hal Daumé III
cs.CL, cs.AI
NLP systems have shown impressive performance at answering questions by retrieving relevant context. However, with the increasingly large models, it is impossible and often undesirable to constrain models' knowledge or reasoning to only the retrieved context. This leads to a mismatch between the information that the mo...
http://arxiv.org/pdf/2305.14331v2
2023-10-26T01:01:11
2023-05-23T17:57:12
{ "title": [ 0.04446411, 0.010620117, -0.036102295, -0.081726074, 0.016326904, -0.043395996, -0.031433105, 0.009628296, -0.021774292, 0.040496826, -0.060699463, 0.023956299, -0.023895264, -0.008728027, -0.027801514, -0.03390503, 0.028198242, 0.01...
http://arxiv.org/abs/2305.14330v2
Large Language Models are Frame-level Directors for Zero-shot Text-to-Video Generation
Susung Hong, Junyoung Seo, Sunghwan Hong, Heeseong Shin, Seungryong Kim
cs.CV, cs.AI, cs.CL
In the paradigm of AI-generated content (AIGC), there has been increasing attention in extending pre-trained text-to-image (T2I) models to text-to-video (T2V) generation. Despite their effectiveness, these frameworks face challenges in maintaining consistent narratives and handling rapid shifts in scene composition or ...
http://arxiv.org/pdf/2305.14330v2
2023-06-01T04:14:59
2023-05-23T17:57:09
{ "title": [ 0.0034217834, -0.051757812, -0.034179688, -0.064941406, -0.07409668, -0.011878967, -0.047454834, 0.041412354, 0.0021915436, 0.050201416, -0.033111572, -0.0020885468, 0.022445679, -0.025283813, -0.010047913, -0.018630981, 0.0014629364, ...
http://arxiv.org/abs/2305.14328v1
Empowering LLM-based Machine Translation with Cultural Awareness
Binwei Yao, Ming Jiang, Diyi Yang, Junjie Hu
cs.CL
Traditional neural machine translation (NMT) systems often fail to translate sentences that contain culturally specific information. Most previous NMT methods have incorporated external cultural knowledge during training, which requires fine-tuning on low-frequency items specific to the culture. Recent in-context learn...
http://arxiv.org/pdf/2305.14328v1
2023-05-23T17:56:33
2023-05-23T17:56:33
{ "title": [ -0.049987793, -0.01802063, -0.028671265, -0.031585693, -0.036315918, -0.004005432, -0.036834717, 0.011764526, -0.002796173, 0.046020508, -0.014381409, 0.00027108192, 0.03466797, -0.03213501, -0.03378296, 0.022232056, -0.0015220642, 0...
http://arxiv.org/abs/2305.14327v2
Dynosaur: A Dynamic Growth Paradigm for Instruction-Tuning Data Curation
Da Yin, Xiao Liu, Fan Yin, Ming Zhong, Hritik Bansal, Jiawei Han, Kai-Wei Chang
cs.CL, cs.AI
Instruction tuning has emerged to enhance the capabilities of large language models (LLMs) to comprehend instructions and generate appropriate responses. Existing methods either manually annotate or employ LLM (e.g., GPT-series) to generate data for instruction tuning. However, they often overlook associating instructi...
http://arxiv.org/pdf/2305.14327v2
2023-10-26T05:10:18
2023-05-23T17:56:26
{ "title": [ -0.045715332, -0.017700195, -0.06500244, -0.05291748, -0.005180359, -0.009994507, -0.027923584, 0.015464783, -0.0049972534, 0.018692017, -0.05407715, 0.009262085, -0.055023193, 0.001789093, -0.043304443, -0.036010742, 0.009346008, 0....
http://arxiv.org/abs/2305.14326v2
TalkUp: Paving the Way for Understanding Empowering Language
Lucille Njoo, Chan Young Park, Octavia Stappart, Marvin Thielk, Yi Chu, Yulia Tsvetkov
cs.CL
Empowering language is important in many real-world contexts, from education to workplace dynamics to healthcare. Though language technologies are growing more prevalent in these contexts, empowerment has seldom been studied in NLP, and moreover, it is inherently challenging to operationalize because of its implicit na...
http://arxiv.org/pdf/2305.14326v2
2023-10-23T06:37:47
2023-05-23T17:55:34
{ "title": [ -0.024108887, -0.002565384, -0.06903076, -0.06585693, 0.014404297, -0.009689331, -0.03692627, 0.015640259, 0.0046958923, 0.0134887695, -0.021057129, 0.0026931763, 0.011566162, -0.029785156, -0.01109314, 0.011131287, -0.021514893, 0.0...
http://arxiv.org/abs/2305.14325v1
Improving Factuality and Reasoning in Language Models through Multiagent Debate
Yilun Du, Shuang Li, Antonio Torralba, Joshua B. Tenenbaum, Igor Mordatch
cs.CL, cs.AI, cs.CV, cs.LG
Large language models (LLMs) have demonstrated remarkable capabilities in language generation, understanding, and few-shot learning in recent years. An extensive body of work has explored how their performance may be further improved through the tools of prompting, ranging from verification, self-consistency, or interm...
http://arxiv.org/pdf/2305.14325v1
2023-05-23T17:55:11
2023-05-23T17:55:11
{ "title": [ -0.0048599243, -0.00819397, -0.029953003, -0.07373047, -0.0055236816, -0.0041999817, -0.056427002, 0.009971619, -0.0036334991, 0.030853271, 0.0010738373, -0.0023612976, 0.02633667, -0.020996094, -0.041137695, 0.0066947937, -0.036102295, ...
http://arxiv.org/abs/2305.14324v2
Ties Matter: Meta-Evaluating Modern Metrics with Pairwise Accuracy and Tie Calibration
Daniel Deutsch, George Foster, Markus Freitag
cs.CL
Kendall's tau is frequently used to meta-evaluate how well machine translation (MT) evaluation metrics score individual translations. Its focus on pairwise score comparisons is intuitive but raises the question of how ties should be handled, a gray area that has motivated different variants in the literature. We demons...
http://arxiv.org/pdf/2305.14324v2
2023-10-17T16:33:33
2023-05-23T17:54:57
{ "title": [ -0.031219482, -0.033569336, -0.021347046, -0.017349243, -0.010856628, -0.0015411377, -0.018371582, -0.04537964, -0.016601562, 0.027954102, -0.0041770935, 0.0014286041, -0.0121154785, -0.03338623, -0.044891357, 0.0002142191, -0.007194519,...
http://arxiv.org/abs/2305.14323v3
ChatCoT: Tool-Augmented Chain-of-Thought Reasoning on Chat-based Large Language Models
Zhipeng Chen, Kun Zhou, Beichen Zhang, Zheng Gong, Wayne Xin Zhao, Ji-Rong Wen
cs.CL
Although large language models (LLMs) have achieved excellent performance in a variety of evaluation benchmarks, they still struggle in complex reasoning tasks which require specific knowledge and multi-hop reasoning. To improve the reasoning abilities, we propose ChatCoT, a tool-augmented chain-of-thought reasoning fr...
http://arxiv.org/pdf/2305.14323v3
2023-11-06T11:00:26
2023-05-23T17:54:33
{ "title": [ -0.017669678, -0.041931152, -0.0015506744, -0.07727051, -0.03111267, 0.03439331, -0.05783081, 0.018539429, 0.025543213, 0.02180481, -0.024856567, -0.007080078, 0.026000977, -0.019119263, -0.04449463, 0.00046277046, -0.021606445, 0.04...
http://arxiv.org/abs/2305.14322v1
RET-LLM: Towards a General Read-Write Memory for Large Language Models
Ali Modarressi, Ayyoob Imani, Mohsen Fayyaz, Hinrich Schütze
cs.CL
Large language models (LLMs) have significantly advanced the field of natural language processing (NLP) through their extensive parameters and comprehensive data utilization. However, existing LLMs lack a dedicated memory unit, limiting their ability to explicitly store and retrieve knowledge for various tasks. In this...
http://arxiv.org/pdf/2305.14322v1
2023-05-23T17:53:38
2023-05-23T17:53:38
{ "title": [ -0.05670166, -0.037994385, -0.042144775, -0.07678223, -0.02720642, 0.022521973, -0.07006836, 0.030532837, 0.016983032, 0.045043945, -0.013542175, -0.0019798279, 0.009498596, -0.011962891, -0.060516357, 0.026687622, -0.014038086, 0.00...
http://arxiv.org/abs/2305.14321v1
ConGraT: Self-Supervised Contrastive Pretraining for Joint Graph and Text Embeddings
William Brannon, Suyash Fulay, Hang Jiang, Wonjune Kang, Brandon Roy, Jad Kabbara, Deb Roy
cs.CL
We propose ConGraT(Contrastive Graph-Text pretraining), a general, self-supervised method for jointly learning separate representations of texts and nodes in a parent (or ``supervening'') graph, where each text is associated with one of the nodes. Datasets fitting this paradigm are common, from social media (users and ...
http://arxiv.org/pdf/2305.14321v1
2023-05-23T17:53:30
2023-05-23T17:53:30
{ "title": [ 0.004940033, 0.0062294006, -0.028793335, -0.04711914, 0.002023697, -0.010856628, -0.027053833, -0.020187378, 0.0071487427, 0.014343262, -0.011314392, 0.035583496, -0.023117065, -0.024673462, -0.014404297, 0.010643005, -0.012565613, 0...
http://arxiv.org/abs/2305.14318v2
CREATOR: Tool Creation for Disentangling Abstract and Concrete Reasoning of Large Language Models
Cheng Qian, Chi Han, Yi R. Fung, Yujia Qin, Zhiyuan Liu, Heng Ji
cs.CL
Large Language Models (LLMs) have made significant progress in utilizing tools, but their ability is limited by API availability and the instability of implicit reasoning, particularly when both planning and execution are involved. To overcome these limitations, we propose CREATOR, a novel framework that enables LLMs t...
http://arxiv.org/pdf/2305.14318v2
2023-10-08T14:12:12
2023-05-23T17:51:52
{ "title": [ -0.013435364, -0.023071289, -0.05886841, -0.066833496, -0.039031982, 0.0017242432, -0.0769043, 0.021194458, -0.0042686462, 0.03253174, -0.030197144, -0.019454956, 0.027648926, -0.036743164, -0.016189575, -0.02960205, -0.0059127808, 0...
http://arxiv.org/abs/2305.14310v2
Navigating Prompt Complexity for Zero-Shot Classification: A Study of Large Language Models in Computational Social Science
Yida Mu, Ben P. Wu, William Thorne, Ambrose Robinson, Nikolaos Aletras, Carolina Scarton, Kalina Bontcheva, Xingyi Song
cs.CL
Instruction-tuned Large Language Models (LLMs) have exhibited impressive language understanding and the capacity to generate responses that follow specific prompts. However, due to the computational demands associated with training these models, their applications often adopt a zero-shot setting. In this paper, we eval...
http://arxiv.org/pdf/2305.14310v2
2023-09-20T19:53:48
2023-05-23T17:48:21
{ "title": [ -0.00066900253, -0.011634827, -0.02558899, -0.04876709, -0.0065689087, -0.014549255, -0.03427124, -0.0021953583, -0.0002565384, 0.02935791, -0.010749817, 0.014190674, -0.01739502, -0.013999939, -0.0513916, -0.008399963, -0.013183594, ...
http://arxiv.org/abs/2305.14307v1
Debiasing should be Good and Bad: Measuring the Consistency of Debiasing Techniques in Language Models
Robert Morabito, Jad Kabbara, Ali Emami
cs.CL, cs.AI
Debiasing methods that seek to mitigate the tendency of Language Models (LMs) to occasionally output toxic or inappropriate text have recently gained traction. In this paper, we propose a standardized protocol which distinguishes methods that yield not only desirable results, but are also consistent with their mechanis...
http://arxiv.org/pdf/2305.14307v1
2023-05-23T17:45:54
2023-05-23T17:45:54
{ "title": [ -0.034423828, 0.01109314, -0.03955078, -0.061798096, 0.012252808, 0.0051193237, -0.076293945, -0.035583496, -0.0042648315, 0.029205322, -0.018096924, -0.0066947937, 0.017044067, -0.021621704, -0.05105591, -0.00055503845, -0.025604248, ...
http://arxiv.org/abs/2305.14303v2
QTSumm: Query-Focused Summarization over Tabular Data
Yilun Zhao, Zhenting Qi, Linyong Nan, Boyu Mi, Yixin Liu, Weijin Zou, Simeng Han, Ruizhe Chen, Xiangru Tang, Yumo Xu, Dragomir Radev, Arman Cohan
cs.CL
People primarily consult tables to conduct data analysis or answer specific questions. Text generation systems that can provide accurate table summaries tailored to users' information needs can facilitate more efficient access to relevant data insights. Motivated by this, we define a new query-focused table summarizati...
http://arxiv.org/pdf/2305.14303v2
2023-11-07T04:53:07
2023-05-23T17:43:51
{ "title": [ 0.0049438477, -0.0020103455, -0.035827637, -0.09887695, 0.016036987, 0.0016469955, -0.047180176, 0.00074625015, 0.012321472, 0.057617188, -0.017089844, 0.0064964294, 0.024658203, -0.00094509125, -0.0647583, 0.01007843, 0.00029587746, ...
http://arxiv.org/abs/2305.14299v1
TaDSE: Template-aware Dialogue Sentence Embeddings
Minsik Oh, Jiwei Li, Guoyin Wang
cs.CL, cs.AI
Learning high quality sentence embeddings from dialogues has drawn increasing attentions as it is essential to solve a variety of dialogue-oriented tasks with low annotation cost. However, directly annotating and gathering utterance relationships in conversations are difficult, while token-level annotations, \eg, entit...
http://arxiv.org/pdf/2305.14299v1
2023-05-23T17:40:41
2023-05-23T17:40:41
{ "title": [ -0.0084991455, -0.009056091, -0.026229858, -0.0635376, -0.0033836365, 0.010116577, -0.058135986, 0.007789612, 0.012771606, 0.023635864, 0.0037593842, 0.007724762, 0.004348755, -0.01826477, -0.041229248, 0.014015198, -0.024429321, 0.0...
http://arxiv.org/abs/2305.14296v1
USB: A Unified Summarization Benchmark Across Tasks and Domains
Kundan Krishna, Prakhar Gupta, Sanjana Ramprasad, Byron C. Wallace, Jeffrey P. Bigham, Zachary C. Lipton
cs.CL, cs.LG
An abundance of datasets exist for training and evaluating models on the task of summary generation.However, these datasets are often derived heuristically, and lack sufficient annotations to support research into all aspects of summarization, such as evidence extraction and controllable summarization. We introduce a b...
http://arxiv.org/pdf/2305.14296v1
2023-05-23T17:39:54
2023-05-23T17:39:54
{ "title": [ -0.017486572, -0.012832642, -0.033599854, -0.09375, 0.0012359619, 0.012466431, -0.068237305, -0.01121521, 0.0027065277, 0.0592041, 0.010650635, 0.021850586, 0.015029907, -0.013061523, -0.04534912, 0.04019165, 0.004497528, 0.034088135...
http://arxiv.org/abs/2305.14293v2
WebIE: Faithful and Robust Information Extraction on the Web
Chenxi Whitehouse, Clara Vania, Alham Fikri Aji, Christos Christodoulopoulos, Andrea Pierleoni
cs.CL
Extracting structured and grounded fact triples from raw text is a fundamental task in Information Extraction (IE). Existing IE datasets are typically collected from Wikipedia articles, using hyperlinks to link entities to the Wikidata knowledge base. However, models trained only on Wikipedia have limitations when appl...
http://arxiv.org/pdf/2305.14293v2
2023-06-15T13:51:36
2023-05-23T17:37:53
{ "title": [ -0.023956299, -0.012382507, -0.062347412, -0.07208252, 0.035339355, -0.006214142, -0.029571533, -0.0024986267, -0.00020885468, 0.027404785, -0.01864624, 0.022750854, 0.0022182465, -0.008522034, -0.03857422, 0.0085372925, -0.0102005005, ...
http://arxiv.org/abs/2305.14292v2
WikiChat: Stopping the Hallucination of Large Language Model Chatbots by Few-Shot Grounding on Wikipedia
Sina J. Semnani, Violet Z. Yao, Heidi C. Zhang, Monica S. Lam
cs.CL
This paper presents the first few-shot LLM-based chatbot that almost never hallucinates and has high conversationality and low latency. WikiChat is grounded on the English Wikipedia, the largest curated free-text corpus. WikiChat generates a response from an LLM, retains only the grounded facts, and combines them wit...
http://arxiv.org/pdf/2305.14292v2
2023-10-27T19:11:55
2023-05-23T17:37:36
{ "title": [ 0.005279541, -0.012428284, -0.0090408325, -0.01499176, -0.025741577, 0.0024909973, -0.00919342, -0.017944336, 0.0042686462, 0.04119873, -0.022155762, 0.0025749207, 0.037017822, -0.011993408, -0.040802002, -0.020568848, 0.0093688965, ...
http://arxiv.org/abs/2305.14291v2
Evaluation of African American Language Bias in Natural Language Generation
Nicholas Deas, Jessi Grieser, Shana Kleiner, Desmond Patton, Elsbeth Turcan, Kathleen McKeown
cs.CL
We evaluate how well LLMs understand African American Language (AAL) in comparison to their performance on White Mainstream English (WME), the encouraged "standard" form of English taught in American classrooms. We measure LLM performance using automatic metrics and human judgments for two tasks: a counterpart generati...
http://arxiv.org/pdf/2305.14291v2
2023-11-13T01:41:43
2023-05-23T17:34:37
{ "title": [ -0.028274536, -0.015022278, -0.03656006, -0.02746582, -0.03378296, -0.0050735474, -0.025665283, 0.009643555, -0.014541626, 0.031951904, -0.025009155, 0.006252289, 0.014709473, -0.029571533, -0.027679443, -0.005252838, -0.029251099, 0...
http://arxiv.org/abs/2305.14288v2
LLM-powered Data Augmentation for Enhanced Cross-lingual Performance
Chenxi Whitehouse, Monojit Choudhury, Alham Fikri Aji
cs.CL
This paper explores the potential of leveraging Large Language Models (LLMs) for data augmentation in multilingual commonsense reasoning datasets where the available training data is extremely limited. To achieve this, we utilise several LLMs, namely Dolly-v2, StableVicuna, ChatGPT, and GPT-4, to augment three datasets...
http://arxiv.org/pdf/2305.14288v2
2023-10-22T22:57:00
2023-05-23T17:33:27
{ "title": [ -0.03363037, -0.013427734, -0.06365967, -0.06185913, -0.016845703, 0.0013093948, -0.07080078, 0.02003479, -0.0009841919, 0.034606934, -0.013717651, 0.014633179, 0.0065956116, -0.02645874, -0.0126571655, 0.02204895, -0.012664795, 0.04...
http://arxiv.org/abs/2305.14283v3
Query Rewriting for Retrieval-Augmented Large Language Models
Xinbei Ma, Yeyun Gong, Pengcheng He, Hai Zhao, Nan Duan
cs.CL
Large Language Models (LLMs) play powerful, black-box readers in the retrieve-then-read pipeline, making remarkable progress in knowledge-intensive tasks. This work introduces a new framework, Rewrite-Retrieve-Read instead of the previous retrieve-then-read for the retrieval-augmented LLMs from the perspective of the q...
http://arxiv.org/pdf/2305.14283v3
2023-10-23T02:34:49
2023-05-23T17:27:50
{ "title": [ -0.020904541, -0.018325806, -0.02178955, -0.09802246, 0.0031719208, 0.034454346, -0.07348633, 0.013412476, 0.018188477, 0.04159546, -0.017410278, -0.01763916, 0.02218628, -0.010017395, -0.05444336, 0.005443573, -0.008018494, 0.002845...