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/2305.15328v2
Visual Programming for Text-to-Image Generation and Evaluation
Jaemin Cho, Abhay Zala, Mohit Bansal
cs.CV, cs.AI, cs.CL, cs.LG
As large language models have demonstrated impressive performance in many domains, recent works have adopted language models (LMs) as controllers of visual modules for vision-and-language tasks. While existing work focuses on equipping LMs with visual understanding, we propose two novel interpretable/explainable visual...
http://arxiv.org/pdf/2305.15328v2
2023-10-27T01:44:27
2023-05-24T16:42:17
{ "title": [ -0.02268982, -0.012519836, -0.018218994, -0.046722412, -0.008712769, 0.0076141357, -0.027511597, -0.020523071, 0.025024414, 0.034606934, -0.026870728, 0.008926392, -0.027038574, -0.012161255, -0.06933594, -0.010650635, 0.0030536652, ...
http://arxiv.org/abs/2305.15321v1
Towards Foundation Models for Relational Databases [Vision Paper]
Liane Vogel, Benjamin Hilprecht, Carsten Binnig
cs.DB, cs.CL
Tabular representation learning has recently gained a lot of attention. However, existing approaches only learn a representation from a single table, and thus ignore the potential to learn from the full structure of relational databases, including neighboring tables that can contain important information for a contextu...
http://arxiv.org/pdf/2305.15321v1
2023-05-24T16:37:35
2023-05-24T16:37:35
{ "title": [ -0.030441284, 0.0026798248, -0.066345215, -0.08972168, 0.029708862, 0.006526947, -0.042938232, 0.02305603, 0.014541626, 0.029220581, -0.028167725, -0.0049362183, -0.015716553, -0.009635925, -0.058746338, -0.0063972473, -0.011314392, ...
http://arxiv.org/abs/2305.15299v4
Science in the Era of ChatGPT, Large Language Models and Generative AI: Challenges for Research Ethics and How to Respond
Evangelos Pournaras
cs.CY, cs.AI, cs.CL
Large language models of artificial intelligence (AI), such as ChatGPT, find remarkable but controversial applicability in science and research. This paper reviews epistemological challenges, ethical and integrity risks in science conduct in the advent of generative AI. This is with the aim to lay new timely foundation...
http://arxiv.org/pdf/2305.15299v4
2023-07-29T12:54:20
2023-05-24T16:23:46
{ "title": [ -0.028839111, -0.019241333, -0.035125732, 0.007911682, -0.022857666, -0.02268982, 0.0031032562, 0.004966736, -0.011428833, 0.04095459, -0.015342712, 0.007843018, 0.00010025501, -0.014099121, -0.034362793, -0.0066947937, -0.013580322, ...
http://arxiv.org/abs/2305.15294v2
Enhancing Retrieval-Augmented Large Language Models with Iterative Retrieval-Generation Synergy
Zhihong Shao, Yeyun Gong, Yelong Shen, Minlie Huang, Nan Duan, Weizhu Chen
cs.CL
Large language models are powerful text processors and reasoners, but are still subject to limitations including outdated knowledge and hallucinations, which necessitates connecting them to the world. Retrieval-augmented large language models have raised extensive attention for grounding model generation on external kn...
http://arxiv.org/pdf/2305.15294v2
2023-10-23T09:58:13
2023-05-24T16:17:36
{ "title": [ -0.012161255, 0.0050621033, -0.03338623, -0.07092285, -0.0030345917, 0.013549805, -0.069885254, -0.0018672943, 0.0317688, 0.06640625, -0.012199402, 0.00674057, 0.018127441, -0.0029640198, -0.074279785, 0.029449463, -0.019210815, 0.02...
http://arxiv.org/abs/2305.15282v2
A Simple and Effective Framework for Strict Zero-Shot Hierarchical Classification
Rohan Bhambhoria, Lei Chen, Xiaodan Zhu
cs.CL, cs.AI
In recent years, large language models (LLMs) have achieved strong performance on benchmark tasks, especially in zero or few-shot settings. However, these benchmarks often do not adequately address the challenges posed in the real-world, such as that of hierarchical classification. In order to address this challenge, w...
http://arxiv.org/pdf/2305.15282v2
2023-05-26T13:57:52
2023-05-24T16:04:26
{ "title": [ 0.0129776, 0.009277344, -0.055480957, -0.081726074, 0.030410767, -0.013114929, -0.028030396, -0.0048980713, 0.0206604, 0.030029297, -0.011787415, 0.011550903, -0.024932861, -0.023223877, -0.06121826, 0.007709503, 0.0004887581, 0.0401...
http://arxiv.org/abs/2305.15275v1
Self-Evolution Learning for Discriminative Language Model Pretraining
Qihuang Zhong, Liang Ding, Juhua Liu, Bo Du, Dacheng Tao
cs.CL
Masked language modeling, widely used in discriminative language model (e.g., BERT) pretraining, commonly adopts a random masking strategy. However, random masking does not consider the importance of the different words in the sentence meaning, where some of them are more worthy to be predicted. Therefore, various mask...
http://arxiv.org/pdf/2305.15275v1
2023-05-24T16:00:54
2023-05-24T16:00:54
{ "title": [ -0.01902771, 0.0110321045, -0.033721924, -0.025650024, -0.0015296936, -0.01689148, -0.0231781, -0.009933472, -0.01802063, 0.034698486, 0.008346558, 0.011528015, -0.027282715, -0.03643799, -0.04107666, -0.0029907227, -0.0043945312, 0....
http://arxiv.org/abs/2305.15273v1
Revisiting Token Dropping Strategy in Efficient BERT Pretraining
Qihuang Zhong, Liang Ding, Juhua Liu, Xuebo Liu, Min Zhang, Bo Du, Dacheng Tao
cs.CL
Token dropping is a recently-proposed strategy to speed up the pretraining of masked language models, such as BERT, by skipping the computation of a subset of the input tokens at several middle layers. It can effectively reduce the training time without degrading much performance on downstream tasks. However, we empiri...
http://arxiv.org/pdf/2305.15273v1
2023-05-24T15:59:44
2023-05-24T15:59:44
{ "title": [ 0.017028809, -0.016525269, -0.032958984, -0.07080078, -0.015586853, -0.008636475, -0.033111572, 0.005054474, -0.026779175, 0.028305054, -0.0090789795, 0.013191223, 0.014907837, -0.02722168, -0.004837036, -0.0058670044, -0.012496948, ...
http://arxiv.org/abs/2305.15269v3
Testing the General Deductive Reasoning Capacity of Large Language Models Using OOD Examples
Abulhair Saparov, Richard Yuanzhe Pang, Vishakh Padmakumar, Nitish Joshi, Seyed Mehran Kazemi, Najoung Kim, He He
cs.CL, cs.AI
Given the intractably large size of the space of proofs, any model that is capable of general deductive reasoning must generalize to proofs of greater complexity. Recent studies have shown that large language models (LLMs) possess some abstract deductive reasoning ability given chain-of-thought prompts. However, they h...
http://arxiv.org/pdf/2305.15269v3
2023-11-03T18:45:56
2023-05-24T15:55:51
{ "title": [ -0.008552551, -0.0137786865, -0.023788452, -0.08355713, 0.013618469, 0.015792847, -0.05053711, -0.0016307831, 0.0059013367, 0.005432129, -0.041656494, -0.008903503, 0.003047943, -0.022323608, -0.029144287, -0.017669678, -0.010467529, ...
http://arxiv.org/abs/2305.15268v1
EvEval: A Comprehensive Evaluation of Event Semantics for Large Language Models
Zhengwei Tao, Zhi Jin, Xiaoying Bai, Haiyan Zhao, Yanlin Feng, Jia Li, Wenpeng Hu
cs.CL, cs.AI
Events serve as fundamental units of occurrence within various contexts. The processing of event semantics in textual information forms the basis of numerous natural language processing (NLP) applications. Recent studies have begun leveraging large language models (LLMs) to address event semantic processing. However, t...
http://arxiv.org/pdf/2305.15268v1
2023-05-24T15:55:40
2023-05-24T15:55:40
{ "title": [ -0.023712158, -0.025619507, -0.038909912, -0.11236572, 0.020935059, 0.01838684, -0.08251953, 0.0154953, 0.027008057, 0.02645874, 0.00043416023, -0.0012750626, 0.026657104, -0.023208618, -0.0390625, 0.03488159, -0.016143799, 0.0368347...
http://arxiv.org/abs/2305.15265v1
Winner-Take-All Column Row Sampling for Memory Efficient Adaptation of Language Model
Zirui Liu, Guanchu Wang, Shaochen Zhong, Zhaozhuo Xu, Daochen Zha, Ruixiang Tang, Zhimeng Jiang, Kaixiong Zhou, Vipin Chaudhary, Shuai Xu, Xia Hu
cs.LG, cs.CL
With the rapid growth in model size, fine-tuning the large pre-trained language model has become increasingly difficult due to its extensive memory usage. Previous works usually focus on reducing the number of trainable parameters in the network. While the model parameters do contribute to memory usage, the primary mem...
http://arxiv.org/pdf/2305.15265v1
2023-05-24T15:52:08
2023-05-24T15:52:08
{ "title": [ -0.03555298, -0.03515625, -0.017028809, -0.045410156, -0.038848877, 0.0021381378, -0.049041748, 0.018127441, 0.020477295, 0.059051514, -0.02079773, 0.01727295, 0.025634766, -0.00044226646, -0.06341553, 0.02482605, 0.0031776428, 0.000...
http://arxiv.org/abs/2305.15262v1
Revisiting Parallel Context Windows: A Frustratingly Simple Alternative and Chain-of-Thought Deterioration
Kejuan Yang, Xiao Liu, Kaiwen Men, Aohan Zeng, Yuxiao Dong, Jie Tang
cs.CL
We identify two crucial limitations in the evaluation of recent parallel-integrated method Parallel Context Windows (PCW), which extends the maximum context lengths of language models, e.g., 2048 for LLaMA, by harnessing window-wise attention and positional embedding techniques. We first show that a simple yet strong b...
http://arxiv.org/pdf/2305.15262v1
2023-05-24T15:48:29
2023-05-24T15:48:29
{ "title": [ -0.0027675629, -0.006225586, -0.022888184, -0.059783936, -0.009117126, -0.008399963, -0.03010559, -0.019470215, -0.0025253296, 0.056610107, -0.034606934, -0.012084961, 0.016845703, -0.01914978, -0.05670166, -0.004421234, 0.020889282, ...
http://arxiv.org/abs/2305.15255v3
Spoken Question Answering and Speech Continuation Using Spectrogram-Powered LLM
Eliya Nachmani, Alon Levkovitch, Roy Hirsch, Julian Salazar, Chulayuth Asawaroengchai, Soroosh Mariooryad, Ehud Rivlin, RJ Skerry-Ryan, Michelle Tadmor Ramanovich
cs.CL, cs.LG, cs.SD, eess.AS
We present a novel approach to adapting pre-trained large language models (LLMs) to perform question answering (QA) and speech continuation. By endowing the LLM with a pre-trained speech encoder, our model becomes able to take speech inputs and generate speech outputs. The entire system is trained end-to-end and operat...
http://arxiv.org/pdf/2305.15255v3
2023-10-20T05:55:39
2023-05-24T15:39:43
{ "title": [ -0.017425537, -0.03213501, -0.023162842, -0.0635376, -0.026977539, 0.013374329, -0.0713501, 0.023666382, 0.009864807, 0.0368042, -0.013687134, 0.010643005, -0.02178955, -0.01751709, -0.06237793, 0.008087158, -0.012336731, 0.042510986...
http://arxiv.org/abs/2305.15233v1
Boosting Cross-lingual Transferability in Multilingual Models via In-Context Learning
Sunkyoung Kim, Dayeon Ki, Yireun Kim, Jinsik Lee
cs.CL, cs.AI
Existing cross-lingual transfer (CLT) prompting methods are only concerned with monolingual demonstration examples in the source language. In this paper, we propose In-CLT, a novel cross-lingual transfer prompting method that leverages both source and target languages to construct the demonstration examples. We conduct...
http://arxiv.org/pdf/2305.15233v1
2023-05-24T15:14:49
2023-05-24T15:14:49
{ "title": [ -0.005634308, 0.0036773682, -0.02494812, -0.066101074, 0.01878357, 0.008132935, -0.048095703, -0.011634827, 0.01675415, 0.03466797, -0.0126571655, 0.019744873, 0.021514893, -0.018508911, -0.02255249, 0.017333984, -0.020889282, 0.0505...
http://arxiv.org/abs/2305.15225v2
SAIL: Search-Augmented Instruction Learning
Hongyin Luo, Yung-Sung Chuang, Yuan Gong, Tianhua Zhang, Yoon Kim, Xixin Wu, Danny Fox, Helen Meng, James Glass
cs.CL
Large language models (LLMs) have been significantly improved by instruction fine-tuning, but still lack transparency and the ability to utilize up-to-date knowledge and information. In this work, we propose search-augmented instruction learning (SAIL), which grounds the language generation and instruction following ab...
http://arxiv.org/pdf/2305.15225v2
2023-06-25T17:56:37
2023-05-24T15:07:30
{ "title": [ -0.019348145, -0.012893677, -0.027404785, -0.0003285408, -0.018051147, -0.024047852, 0.0033416748, -0.029174805, -0.024475098, 0.007259369, -0.034057617, 0.010002136, -0.024520874, -0.023651123, -0.0012178421, -0.020935059, -0.01965332, ...
http://arxiv.org/abs/2305.15222v1
Neural Summarization of Electronic Health Records
Koyena Pal, Seyed Ali Bahrainian, Laura Mercurio, Carsten Eickhoff
cs.CL, cs.AI, cs.IR
Hospital discharge documentation is among the most essential, yet time-consuming documents written by medical practitioners. The objective of this study was to automatically generate hospital discharge summaries using neural network summarization models. We studied various data preparation and neural network training t...
http://arxiv.org/pdf/2305.15222v1
2023-05-24T15:05:53
2023-05-24T15:05:53
{ "title": [ -0.0020961761, 0.009300232, -0.03503418, -0.081726074, 0.008850098, 0.0047187805, -0.0519104, 0.0006957054, -0.021438599, 0.062805176, -0.01525116, 0.017425537, 0.012786865, -0.010681152, -0.045166016, 0.008369446, 0.012023926, 0.020...
http://arxiv.org/abs/2305.15212v1
Towards Adaptive Prefix Tuning for Parameter-Efficient Language Model Fine-tuning
Zhen-Ru Zhang, Chuanqi Tan, Haiyang Xu, Chengyu Wang, Jun Huang, Songfang Huang
cs.CL
Fine-tuning large pre-trained language models on various downstream tasks with whole parameters is prohibitively expensive. Hence, Parameter-efficient fine-tuning has attracted attention that only optimizes a few task-specific parameters with the frozen pre-trained model. In this work, we focus on prefix tuning, which ...
http://arxiv.org/pdf/2305.15212v1
2023-05-24T14:51:01
2023-05-24T14:51:01
{ "title": [ -0.04864502, -0.023147583, -0.026046753, -0.07507324, -0.017242432, 0.0231781, -0.07684326, 0.01878357, -0.010498047, 0.038909912, 0.024398804, 0.0011825562, 0.007926941, -0.031829834, -0.043945312, 0.025283813, -0.021499634, 0.03152...
http://arxiv.org/abs/2305.15186v1
SciReviewGen: A Large-scale Dataset for Automatic Literature Review Generation
Tetsu Kasanishi, Masaru Isonuma, Junichiro Mori, Ichiro Sakata
cs.CL, cs.AI
Automatic literature review generation is one of the most challenging tasks in natural language processing. Although large language models have tackled literature review generation, the absence of large-scale datasets has been a stumbling block to the progress. We release SciReviewGen, consisting of over 10,000 literat...
http://arxiv.org/pdf/2305.15186v1
2023-05-24T14:26:30
2023-05-24T14:26:30
{ "title": [ -0.035583496, -0.008293152, -0.04626465, -0.04586792, 0.0062789917, 0.0065307617, -0.04449463, -0.028411865, -0.005218506, 0.029251099, -0.019470215, 0.025604248, -0.018478394, 0.004386902, -0.05331421, 0.006980896, -0.009216309, 0.0...
http://arxiv.org/abs/2305.15183v1
Are Pre-trained Language Models Useful for Model Ensemble in Chinese Grammatical Error Correction?
Chenming Tang, Xiuyu Wu, Yunfang Wu
cs.CL
Model ensemble has been in widespread use for Grammatical Error Correction (GEC), boosting model performance. We hypothesize that model ensemble based on the perplexity (PPL) computed by pre-trained language models (PLMs) should benefit the GEC system. To this end, we explore several ensemble strategies based on strong...
http://arxiv.org/pdf/2305.15183v1
2023-05-24T14:18:52
2023-05-24T14:18:52
{ "title": [ 0.006652832, 0.0064353943, -0.006801605, -0.023071289, -0.029342651, -0.004634857, -0.044433594, -0.01878357, 0.0016307831, 0.038330078, -0.031585693, -0.017974854, 0.038726807, -0.03250122, -0.03161621, -0.00894928, -0.028762817, 0....
http://arxiv.org/abs/2305.15182v2
HiTIN: Hierarchy-aware Tree Isomorphism Network for Hierarchical Text Classification
He Zhu, Chong Zhang, Junjie Huang, Junran Wu, Ke Xu
cs.CL
Hierarchical text classification (HTC) is a challenging subtask of multi-label classification as the labels form a complex hierarchical structure. Existing dual-encoder methods in HTC achieve weak performance gains with huge memory overheads and their structure encoders heavily rely on domain knowledge. Under such obse...
http://arxiv.org/pdf/2305.15182v2
2023-06-09T08:53:14
2023-05-24T14:14:08
{ "title": [ -0.01977539, 0.012039185, -0.04953003, -0.06323242, 0.009399414, 0.016189575, -0.041931152, -0.0039863586, -0.014518738, 0.059143066, -0.024124146, 0.03164673, -0.0045547485, 0.003238678, -0.06365967, -0.007461548, -0.02923584, 0.039...
http://arxiv.org/abs/2305.15175v1
Pre-training Multi-party Dialogue Models with Latent Discourse Inference
Yiyang Li, Xinting Huang, Wei Bi, Hai Zhao
cs.CL
Multi-party dialogues are more difficult for models to understand than one-to-one two-party dialogues, since they involve multiple interlocutors, resulting in interweaving reply-to relations and information flows. To step over these obstacles, an effective way is to pre-train a model that understands the discourse stru...
http://arxiv.org/pdf/2305.15175v1
2023-05-24T14:06:27
2023-05-24T14:06:27
{ "title": [ -0.0066604614, -0.0012493134, -0.04156494, -0.08050537, 0.005634308, -0.0075263977, -0.064941406, 0.012077332, -0.0044403076, 0.0138549805, 0.005680084, 0.008728027, -0.021331787, -0.02798462, -0.024108887, 0.0004734993, -0.019515991, ...
http://arxiv.org/abs/2305.15138v1
Topic-Guided Self-Introduction Generation for Social Media Users
Chunpu Xu, Jing Li, Piji Li, Min Yang
cs.CL, cs.AI, cs.LG
Millions of users are active on social media. To allow users to better showcase themselves and network with others, we explore the auto-generation of social media self-introduction, a short sentence outlining a user's personal interests. While most prior work profiles users with tags (e.g., ages), we investigate senten...
http://arxiv.org/pdf/2305.15138v1
2023-05-24T13:35:08
2023-05-24T13:35:08
{ "title": [ -0.007030487, 0.01977539, -0.031341553, -0.026992798, 0.0020751953, -0.0031719208, -0.018005371, -0.029937744, -0.035064697, 0.043640137, -0.011459351, 0.060699463, -0.029342651, -0.009292603, -0.020111084, 0.002122879, 0.0021133423, ...
http://arxiv.org/abs/2305.15130v1
On Degrees of Freedom in Defining and Testing Natural Language Understanding
Saku Sugawara, Shun Tsugita
cs.CL, cs.AI
Natural language understanding (NLU) studies often exaggerate or underestimate the capabilities of systems, thereby limiting the reproducibility of their findings. These erroneous evaluations can be attributed to the difficulty of defining and testing NLU adequately. In this position paper, we reconsider this challenge...
http://arxiv.org/pdf/2305.15130v1
2023-05-24T13:25:20
2023-05-24T13:25:20
{ "title": [ -0.030776978, -0.0064735413, -0.04232788, -0.05432129, 0.0033798218, -0.0040283203, -0.030700684, -0.0055351257, -0.010871887, -0.0009431839, -0.039154053, 0.0036640167, -0.026550293, -0.0104599, -0.04425049, -0.023513794, -0.019104004, ...
http://arxiv.org/abs/2305.15119v1
Another Dead End for Morphological Tags? Perturbed Inputs and Parsing
Alberto Muñoz-Ortiz, David Vilares
cs.CL
The usefulness of part-of-speech tags for parsing has been heavily questioned due to the success of word-contextualized parsers. Yet, most studies are limited to coarse-grained tags and high quality written content; while we know little about their influence when it comes to models in production that face lexical error...
http://arxiv.org/pdf/2305.15119v1
2023-05-24T13:11:04
2023-05-24T13:11:04
{ "title": [ 0.040008545, -0.0006556511, 0.02053833, -0.037109375, -0.030059814, -0.0020961761, -0.022232056, -0.045288086, 0.0005598068, 0.013145447, -0.038238525, 0.01687622, -0.0129470825, -0.036010742, 0.0051460266, -0.02494812, -0.01965332, ...
http://arxiv.org/abs/2305.15108v1
The Role of Output Vocabulary in T2T LMs for SPARQL Semantic Parsing
Debayan Banerjee, Pranav Ajit Nair, Ricardo Usbeck, Chris Biemann
cs.CL
In this work, we analyse the role of output vocabulary for text-to-text (T2T) models on the task of SPARQL semantic parsing. We perform experiments within the the context of knowledge graph question answering (KGQA), where the task is to convert questions in natural language to the SPARQL query language. We observe tha...
http://arxiv.org/pdf/2305.15108v1
2023-05-24T12:55:04
2023-05-24T12:55:04
{ "title": [ -0.01789856, 0.0044136047, -0.048431396, -0.071777344, -0.005367279, 0.0047416687, -0.07244873, 0.0030536652, -0.022384644, 0.029647827, -0.02986145, 0.008277893, -0.016677856, -0.017166138, -0.031311035, -0.008934021, -0.0020942688, ...
http://arxiv.org/abs/2305.15099v1
Fourier Transformer: Fast Long Range Modeling by Removing Sequence Redundancy with FFT Operator
Ziwei He, Meng Yang, Minwei Feng, Jingcheng Yin, Xinbing Wang, Jingwen Leng, Zhouhan Lin
cs.CL
The transformer model is known to be computationally demanding, and prohibitively costly for long sequences, as the self-attention module uses a quadratic time and space complexity with respect to sequence length. Many researchers have focused on designing new forms of self-attention or introducing new parameters to ov...
http://arxiv.org/pdf/2305.15099v1
2023-05-24T12:33:06
2023-05-24T12:33:06
{ "title": [ 0.012741089, 0.047912598, -0.036590576, -0.032196045, 0.027832031, -0.030776978, -0.025344849, -0.05480957, 0.00415802, 0.049072266, 0.0024299622, -0.028564453, 0.062042236, -0.052093506, -0.019332886, 0.004245758, 0.011260986, 0.026...
http://arxiv.org/abs/2305.15098v1
Referral Augmentation for Zero-Shot Information Retrieval
Michael Tang, Shunyu Yao, John Yang, Karthik Narasimhan
cs.CL
We propose Referral-Augmented Retrieval (RAR), a simple technique that concatenates document indices with referrals, i.e. text from other documents that cite or link to the given document, to provide significant performance gains for zero-shot information retrieval. The key insight behind our method is that referrals p...
http://arxiv.org/pdf/2305.15098v1
2023-05-24T12:28:35
2023-05-24T12:28:35
{ "title": [ 0.04751587, -0.005935669, -0.024093628, -0.05239868, -0.0010185242, -0.025115967, 0.016830444, -0.0065727234, 0.01878357, 0.04949951, -0.021575928, 0.0019664764, 0.014198303, -0.011276245, -0.05206299, 0.017700195, -0.0025997162, 0.0...
http://arxiv.org/abs/2305.15096v2
Dynamic Masking Rate Schedules for MLM Pretraining
Zachary Ankner, Naomi Saphra, Davis Blalock, Jonathan Frankle, Matthew L. Leavitt
cs.CL, cs.AI
Most works on transformers trained with the Masked Language Modeling (MLM) objective use the original BERT model's fixed masking rate of 15%. We propose to instead dynamically schedule the masking rate throughout training. We find that linearly decreasing the masking rate over the course of pretraining improves average...
http://arxiv.org/pdf/2305.15096v2
2023-09-14T21:26:14
2023-05-24T12:24:12
{ "title": [ 0.004699707, 0.019805908, -0.07330322, -0.035614014, -0.008407593, -0.02633667, -0.0317688, -0.0019340515, -0.07366943, 0.03579712, -0.0015420914, 0.029968262, 0.009284973, -0.044677734, 0.05593872, -0.0070343018, 0.02407837, 0.03692...
http://arxiv.org/abs/2305.15093v2
C-STS: Conditional Semantic Textual Similarity
Ameet Deshpande, Carlos E. Jimenez, Howard Chen, Vishvak Murahari, Victoria Graf, Tanmay Rajpurohit, Ashwin Kalyan, Danqi Chen, Karthik Narasimhan
cs.CL, cs.AI, cs.LG
Semantic textual similarity (STS), a cornerstone task in NLP, measures the degree of similarity between a pair of sentences, and has broad application in fields such as information retrieval and natural language understanding. However, sentence similarity can be inherently ambiguous, depending on the specific aspect of...
http://arxiv.org/pdf/2305.15093v2
2023-11-06T18:48:43
2023-05-24T12:18:50
{ "title": [ -0.036346436, 0.002155304, -0.041107178, -0.061706543, 0.019210815, 0.011627197, -0.04135132, -0.02532959, 0.015777588, 0.03466797, -0.030776978, 0.030395508, 0.02079773, -0.0019168854, -0.03652954, 0.01108551, -0.0056152344, 0.02502...
http://arxiv.org/abs/2305.15090v2
STAR: Improving Low-Resource Information Extraction by Structure-to-Text Data Generation with Large Language Models
Mingyu Derek Ma, Xiaoxuan Wang, Po-Nien Kung, P. Jeffrey Brantingham, Nanyun Peng, Wei Wang
cs.CL, cs.AI
Information extraction tasks such as event extraction require an in-depth understanding of the output structure and sub-task dependencies. They heavily rely on task-specific training data in the form of (passage, target structure) pairs to obtain reasonable performance. However, obtaining such data through human annota...
http://arxiv.org/pdf/2305.15090v2
2023-09-30T08:37:28
2023-05-24T12:15:19
{ "title": [ -0.04598999, -0.015571594, -0.049926758, -0.07421875, -0.010032654, 0.0178833, -0.0657959, 0.0039749146, 0.008888245, 0.02154541, -0.021636963, 0.017501831, -0.006263733, -0.005722046, -0.040130615, 0.0048675537, -0.013702393, 0.0367...
http://arxiv.org/abs/2306.05535v1
Detecting Check-Worthy Claims in Political Debates, Speeches, and Interviews Using Audio Data
Petar Ivanov, Ivan Koychev, Momchil Hardalov, Preslav Nakov
cs.CL, cs.AI, cs.IR, cs.LG, cs.SD, eess.AS, 68T50, F.2.2; I.2.7
A large portion of society united around the same vision and ideas carries enormous energy. That is precisely what political figures would like to accumulate for their cause. With this goal in mind, they can sometimes resort to distorting or hiding the truth, unintentionally or on purpose, which opens the door for misi...
http://arxiv.org/pdf/2306.05535v1
2023-05-24T12:09:42
2023-05-24T12:09:42
{ "title": [ 0.049316406, 0.0062026978, -0.03253174, -0.080322266, 0.02520752, -0.033355713, -0.023452759, -0.0071792603, -0.055511475, 0.019424438, -0.012107849, 0.015296936, -0.01184082, -0.019836426, -0.024734497, -0.030639648, -0.03010559, 0....
http://arxiv.org/abs/2305.15087v1
Pento-DIARef: A Diagnostic Dataset for Learning the Incremental Algorithm for Referring Expression Generation from Examples
Philipp Sadler, David Schlangen
cs.CL, cs.CV
NLP tasks are typically defined extensionally through datasets containing example instantiations (e.g., pairs of image i and text t), but motivated intensionally through capabilities invoked in verbal descriptions of the task (e.g., "t is a description of i, for which the content of i needs to be recognised and underst...
http://arxiv.org/pdf/2305.15087v1
2023-05-24T12:05:53
2023-05-24T12:05:53
{ "title": [ 0.013671875, -0.004940033, -0.031829834, -0.070007324, -0.0035591125, -0.00187397, -0.054748535, 0.0017719269, 0.016448975, 0.017242432, -0.016067505, 0.002746582, 0.043121338, -0.040222168, -0.0039100647, 0.013809204, -0.015350342, ...
http://arxiv.org/abs/2305.15083v2
Eliciting the Translation Ability of Large Language Models via Multilingual Finetuning with Translation Instructions
Jiahuan Li, Hao Zhou, Shujian Huang, Shanbo Cheng, Jiajun Chen
cs.CL
Large-scale Pretrained Language Models (LLMs), such as ChatGPT and GPT4, have shown strong abilities in multilingual translations, without being explicitly trained on parallel corpora. It is interesting how the LLMs obtain their ability to carry out translation instructions for different languages. In this paper, we pr...
http://arxiv.org/pdf/2305.15083v2
2023-06-30T02:32:11
2023-05-24T12:00:24
{ "title": [ -0.035186768, -0.016403198, -0.03201294, -0.051116943, -0.009887695, -0.00011539459, -0.056610107, -0.0035114288, -0.0047035217, 0.04660034, -0.008255005, 0.023513794, 0.0030460358, -0.013076782, -0.058166504, 0.0132751465, -0.0012617111...
http://arxiv.org/abs/2305.15080v2
Visually-Situated Natural Language Understanding with Contrastive Reading Model and Frozen Large Language Models
Geewook Kim, Hodong Lee, Daehee Kim, Haeji Jung, Sanghee Park, Yoonsik Kim, Sangdoo Yun, Taeho Kil, Bado Lee, Seunghyun Park
cs.CL, cs.AI
Recent advances in Large Language Models (LLMs) have stimulated a surge of research aimed at extending their applications to the visual domain. While these models exhibit promise in generating abstract image captions and facilitating natural conversations, their performance on text-rich images still requires improvemen...
http://arxiv.org/pdf/2305.15080v2
2023-10-26T12:51:07
2023-05-24T11:59:13
{ "title": [ -0.013671875, -0.009651184, -0.008956909, -0.0519104, -0.030654907, 0.0049095154, -0.03677368, -0.0104904175, 0.030166626, 0.021392822, -0.044128418, -0.0032958984, -0.010726929, -0.0047073364, -0.061584473, -0.01361084, -0.015464783, ...
http://arxiv.org/abs/2305.15077v2
Contrastive Learning of Sentence Embeddings from Scratch
Junlei Zhang, Zhenzhong Lan, Junxian He
cs.CL
Contrastive learning has been the dominant approach to train state-of-the-art sentence embeddings. Previous studies have typically learned sentence embeddings either through the use of human-annotated natural language inference (NLI) data or via large-scale unlabeled sentences in an unsupervised manner. However, even i...
http://arxiv.org/pdf/2305.15077v2
2023-10-24T09:56:46
2023-05-24T11:56:21
{ "title": [ 0.008300781, 0.00039887428, -0.007835388, -0.048828125, 0.002708435, -0.00409317, -0.013496399, -0.015670776, 0.02619934, 0.018234253, -0.029663086, 0.019744873, -0.009017944, -0.0082473755, -0.0423584, 0.0015201569, -0.015625, 0.065...
http://arxiv.org/abs/2305.15076v2
Meta-Learning Online Adaptation of Language Models
Nathan Hu, Eric Mitchell, Christopher D. Manning, Chelsea Finn
cs.CL
Large language models encode impressively broad world knowledge in their parameters. However, the knowledge in static language models falls out of date, limiting the model's effective "shelf life." While online fine-tuning can reduce this degradation, we find that naively fine-tuning on a stream of documents leads to a...
http://arxiv.org/pdf/2305.15076v2
2023-10-20T22:49:24
2023-05-24T11:56:20
{ "title": [ -0.035247803, -0.018875122, -0.022659302, -0.049560547, -0.014953613, 0.014564514, -0.06335449, -0.0062713623, 0.012077332, 0.048675537, -0.0020256042, -0.00818634, 0.00617218, -0.01676941, -0.05319214, 0.019989014, 0.0033302307, 0.0...
http://arxiv.org/abs/2305.15075v1
HuatuoGPT, towards Taming Language Model to Be a Doctor
Hongbo Zhang, Junying Chen, Feng Jiang, Fei Yu, Zhihong Chen, Jianquan Li, Guiming Chen, Xiangbo Wu, Zhiyi Zhang, Qingying Xiao, Xiang Wan, Benyou Wang, Haizhou Li
cs.CL, cs.AI
In this paper, we present HuatuoGPT, a large language model (LLM) for medical consultation. The core recipe of HuatuoGPT is to leverage both \textit{distilled data from ChatGPT} and \textit{real-world data from doctors} in the supervised fine-tuned stage. The responses of ChatGPT are usually detailed, well-presented an...
http://arxiv.org/pdf/2305.15075v1
2023-05-24T11:56:01
2023-05-24T11:56:01
{ "title": [ 0.016540527, -0.0013380051, -0.03616333, 0.023727417, -0.02885437, -0.030319214, 0.0071029663, -0.03567505, -0.00598526, 0.033355713, 0.010108948, -0.03060913, 0.06750488, -0.06402588, 0.003610611, -0.000213027, -0.005718231, 0.02528...
http://arxiv.org/abs/2305.15074v3
Have LLMs Advanced Enough? A Challenging Problem Solving Benchmark For Large Language Models
Daman Arora, Himanshu Gaurav Singh, Mausam
cs.CL, cs.AI
The performance of large language models (LLMs) on existing reasoning benchmarks has significantly improved over the past years. In response, we present JEEBench, a considerably more challenging benchmark dataset for evaluating the problem solving abilities of LLMs. We curate 515 challenging pre-engineering mathematics...
http://arxiv.org/pdf/2305.15074v3
2023-10-23T11:55:58
2023-05-24T11:55:59
{ "title": [ -0.04736328, -0.020324707, -0.03253174, -0.053253174, -0.021743774, 0.0026683807, -0.07116699, 0.001832962, 0.007106781, 0.016326904, -0.020004272, 0.014022827, -0.01777649, -0.016494751, -0.034179688, 0.0040283203, -0.01776123, 0.02...
http://arxiv.org/abs/2305.15070v3
Annotation Imputation to Individualize Predictions: Initial Studies on Distribution Dynamics and Model Predictions
London Lowmanstone, Ruyuan Wan, Risako Owan, Jaehyung Kim, Dongyeop Kang
cs.CL
Annotating data via crowdsourcing is time-consuming and expensive. Due to these costs, dataset creators often have each annotator label only a small subset of the data. This leads to sparse datasets with examples that are marked by few annotators. The downside of this process is that if an annotator doesn't get to labe...
http://arxiv.org/pdf/2305.15070v3
2023-10-05T07:10:25
2023-05-24T11:54:46
{ "title": [ -0.012580872, -0.012611389, -0.011810303, -0.06933594, -0.020980835, 0.011558533, -0.055480957, -0.010894775, -0.014785767, 0.04244995, 0.019210815, 0.02532959, -0.020309448, -0.034820557, -0.056030273, 0.018051147, -0.00026774406, 0...
http://arxiv.org/abs/2305.15068v2
ToMChallenges: A Principle-Guided Dataset and Diverse Evaluation Tasks for Exploring Theory of Mind
Xiaomeng Ma, Lingyu Gao, Qihui Xu
cs.CL, cs.AI
Theory of Mind (ToM), the capacity to comprehend the mental states of distinct individuals, is essential for numerous practical applications. With the development of large language models (LLMs), there is a heated debate about whether they are able to perform ToM tasks. Previous studies have used different tasks and pr...
http://arxiv.org/pdf/2305.15068v2
2023-10-23T22:16:15
2023-05-24T11:54:07
{ "title": [ -0.0012073517, -0.013633728, -0.023773193, -0.019134521, 0.0109939575, -0.024429321, 0.010375977, -0.024139404, -0.02281189, 0.024963379, 0.015327454, 0.0056610107, -0.0062408447, -0.033569336, -0.044067383, 0.0049934387, -0.012359619, ...
http://arxiv.org/abs/2305.15067v1
Not All Metrics Are Guilty: Improving NLG Evaluation with LLM Paraphrasing
Tianyi Tang, Hongyuan Lu, Yuchen Eleanor Jiang, Haoyang Huang, Dongdong Zhang, Wayne Xin Zhao, Furu Wei
cs.CL
Most research about natural language generation (NLG) relies on evaluation benchmarks with limited references for a sample, which may result in poor correlations with human judgements. The underlying reason is that one semantic meaning can actually be expressed in different forms, and the evaluation with a single or fe...
http://arxiv.org/pdf/2305.15067v1
2023-05-24T11:53:29
2023-05-24T11:53:29
{ "title": [ -0.040405273, -0.025238037, -0.024642944, -0.017868042, -0.04232788, -0.0021743774, -0.042663574, -0.018005371, -0.03564453, 0.023223877, -0.041809082, 0.031158447, -0.013725281, -0.0052337646, -0.029266357, -0.02029419, -0.023284912, ...
http://arxiv.org/abs/2305.15066v2
GPT4Graph: Can Large Language Models Understand Graph Structured Data ? An Empirical Evaluation and Benchmarking
Jiayan Guo, Lun Du, Hengyu Liu, Mengyu Zhou, Xinyi He, Shi Han
cs.AI, cs.CL
Large language models~(LLM) like ChatGPT have become indispensable to artificial general intelligence~(AGI), demonstrating excellent performance in various natural language processing tasks. In the real world, graph data is ubiquitous and an essential part of AGI and prevails in domains like social network analysis, bi...
http://arxiv.org/pdf/2305.15066v2
2023-07-11T15:08:00
2023-05-24T11:53:19
{ "title": [ -0.04385376, -0.017028809, -0.04901123, -0.06762695, -0.0096206665, 0.008163452, -0.07647705, 0.006072998, 0.0135650635, 0.0053100586, -0.0154418945, -0.014305115, -0.014556885, -0.03213501, -0.034729004, -0.0015640259, -0.015838623, ...
http://arxiv.org/abs/2305.15065v1
Inference-Time Policy Adapters (IPA): Tailoring Extreme-Scale LMs without Fine-tuning
Ximing Lu, Faeze Brahman, Peter West, Jaehun Jang, Khyathi Chandu, Abhilasha Ravichander, Lianhui Qin, Prithviraj Ammanabrolu, Liwei Jiang, Sahana Ramnath, Nouha Dziri, Jillian Fisher, Bill Yuchen Lin, Skyler Hallinan, Xiang Ren, Sean Welleck, Yejin Choi
cs.CL
Large language models excel at a variety of language tasks when prompted with examples or instructions. Yet controlling these models through prompting alone is limited. Tailoring language models through fine-tuning (e.g., via reinforcement learning) can be effective, but it is expensive and requires model access. We ...
http://arxiv.org/pdf/2305.15065v1
2023-05-24T11:52:55
2023-05-24T11:52:55
{ "title": [ -0.014831543, -0.0021953583, -0.048217773, -0.05935669, 0.0013866425, -0.020690918, -0.010231018, -0.0021858215, -0.021987915, 0.0067634583, -0.050933838, 0.013832092, -0.025115967, -0.023468018, -0.0029258728, -0.026992798, 0.0074386597...
http://arxiv.org/abs/2305.15064v3
AutoPlan: Automatic Planning of Interactive Decision-Making Tasks With Large Language Models
Siqi Ouyang, Lei Li
cs.CL
Recent large language models (LLMs) are promising for making decisions in grounded environments. However, LLMs frequently fail in complex decision-making tasks due to the misalignment between the pre-trained knowledge in LLMs and the actual rules in the environment. Existing methods require either costly gradient compu...
http://arxiv.org/pdf/2305.15064v3
2023-10-26T16:44:39
2023-05-24T11:52:23
{ "title": [ -0.02798462, -0.009902954, -0.0345459, -0.059295654, 0.008003235, 0.008255005, -0.049072266, -0.015655518, -0.01197052, 0.015388489, -0.025115967, 0.009689331, -0.014808655, -0.017547607, -0.0066070557, -0.014556885, -0.0013742447, 0...
http://arxiv.org/abs/2305.15062v2
Lawyer LLaMA Technical Report
Quzhe Huang, Mingxu Tao, Chen Zhang, Zhenwei An, Cong Jiang, Zhibin Chen, Zirui Wu, Yansong Feng
cs.CL, cs.AI
Large Language Models (LLMs), like LLaMA, have exhibited remarkable performance across various tasks. Nevertheless, when deployed to specific domains such as law or medicine, the models still confront the challenge of a deficiency in domain-specific knowledge and an inadequate capability to leverage that knowledge to r...
http://arxiv.org/pdf/2305.15062v2
2023-10-14T02:14:51
2023-05-24T11:52:07
{ "title": [ -0.024520874, -0.01953125, -0.031463623, -0.0057144165, -0.051361084, -0.010108948, -0.05267334, 0.0074386597, -0.046325684, 0.05355835, -0.009178162, 0.00016629696, 0.045562744, -0.031677246, 0.048309326, -0.0065574646, -0.0007414818, ...
http://arxiv.org/abs/2305.15060v1
Who Wrote this Code? Watermarking for Code Generation
Taehyun Lee, Seokhee Hong, Jaewoo Ahn, Ilgee Hong, Hwaran Lee, Sangdoo Yun, Jamin Shin, Gunhee Kim
cs.CL
Large language models for code have recently shown remarkable performance in generating executable code. However, this rapid advancement has been accompanied by many legal and ethical concerns, such as code licensing issues, code plagiarism, and malware generation, making watermarking machine-generated code a very time...
http://arxiv.org/pdf/2305.15060v1
2023-05-24T11:49:52
2023-05-24T11:49:52
{ "title": [ -0.023880005, -0.0078125, -0.055145264, -0.044830322, -0.018615723, -0.018157959, -0.055603027, -0.000875473, -0.029632568, 0.0064353943, -0.008590698, 0.017807007, -0.022888184, -0.042816162, 0.013130188, -0.024398804, 0.0053863525, ...
http://arxiv.org/abs/2305.15057v1
Learning the String Partial Order
Tianyu Liu, Afra Amini, Mrinmaya Sachan, Ryan Cotterell
cs.CL
We show that most structured prediction problems can be solved in linear time and space by considering them as partial orderings of the tokens in the input string. Our method computes real numbers for each token in an input string and sorts the tokens accordingly, resulting in as few as 2 total orders of the tokens in ...
http://arxiv.org/pdf/2305.15057v1
2023-05-24T11:47:35
2023-05-24T11:47:35
{ "title": [ 0.015365601, -0.0135269165, -0.03213501, -0.09863281, -0.025466919, -0.00831604, -0.049468994, 0.04864502, -0.008407593, 0.04626465, -0.007286072, -0.0211792, 0.016174316, -0.042755127, -0.029571533, -0.0024909973, 0.0087890625, 0.02...
http://arxiv.org/abs/2305.15056v1
Reasoning over Hierarchical Question Decomposition Tree for Explainable Question Answering
Jiajie Zhang, Shulin Cao, Tingjia Zhang, Xin Lv, Jiaxin Shi, Qi Tian, Juanzi Li, Lei Hou
cs.CL
Explainable question answering (XQA) aims to answer a given question and provide an explanation why the answer is selected. Existing XQA methods focus on reasoning on a single knowledge source, e.g., structured knowledge bases, unstructured corpora, etc. However, integrating information from heterogeneous knowledge sou...
http://arxiv.org/pdf/2305.15056v1
2023-05-24T11:45:59
2023-05-24T11:45:59
{ "title": [ -0.0131073, -0.008605957, -0.01423645, -0.103393555, -0.0035533905, 0.03894043, -0.050689697, 0.021316528, 0.018051147, 0.013206482, -0.023513794, 0.0025806427, -0.0011892319, -0.00793457, -0.0418396, -0.008483887, -0.0054244995, 0.0...
http://arxiv.org/abs/2305.15054v2
A Mechanistic Interpretation of Arithmetic Reasoning in Language Models using Causal Mediation Analysis
Alessandro Stolfo, Yonatan Belinkov, Mrinmaya Sachan
cs.CL, cs.LG
Mathematical reasoning in large language models (LMs) has garnered significant attention in recent work, but there is a limited understanding of how these models process and store information related to arithmetic tasks within their architecture. In order to improve our understanding of this aspect of language models, ...
http://arxiv.org/pdf/2305.15054v2
2023-10-20T12:13:27
2023-05-24T11:43:47
{ "title": [ -0.021621704, -0.019851685, -0.040924072, -0.06100464, -0.027359009, -0.014877319, -0.06378174, 0.020935059, -0.005306244, 0.02658081, 0.015838623, -0.0048065186, 0.033477783, -0.0501709, -0.010940552, 0.019851685, -0.02822876, 0.045...
http://arxiv.org/abs/2305.15053v1
Decomposing Complex Queries for Tip-of-the-tongue Retrieval
Kevin Lin, Kyle Lo, Joseph E. Gonzalez, Dan Klein
cs.CL, cs.IR
When re-finding items, users who forget or are uncertain about identifying details often rely on creative strategies for expressing their information needs -- complex queries that describe content elements (e.g., book characters or events), information beyond the document text (e.g., descriptions of book covers), or pe...
http://arxiv.org/pdf/2305.15053v1
2023-05-24T11:43:40
2023-05-24T11:43:40
{ "title": [ 0.023620605, -0.0002706051, -0.005252838, -0.06222534, 0.024398804, 0.00793457, -0.004940033, -0.008773804, 0.02017212, 0.047851562, -0.015464783, -0.004840851, 0.017913818, -0.009407043, -0.060546875, 0.016174316, 0.016983032, 0.002...
http://arxiv.org/abs/2305.15051v1
A Monte Carlo Language Model Pipeline for Zero-Shot Sociopolitical Event Extraction
Erica Cai, Brendan O'Connor
cs.CL
We consider dyadic zero-shot event extraction (EE) to identify actions between pairs of actors. The \emph{zero-shot} setting allows social scientists or other non-computational researchers to extract any customized, user-specified set of events without training, resulting in a \emph{dyadic} event database, allowing ins...
http://arxiv.org/pdf/2305.15051v1
2023-05-24T11:41:33
2023-05-24T11:41:33
{ "title": [ 0.009284973, -0.005695343, -0.026992798, -0.09106445, 0.00040125847, -0.001405716, -0.0602417, 0.0070991516, 0.012306213, 0.01473999, 0.008575439, 0.011161804, 0.00472641, -0.020553589, -0.034362793, 0.017990112, -0.03314209, 0.02699...
http://arxiv.org/abs/2305.15048v1
Ranger: A Toolkit for Effect-Size Based Multi-Task Evaluation
Mete Sertkan, Sophia Althammer, Sebastian Hofstätter
cs.CL, cs.IR
In this paper, we introduce Ranger - a toolkit to facilitate the easy use of effect-size-based meta-analysis for multi-task evaluation in NLP and IR. We observed that our communities often face the challenge of aggregating results over incomparable metrics and scenarios, which makes conclusions and take-away messages l...
http://arxiv.org/pdf/2305.15048v1
2023-05-24T11:38:39
2023-05-24T11:38:39
{ "title": [ -0.015853882, -0.016494751, -0.05657959, -0.035308838, -0.0041656494, -0.026947021, -0.03201294, -0.019470215, -0.0289917, 0.031707764, 0.009101868, 0.013626099, -0.027435303, -0.028396606, -0.038085938, -0.010231018, -0.017166138, 0...
http://arxiv.org/abs/2305.15047v2
Ghostbuster: Detecting Text Ghostwritten by Large Language Models
Vivek Verma, Eve Fleisig, Nicholas Tomlin, Dan Klein
cs.CL, cs.AI
We introduce Ghostbuster, a state-of-the-art system for detecting AI-generated text. Our method works by passing documents through a series of weaker language models, running a structured search over possible combinations of their features, and then training a classifier on the selected features to predict whether docu...
http://arxiv.org/pdf/2305.15047v2
2023-11-13T18:43:11
2023-05-24T11:37:10
{ "title": [ -0.015914917, 0.006828308, -0.014083862, -0.033111572, 0.002111435, 0.0011940002, -0.03842163, -0.039642334, 0.00818634, 0.037963867, -0.026535034, -0.005580902, 0.008010864, 0.0035324097, -0.0647583, -0.00061130524, -0.03414917, 0.0...
http://arxiv.org/abs/2305.15045v1
SETI: Systematicity Evaluation of Textual Inference
Xiyan Fu, Anette Frank
cs.CL
We propose SETI (Systematicity Evaluation of Textual Inference), a novel and comprehensive benchmark designed for evaluating pre-trained language models (PLMs) for their systematicity capabilities in the domain of textual inference. Specifically, SETI offers three different NLI tasks and corresponding datasets to evalu...
http://arxiv.org/pdf/2305.15045v1
2023-05-24T11:35:31
2023-05-24T11:35:31
{ "title": [ -0.04660034, -0.022827148, -0.041168213, -0.070495605, -0.007827759, 0.011917114, -0.06359863, -0.0074157715, 0.022506714, 0.028457642, -0.024765015, 0.0031795502, 0.019012451, -0.013664246, -0.048828125, 0.020248413, -0.013198853, 0...
http://arxiv.org/abs/2305.15044v1
Is Summary Useful or Not? An Extrinsic Human Evaluation of Text Summaries on Downstream Tasks
Xiao Pu, Mingqi Gao, Xiaojun Wan
cs.CL
Research on automated text summarization relies heavily on human and automatic evaluation. While recent work on human evaluation mainly adopted intrinsic evaluation methods, judging the generic quality of text summaries, e.g. informativeness and coherence, our work focuses on evaluating the usefulness of text summaries...
http://arxiv.org/pdf/2305.15044v1
2023-05-24T11:34:39
2023-05-24T11:34:39
{ "title": [ -0.028030396, 0.00075149536, -0.046295166, -0.033813477, -0.010124207, -0.023345947, -0.040222168, -0.020141602, 0.0036849976, 0.053588867, -0.029388428, 0.02456665, 0.0044517517, 0.021499634, -0.078430176, -0.0037212372, -0.014877319, ...
http://arxiv.org/abs/2305.15041v1
Generating Faithful Synthetic Data with Large Language Models: A Case Study in Computational Social Science
Veniamin Veselovsky, Manoel Horta Ribeiro, Akhil Arora, Martin Josifoski, Ashton Anderson, Robert West
cs.CL
Large Language Models (LLMs) have democratized synthetic data generation, which in turn has the potential to simplify and broaden a wide gamut of NLP tasks. Here, we tackle a pervasive problem in synthetic data generation: its generative distribution often differs from the distribution of real-world data researchers ca...
http://arxiv.org/pdf/2305.15041v1
2023-05-24T11:27:59
2023-05-24T11:27:59
{ "title": [ -0.030807495, -0.0021438599, -0.03857422, -0.068847656, 0.0011663437, 0.013877869, -0.06939697, -0.0099487305, -0.011276245, 0.029312134, -0.01625061, 0.0073661804, -0.01826477, -0.0075035095, -0.039794922, -0.0042915344, -0.027923584, ...
http://arxiv.org/abs/2305.15040v2
Active Learning for Natural Language Generation
Yotam Perlitz, Ariel Gera, Michal Shmueli-Scheuer, Dafna Sheinwald, Noam Slonim, Liat Ein-Dor
cs.CL
The field of Natural Language Generation (NLG) suffers from a severe shortage of labeled data due to the extremely expensive and time-consuming process involved in manual annotation. A natural approach for coping with this problem is active learning (AL), a well-known machine learning technique for improving annotation...
http://arxiv.org/pdf/2305.15040v2
2023-10-17T14:37:10
2023-05-24T11:27:53
{ "title": [ -0.03366089, -0.016143799, -0.01939392, -0.061187744, -0.0005350113, 0.016693115, -0.037963867, 0.0034389496, 0.0033798218, 0.03439331, -0.007785797, 0.009483337, -0.025619507, -0.021377563, -0.035461426, 0.010803223, -0.007865906, 0...
http://arxiv.org/abs/2305.15038v2
Is GPT-4 a Good Data Analyst?
Liying Cheng, Xingxuan Li, Lidong Bing
cs.CL
As large language models (LLMs) have demonstrated their powerful capabilities in plenty of domains and tasks, including context understanding, code generation, language generation, data storytelling, etc., many data analysts may raise concerns if their jobs will be replaced by artificial intelligence (AI). This controv...
http://arxiv.org/pdf/2305.15038v2
2023-10-23T02:10:58
2023-05-24T11:26:59
{ "title": [ 0.0053863525, 0.010032654, -0.022750854, -0.010063171, -0.01486969, -0.012481689, -0.012367249, -0.020202637, -0.009780884, 0.021347046, 0.0126571655, -0.04058838, -0.052734375, -0.053344727, -0.032318115, -0.038238525, 0.00018811226, ...
http://arxiv.org/abs/2305.15035v2
Self-ICL: Zero-Shot In-Context Learning with Self-Generated Demonstrations
Wei-Lin Chen, Cheng-Kuang Wu, Yun-Nung Chen, Hsin-Hsi Chen
cs.CL
Large language models (LLMs) have exhibited striking in-context learning (ICL) ability to adapt to target tasks with a few input-output demonstrations. For better ICL, different methods are proposed to select representative demonstrations from existing training corpora. However, such settings are not aligned with real-...
http://arxiv.org/pdf/2305.15035v2
2023-10-23T14:50:57
2023-05-24T11:22:34
{ "title": [ 0.017181396, 0.0026378632, -0.032073975, -0.07171631, 0.016342163, -0.010597229, -0.008369446, 0.00045657158, 0.004825592, 0.02420044, -0.0158844, 0.012878418, -0.038604736, -0.020584106, -0.03314209, -0.00541687, -0.00094127655, 0.0...
http://arxiv.org/abs/2305.15033v1
SmartTrim: Adaptive Tokens and Parameters Pruning for Efficient Vision-Language Models
Zekun Wang, Jingchang Chen, Wangchunshu Zhou, Ming Liu, Bing Qin
cs.CL
Despite achieving remarkable performance on various vision-language tasks, Transformer-based pretrained vision-language models (VLMs) still suffer from efficiency issues arising from long inputs and numerous parameters, limiting their real-world applications. However, the huge computation is redundant for most samples ...
http://arxiv.org/pdf/2305.15033v1
2023-05-24T11:18:00
2023-05-24T11:18:00
{ "title": [ -0.031402588, -0.019958496, -0.033294678, -0.06359863, -0.0056114197, 0.0075263977, -0.060516357, 0.0019702911, 0.038757324, 0.009460449, -0.012008667, -0.0068130493, 0.0109939575, -0.021469116, -0.038604736, 0.0046691895, 0.0143585205, ...
http://arxiv.org/abs/2305.15032v1
How to Distill your BERT: An Empirical Study on the Impact of Weight Initialisation and Distillation Objectives
Xinpeng Wang, Leonie Weissweiler, Hinrich Schütze, Barbara Plank
cs.CL, cs.AI, cs.LG
Recently, various intermediate layer distillation (ILD) objectives have been shown to improve compression of BERT models via Knowledge Distillation (KD). However, a comprehensive evaluation of the objectives in both task-specific and task-agnostic settings is lacking. To the best of our knowledge, this is the first wor...
http://arxiv.org/pdf/2305.15032v1
2023-05-24T11:16:09
2023-05-24T11:16:09
{ "title": [ 0.00038695335, -0.015472412, -0.060546875, -0.07489014, -0.02319336, -0.01651001, -0.07067871, 0.0024089813, -0.018859863, 0.02178955, -0.013946533, -0.0044555664, -0.007751465, -0.040924072, -0.010978699, -0.022888184, -0.006454468, ...
http://arxiv.org/abs/2305.15028v2
ImageNetVC: Zero- and Few-Shot Visual Commonsense Evaluation on 1000 ImageNet Categories
Heming Xia, Qingxiu Dong, Lei Li, Jingjing Xu, Tianyu Liu, Ziwei Qin, Zhifang Sui
cs.CL, cs.AI, cs.CV
Recently, Large Language Models (LLMs) have been serving as general-purpose interfaces, posing a significant demand for comprehensive visual knowledge. However, it remains unclear how well current LLMs and their visually augmented counterparts (VaLMs) can master visual commonsense knowledge. To investigate this, we pro...
http://arxiv.org/pdf/2305.15028v2
2023-10-20T20:41:46
2023-05-24T11:14:31
{ "title": [ 0.02128601, -0.0070533752, -0.021362305, -0.03186035, 0.0070495605, -0.02229309, -0.0031147003, -0.03189087, 0.04348755, 0.014305115, -0.032196045, 0.0056152344, -0.02217102, -0.01777649, -0.052978516, 0.010147095, -0.007774353, 0.04...
http://arxiv.org/abs/2305.15025v2
Dior-CVAE: Pre-trained Language Models and Diffusion Priors for Variational Dialog Generation
Tianyu Yang, Thy Thy Tran, Iryna Gurevych
cs.CL
Current variational dialog models have employed pre-trained language models (PLMs) to parameterize the likelihood and posterior distributions. However, the Gaussian assumption made on the prior distribution is incompatible with these distributions, thus restricting the diversity of generated responses. These models als...
http://arxiv.org/pdf/2305.15025v2
2023-10-23T18:51:10
2023-05-24T11:06:52
{ "title": [ -0.020355225, -0.016403198, -0.054992676, -0.071777344, -0.0036468506, -0.0031661987, -0.064697266, 0.01512146, 0.016738892, 0.022369385, 0.012145996, 0.0059928894, -0.021896362, -0.031204224, -0.04559326, 0.010971069, -0.018051147, ...
http://arxiv.org/abs/2305.15024v1
ChatAgri: Exploring Potentials of ChatGPT on Cross-linguistic Agricultural Text Classification
Biao Zhao, Weiqiang Jin, Javier Del Ser, Guang Yang
cs.CL, cs.AI
In the era of sustainable smart agriculture, a massive amount of agricultural news text is being posted on the Internet, in which massive agricultural knowledge has been accumulated. In this context, it is urgent to explore effective text classification techniques for users to access the required agricultural knowledge...
http://arxiv.org/pdf/2305.15024v1
2023-05-24T11:06:23
2023-05-24T11:06:23
{ "title": [ -0.018722534, -0.008483887, -0.0024414062, 0.014755249, -0.014877319, 0.0071144104, -0.015007019, -0.026306152, -0.011672974, 0.03756714, -0.032409668, 0.022781372, -0.008125305, -0.005748749, -0.033599854, -0.010429382, -0.053497314, ...
http://arxiv.org/abs/2305.15020v3
An Efficient Multilingual Language Model Compression through Vocabulary Trimming
Asahi Ushio, Yi Zhou, Jose Camacho-Collados
cs.CL
Multilingual language model (LM) have become a powerful tool in NLP especially for non-English languages. Nevertheless, model parameters of multilingual LMs remain large due to the larger embedding matrix of the vocabulary covering tokens in different languages. On the contrary, monolingual LMs can be trained in a targ...
http://arxiv.org/pdf/2305.15020v3
2023-10-19T10:30:02
2023-05-24T11:00:33
{ "title": [ -0.06573486, -0.017318726, -0.038116455, -0.053710938, -0.007949829, 0.03338623, -0.07989502, 0.015129089, 0.0047721863, 0.043182373, 0.019699097, 0.00033593178, 0.009567261, -0.026977539, -0.04458618, 0.021850586, 0.0032405853, 0.01...
http://arxiv.org/abs/2305.15017v2
Calc-X and Calcformers: Empowering Arithmetical Chain-of-Thought through Interaction with Symbolic Systems
Marek Kadlčík, Michal Štefánik, Ondřej Sotolář, Vlastimil Martinek
cs.LG, cs.AI, cs.CL
Despite outstanding performance in many tasks, language models are notoriously inclined to make factual errors in tasks requiring arithmetic computation. We address this deficiency by creating Calc-X, a collection of datasets that demonstrates the appropriate use of a calculator in reasoning chains. Calc-X is suitable ...
http://arxiv.org/pdf/2305.15017v2
2023-10-23T15:23:22
2023-05-24T10:58:20
{ "title": [ 0.017242432, -0.018463135, -0.03894043, -0.04055786, -0.01386261, -0.03543091, -0.0016012192, -0.0101623535, 0.00907135, 0.031234741, -0.03427124, -0.0064811707, 0.011940002, -0.02583313, -0.029464722, 0.0014734268, 0.0015630722, 0.0...
http://arxiv.org/abs/2305.15014v1
Unlocking Temporal Question Answering for Large Language Models Using Code Execution
Xingxuan Li, Liying Cheng, Qingyu Tan, Hwee Tou Ng, Shafiq Joty, Lidong Bing
cs.CL
Large language models (LLMs) have made significant progress in natural language processing (NLP), and are utilized extensively in various applications. Recent works, such as chain-of-thought (CoT), have shown that intermediate reasoning steps can improve the performance of LLMs for complex reasoning tasks, such as math...
http://arxiv.org/pdf/2305.15014v1
2023-05-24T10:57:53
2023-05-24T10:57:53
{ "title": [ -0.0011882782, -0.029388428, -0.028015137, -0.082336426, -0.006286621, -0.008834839, -0.0418396, 0.039611816, 0.023635864, 0.025131226, -0.026321411, -0.01171875, 0.0022296906, -0.0082092285, -0.0657959, -0.00014007092, 0.0007033348, ...
http://arxiv.org/abs/2305.15011v2
Bactrian-X: Multilingual Replicable Instruction-Following Models with Low-Rank Adaptation
Haonan Li, Fajri Koto, Minghao Wu, Alham Fikri Aji, Timothy Baldwin
cs.CL
Instruction tuning has shown great promise in improving the performance of large language models. However, research on multilingual instruction tuning has been limited due to the scarcity of high-quality instruction-response datasets across different languages. To bridge this gap, we present Bactrian-X, a comprehensive...
http://arxiv.org/pdf/2305.15011v2
2023-10-10T07:46:44
2023-05-24T10:50:31
{ "title": [ -0.05215454, -0.015205383, -0.011459351, -0.051361084, -0.032562256, 0.026016235, -0.100097656, -0.023040771, -0.0013790131, 0.034301758, -0.018493652, 0.011390686, -0.010604858, -0.0077667236, -0.025619507, 0.0065689087, -0.010978699, ...
http://arxiv.org/abs/2305.15010v1
Injecting Knowledge into Biomedical Pre-trained Models via Polymorphism and Synonymous Substitution
Hongbo Zhang, Xiang Wan, Benyou Wang
cs.CL
Pre-trained language models (PLMs) were considered to be able to store relational knowledge present in the training data. However, some relational knowledge seems to be discarded unsafely in PLMs due to \textbf{report bias}: low-frequency relational knowledge might be underexpressed compared to high-frequency one in PL...
http://arxiv.org/pdf/2305.15010v1
2023-05-24T10:48:53
2023-05-24T10:48:53
{ "title": [ 0.0031471252, -0.022720337, -0.014274597, -0.090026855, -0.012771606, 0.011138916, -0.06677246, -0.012893677, -0.013130188, 0.012649536, 0.0047073364, 0.0028247833, -0.023651123, -0.040618896, -0.0062026978, -0.014717102, 0.0053253174, ...
http://arxiv.org/abs/2305.15008v1
Are Chatbots Ready for Privacy-Sensitive Applications? An Investigation into Input Regurgitation and Prompt-Induced Sanitization
Aman Priyanshu, Supriti Vijay, Ayush Kumar, Rakshit Naidu, Fatemehsadat Mireshghallah
cs.CL, cs.AI, cs.CY
LLM-powered chatbots are becoming widely adopted in applications such as healthcare, personal assistants, industry hiring decisions, etc. In many of these cases, chatbots are fed sensitive, personal information in their prompts, as samples for in-context learning, retrieved records from a database, or as part of the co...
http://arxiv.org/pdf/2305.15008v1
2023-05-24T10:48:05
2023-05-24T10:48:05
{ "title": [ 0.010978699, -0.00015079975, -0.022644043, -0.050445557, 0.005962372, -0.0072669983, -0.023620605, -0.010505676, -0.01576233, 0.044921875, -0.015136719, 0.016311646, 0.0035190582, -0.008163452, -0.048706055, -0.00756073, -0.01096344, ...
http://arxiv.org/abs/2305.15005v1
Sentiment Analysis in the Era of Large Language Models: A Reality Check
Wenxuan Zhang, Yue Deng, Bing Liu, Sinno Jialin Pan, Lidong Bing
cs.CL
Sentiment analysis (SA) has been a long-standing research area in natural language processing. It can offer rich insights into human sentiments and opinions and has thus seen considerable interest from both academia and industry. With the advent of large language models (LLMs) such as ChatGPT, there is a great potentia...
http://arxiv.org/pdf/2305.15005v1
2023-05-24T10:45:25
2023-05-24T10:45:25
{ "title": [ -0.0068244934, -0.014480591, -0.015914917, -0.043304443, 0.0053482056, -0.0019893646, -0.017593384, -0.01600647, 0.021835327, 0.033813477, -0.02861023, -0.014717102, -0.0022449493, -0.0069274902, -0.0770874, -0.00056505203, -0.006477356,...
http://arxiv.org/abs/2305.15004v3
LLMDet: A Third Party Large Language Models Generated Text Detection Tool
Kangxi Wu, Liang Pang, Huawei Shen, Xueqi Cheng, Tat-Seng Chua
cs.CL
Generated texts from large language models (LLMs) are remarkably close to high-quality human-authored text, raising concerns about their potential misuse in spreading false information and academic misconduct. Consequently, there is an urgent need for a highly practical detection tool capable of accurately identifying ...
http://arxiv.org/pdf/2305.15004v3
2023-11-03T14:31:09
2023-05-24T10:45:16
{ "title": [ -0.025177002, -0.001701355, -0.02734375, -0.035461426, 0.0055618286, -0.0065345764, -0.035125732, -0.023971558, -0.0042800903, 0.034484863, -0.022033691, 0.00894165, 0.0032672882, -0.003929138, -0.04348755, -0.00005722046, -0.023544312, ...
http://arxiv.org/abs/2305.15002v1
A RelEntLess Benchmark for Modelling Graded Relations between Named Entities
Asahi Ushio, Jose Camacho Collados, Steven Schockaert
cs.CL, cs.LG
Relations such as "is influenced by", "is known for" or "is a competitor of" are inherently graded: we can rank entity pairs based on how well they satisfy these relations, but it is hard to draw a line between those pairs that satisfy them and those that do not. Such graded relations play a central role in many applic...
http://arxiv.org/pdf/2305.15002v1
2023-05-24T10:41:24
2023-05-24T10:41:24
{ "title": [ -0.006717682, -0.007911682, -0.039520264, -0.11053467, 0.015625, 0.0053863525, -0.06021118, 0.009262085, 0.029037476, 0.01626587, -0.014724731, 0.0037384033, 0.007774353, -0.0129776, -0.040618896, 0.016143799, -0.01309967, 0.04904174...
http://arxiv.org/abs/2305.14999v2
The Art of SOCRATIC QUESTIONING: Recursive Thinking with Large Language Models
Jingyuan Qi, Zhiyang Xu, Ying Shen, Minqian Liu, Di Jin, Qifan Wang, Lifu Huang
cs.CL
Chain-of-Thought (CoT) prompting enables large language models to solve complex reasoning problems by generating intermediate steps. However, confined by its inherent single-pass and sequential generation process, CoT heavily relies on the initial decisions, causing errors in early steps to accumulate and impact the fi...
http://arxiv.org/pdf/2305.14999v2
2023-11-02T02:40:48
2023-05-24T10:36:14
{ "title": [ -0.021881104, -0.017318726, -0.020599365, -0.04058838, -0.0005145073, -0.005748749, -0.030166626, -0.01953125, -0.006729126, -0.0048179626, -0.0076522827, -0.017318726, -0.03439331, -0.041931152, -0.037597656, -0.021316528, -0.017242432,...
http://arxiv.org/abs/2305.14998v1
An Examination of the Robustness of Reference-Free Image Captioning Evaluation Metrics
Saba Ahmadi, Aishwarya Agrawal
cs.CL, cs.AI, cs.CV, cs.LG
Recently, reference-free metrics such as CLIPScore (Hessel et al., 2021) and UMIC (Lee et al., 2021) have been proposed for automatic evaluation of image captions, demonstrating a high correlation with human judgment. In this work, our focus lies in evaluating the robustness of these metrics in scenarios that require d...
http://arxiv.org/pdf/2305.14998v1
2023-05-24T10:36:12
2023-05-24T10:36:12
{ "title": [ 0.023620605, 0.010894775, -0.00687027, -0.010375977, 0.0056495667, -0.0284729, 0.024932861, -0.04095459, -0.0012626648, 0.06439209, -0.019622803, 0.016616821, 0.013389587, 0.0002655983, -0.060180664, 0.0053367615, -0.008720398, 0.044...
http://arxiv.org/abs/2305.16344v1
Leveraging LLMs for KPIs Retrieval from Hybrid Long-Document: A Comprehensive Framework and Dataset
Chongjian Yue, Xinrun Xu, Xiaojun Ma, Lun Du, Hengyu Liu, Zhiming Ding, Yanbing Jiang, Shi Han, Dongmei Zhang
cs.CL, cs.AI
Large Language Models (LLMs) demonstrate exceptional performance in textual understanding and tabular reasoning tasks. However, their ability to comprehend and analyze hybrid text, containing textual and tabular data, remains underexplored. In this research, we specialize in harnessing the potential of LLMs to comprehe...
http://arxiv.org/pdf/2305.16344v1
2023-05-24T10:35:58
2023-05-24T10:35:58
{ "title": [ -0.0390625, -0.0107040405, -0.031707764, -0.008888245, -0.025146484, -0.012756348, -0.023376465, -0.017303467, -0.008239746, 0.0116119385, -0.060028076, -0.0010375977, -0.008094788, -0.020309448, -0.03375244, -0.055023193, -0.015945435, ...
http://arxiv.org/abs/2305.14996v2
The ACL OCL Corpus: Advancing Open Science in Computational Linguistics
Shaurya Rohatgi, Yanxia Qin, Benjamin Aw, Niranjana Unnithan, Min-Yen Kan
cs.CL, cs.DL
We present ACL OCL, a scholarly corpus derived from the ACL Anthology to assist Open scientific research in the Computational Linguistics domain. Integrating and enhancing the previous versions of the ACL Anthology, the ACL OCL contributes metadata, PDF files, citation graphs and additional structured full texts with s...
http://arxiv.org/pdf/2305.14996v2
2023-10-24T05:18:32
2023-05-24T10:35:56
{ "title": [ -0.031280518, -0.006072998, -0.027297974, -0.08148193, 0.009788513, 0.026824951, -0.06756592, -0.004634857, -0.0010738373, 0.03387451, 0.0048294067, 0.006290436, 0.009529114, -0.024032593, -0.035614014, 0.015037537, -0.016464233, 0.0...
http://arxiv.org/abs/2305.14994v3
RefGPT: Dialogue Generation of GPT, by GPT, and for GPT
Dongjie Yang, Ruifeng Yuan, Yuantao Fan, Yifei Yang, Zili Wang, Shusen Wang, Hai Zhao
cs.CL
Large Language Models (LLMs) have attained the impressive capability to resolve a wide range of NLP tasks by fine-tuning high-quality instruction data. However, collecting human-written data of high quality, especially multi-turn dialogues, is expensive and unattainable for most people. Though previous studies have use...
http://arxiv.org/pdf/2305.14994v3
2023-10-19T00:45:51
2023-05-24T10:30:42
{ "title": [ -0.028717041, -0.022201538, -0.03744507, -0.01878357, -0.037475586, -0.0047721863, -0.064819336, -0.014335632, -0.0068244934, 0.0030212402, -0.025924683, -0.009796143, 0.0024337769, -0.056762695, 0.0049362183, -0.006942749, -0.025436401,...
http://arxiv.org/abs/2305.14993v1
How To Control Text Simplification? An Empirical Study of Control Tokens for Meaning Preserving Controlled Simplification
Sweta Agrawal, Marine Carpuat
cs.CL
Text simplification rewrites text to be more readable for a specific audience, while preserving its meaning. However, determining what makes a text easy to read depends on who are the intended readers. Recent work has introduced a wealth of techniques to control output simplicity, ranging from specifying the desired re...
http://arxiv.org/pdf/2305.14993v1
2023-05-24T10:29:45
2023-05-24T10:29:45
{ "title": [ -0.025772095, -0.017669678, -0.046539307, -0.046081543, -0.0060424805, -0.0052604675, -0.039733887, -0.012397766, -0.019958496, 0.049682617, -0.005710602, 0.039276123, 0.016799927, -0.010131836, -0.038330078, 0.0044555664, -0.009407043, ...
http://arxiv.org/abs/2305.14992v2
Reasoning with Language Model is Planning with World Model
Shibo Hao, Yi Gu, Haodi Ma, Joshua Jiahua Hong, Zhen Wang, Daisy Zhe Wang, Zhiting Hu
cs.CL, cs.AI, cs.LG
Large language models (LLMs) have shown remarkable reasoning capabilities, especially when prompted to generate intermediate reasoning steps (e.g., Chain-of-Thought, CoT). However, LLMs can still struggle with problems that are easy for humans, such as generating action plans for executing tasks in a given environment,...
http://arxiv.org/pdf/2305.14992v2
2023-10-23T07:24:28
2023-05-24T10:28:28
{ "title": [ -0.009597778, -0.021621704, -0.014122009, -0.066711426, -0.07128906, 0.019989014, -0.08135986, 0.016860962, -0.06817627, 0.01838684, -0.02861023, -0.015930176, 0.031097412, -0.05697632, 0.05630493, -0.028793335, 0.015556335, 0.021713...
http://arxiv.org/abs/2305.14991v1
MuLER: Detailed and Scalable Reference-based Evaluation
Taelin Karidi, Leshem Choshen, Gal Patel, Omri Abend
cs.CL, cs.AI
We propose a novel methodology (namely, MuLER) that transforms any reference-based evaluation metric for text generation, such as machine translation (MT) into a fine-grained analysis tool. Given a system and a metric, MuLER quantifies how much the chosen metric penalizes specific error types (e.g., errors in transla...
http://arxiv.org/pdf/2305.14991v1
2023-05-24T10:26:13
2023-05-24T10:26:13
{ "title": [ -0.031951904, -0.011924744, -0.05053711, -0.07458496, 0.024780273, 0.017089844, -0.048797607, -0.02017212, -0.009567261, 0.025665283, 0.003921509, 0.0002644062, -0.004322052, -0.040802002, -0.0362854, 0.016525269, -0.02067566, 0.0560...
http://arxiv.org/abs/2306.04657v1
Improving Empathetic Dialogue Generation by Dynamically Infusing Commonsense Knowledge
Hua Cai, Xuli Shen, Qing Xu, Weilin Shen, Xiaomei Wang, Weifeng Ge, Xiaoqing Zheng, Xiangyang Xue
cs.CL, cs.AI
In empathetic conversations, individuals express their empathy towards others. Previous work has mainly focused on generating empathetic responses by utilizing the speaker's emotion. Besides, external commonsense knowledge has been applied to enhance the system's understandings of the speaker's situation. However, give...
http://arxiv.org/pdf/2306.04657v1
2023-05-24T10:25:12
2023-05-24T10:25:12
{ "title": [ 0.00071048737, -0.032440186, -0.039611816, -0.054107666, -0.012290955, -0.017044067, -0.01876831, 0.017593384, 0.016601562, 0.012794495, -0.003522873, -0.0017871857, -0.009857178, -0.035217285, -0.037384033, -0.006198883, -0.013725281, ...
http://arxiv.org/abs/2305.14989v2
Dolphin: A Challenging and Diverse Benchmark for Arabic NLG
El Moatez Billah Nagoudi, AbdelRahim Elmadany, Ahmed El-Shangiti, Muhammad Abdul-Mageed
cs.CL
We present Dolphin, a novel benchmark that addresses the need for a natural language generation (NLG) evaluation framework dedicated to the wide collection of Arabic languages and varieties. The proposed benchmark encompasses a broad range of 13 different NLG tasks, including dialogue generation, question answering, ma...
http://arxiv.org/pdf/2305.14989v2
2023-10-24T17:48:43
2023-05-24T10:24:10
{ "title": [ -0.030212402, -0.030822754, -0.027908325, -0.06695557, -0.0051460266, 0.00055933, -0.0395813, 0.021728516, 0.006084442, 0.023483276, -0.01461792, 0.012588501, -0.022903442, -0.012565613, -0.05557251, 0.0035533905, -0.00096178055, 0.0...
http://arxiv.org/abs/2305.14987v2
Investigating Table-to-Text Generation Capabilities of LLMs in Real-World Information Seeking Scenarios
Yilun Zhao, Haowei Zhang, Shengyun Si, Linyong Nan, Xiangru Tang, Arman Cohan
cs.CL
Tabular data is prevalent across various industries, necessitating significant time and effort for users to understand and manipulate for their information-seeking purposes. The advancements in large language models (LLMs) have shown enormous potential to improve user efficiency. However, the adoption of LLMs in real-w...
http://arxiv.org/pdf/2305.14987v2
2023-10-30T22:00:25
2023-05-24T10:22:30
{ "title": [ -0.04171753, -0.013542175, -0.025497437, 0.0017271042, -0.040802002, -0.0031757355, -0.036590576, -0.028030396, -0.022644043, 0.04547119, -0.03881836, 0.046295166, -0.012138367, -0.0031032562, -0.03778076, -0.019363403, -0.006576538, ...
http://arxiv.org/abs/2305.14985v1
IdealGPT: Iteratively Decomposing Vision and Language Reasoning via Large Language Models
Haoxuan You, Rui Sun, Zhecan Wang, Long Chen, Gengyu Wang, Hammad A. Ayyubi, Kai-Wei Chang, Shih-Fu Chang
cs.CV, cs.CL
The field of vision-and-language (VL) understanding has made unprecedented progress with end-to-end large pre-trained VL models (VLMs). However, they still fall short in zero-shot reasoning tasks that require multi-step inferencing. To achieve this goal, previous works resort to a divide-and-conquer pipeline. In this p...
http://arxiv.org/pdf/2305.14985v1
2023-05-24T10:19:57
2023-05-24T10:19:57
{ "title": [ -0.018737793, -0.029434204, -0.03314209, -0.06640625, -0.0050849915, 0.0010528564, -0.048950195, 0.008384705, 0.01876831, 0.014328003, -0.020767212, -0.0063934326, 0.010917664, -0.011856079, -0.031951904, 0.0038909912, 0.00869751, 0....
http://arxiv.org/abs/2305.14982v1
Benchmarking Arabic AI with Large Language Models
Ahmed Abdelali, Hamdy Mubarak, Shammur Absar Chowdhury, Maram Hasanain, Basel Mousi, Sabri Boughorbel, Yassine El Kheir, Daniel Izham, Fahim Dalvi, Majd Hawasly, Nizi Nazar, Yousseif Elshahawy, Ahmed Ali, Nadir Durrani, Natasa Milic-Frayling, Firoj Alam
cs.CL, cs.AI, 68T50, F.2.2; I.2.7
With large Foundation Models (FMs), language technologies (AI in general) are entering a new paradigm: eliminating the need for developing large-scale task-specific datasets and supporting a variety of tasks through set-ups ranging from zero-shot to few-shot learning. However, understanding FMs capabilities requires a ...
http://arxiv.org/pdf/2305.14982v1
2023-05-24T10:16:16
2023-05-24T10:16:16
{ "title": [ -0.04598999, -0.02192688, -0.01335144, -0.051696777, -0.014160156, 0.008682251, -0.046539307, 0.0030708313, -0.009864807, 0.023971558, -0.00062799454, -0.004306793, -0.016799927, -0.0284729, -0.04611206, 0.0018358231, 0.008117676, 0....
http://arxiv.org/abs/2305.14981v1
Improving Factuality of Abstractive Summarization without Sacrificing Summary Quality
Tanay Dixit, Fei Wang, Muhao Chen
cs.CL, cs.AI, cs.LG
Improving factual consistency of abstractive summarization has been a widely studied topic. However, most of the prior works on training factuality-aware models have ignored the negative effect it has on summary quality. We propose EFACTSUM (i.e., Effective Factual Summarization), a candidate summary generation and ran...
http://arxiv.org/pdf/2305.14981v1
2023-05-24T10:15:17
2023-05-24T10:15:17
{ "title": [ -0.0015506744, 0.011764526, -0.039031982, -0.06402588, -0.016220093, -0.011894226, -0.05456543, -0.027801514, 0.0040130615, 0.041381836, -0.020385742, 0.027114868, -0.008369446, 0.0036735535, -0.04949951, 0.009796143, -0.014297485, 0...
http://arxiv.org/abs/2305.14976v2
GPTAraEval: A Comprehensive Evaluation of ChatGPT on Arabic NLP
Md Tawkat Islam Khondaker, Abdul Waheed, El Moatez Billah Nagoudi, Muhammad Abdul-Mageed
cs.CL, cs.LG
ChatGPT's emergence heralds a transformative phase in NLP, particularly demonstrated through its excellent performance on many English benchmarks. However, the model's efficacy across diverse linguistic contexts remains largely uncharted territory. This work aims to bridge this knowledge gap, with a primary focus on as...
http://arxiv.org/pdf/2305.14976v2
2023-10-21T05:16:24
2023-05-24T10:12:39
{ "title": [ -0.017807007, -0.0262146, -0.016204834, -0.04309082, -0.0118637085, 0.0116119385, -0.026306152, -0.0031089783, -0.005847931, 0.007499695, -0.0043754578, 0.012542725, -0.037261963, -0.038391113, -0.01966858, 0.007106781, -0.016647339, ...
http://arxiv.org/abs/2305.14975v2
Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedback
Katherine Tian, Eric Mitchell, Allan Zhou, Archit Sharma, Rafael Rafailov, Huaxiu Yao, Chelsea Finn, Christopher D. Manning
cs.CL
A trustworthy real-world prediction system should produce well-calibrated confidence scores; that is, its confidence in an answer should be indicative of the likelihood that the answer is correct, enabling deferral to an expert in cases of low-confidence predictions. Recent studies have shown that unsupervised pre-trai...
http://arxiv.org/pdf/2305.14975v2
2023-10-24T04:27:42
2023-05-24T10:12:33
{ "title": [ -0.047027588, 0.011245728, -0.03475952, -0.01586914, -0.018798828, -0.0053901672, -0.06982422, -0.05267334, 0.008415222, 0.023117065, -0.05819702, 0.015853882, -0.037139893, -0.0067443848, -0.061767578, -0.020721436, -0.0016374588, 0...
http://arxiv.org/abs/2305.14973v1
OverPrompt: Enhancing ChatGPT Capabilities through an Efficient In-Context Learning Approach
Jiazheng Li, Runcong Zhao, Yulan He, Lin Gui
cs.CL
The exceptional performance of pre-trained large language models has revolutionised various applications, but their adoption in production environments is hindered by prohibitive costs and inefficiencies, particularly when utilising long prompts. This paper proposes OverPrompt, an in-context learning method aimed at im...
http://arxiv.org/pdf/2305.14973v1
2023-05-24T10:08:04
2023-05-24T10:08:04
{ "title": [ 0.012763977, -0.016983032, -0.018539429, -0.04019165, 0.013465881, -0.004688263, -0.00021398067, -0.007965088, -0.0030117035, 0.018005371, -0.04296875, 0.030532837, -0.0032920837, -0.0132369995, -0.0065345764, 0.00015902519, -0.016799927...
http://arxiv.org/abs/2305.16343v1
A Distributed Automatic Domain-Specific Multi-Word Term Recognition Architecture using Spark Ecosystem
Ciprian-Octavian Truică, Neculai-Ovidiu Istrate, Elena-Simona Apostol
cs.CL, cs.AI
Automatic Term Recognition is used to extract domain-specific terms that belong to a given domain. In order to be accurate, these corpus and language-dependent methods require large volumes of textual data that need to be processed to extract candidate terms that are afterward scored according to a given metric. To imp...
http://arxiv.org/pdf/2305.16343v1
2023-05-24T10:05:59
2023-05-24T10:05:59
{ "title": [ -0.024490356, -0.004447937, -0.06451416, -0.06964111, -0.0000030994415, 0.006046295, -0.05657959, 0.02571106, -0.009712219, 0.06774902, -0.038269043, 0.012580872, 0.019958496, 0.019348145, -0.06945801, 0.0011453629, -0.013549805, 0.0...
http://arxiv.org/abs/2305.14970v1
Getting Sick After Seeing a Doctor? Diagnosing and Mitigating Knowledge Conflicts in Event Temporal Reasoning
Tianqing Fang, Zhaowei Wang, Wenxuan Zhou, Hongming Zhang, Yangqiu Song, Muhao Chen
cs.CL, cs.AI
Event temporal reasoning aims at identifying the temporal relations between two or more events. However, knowledge conflicts arise when there is a mismatch between the actual temporal relations of events in the context and the prior knowledge or biases learned by the model. We first systematically define distinct kinds...
http://arxiv.org/pdf/2305.14970v1
2023-05-24T10:04:06
2023-05-24T10:04:06
{ "title": [ 0.0657959, 0.000108122826, -0.03829956, -0.1262207, 0.037200928, -0.029403687, -0.024017334, 0.0012540817, -0.0054244995, 0.018096924, -0.020706177, 0.010940552, -0.010284424, -0.020706177, -0.027130127, -0.02696228, 0.032989502, -0....
http://arxiv.org/abs/2305.14965v1
Tricking LLMs into Disobedience: Understanding, Analyzing, and Preventing Jailbreaks
Abhinav Rao, Sachin Vashistha, Atharva Naik, Somak Aditya, Monojit Choudhury
cs.CL
Recent explorations with commercial Large Language Models (LLMs) have shown that non-expert users can jailbreak LLMs by simply manipulating the prompts; resulting in degenerate output behavior, privacy and security breaches, offensive outputs, and violations of content regulator policies. Limited formal studies have be...
http://arxiv.org/pdf/2305.14965v1
2023-05-24T09:57:37
2023-05-24T09:57:37
{ "title": [ -0.035125732, 0.009063721, -0.043395996, -0.007621765, -0.0025024414, -0.020248413, -0.012771606, -0.01928711, -0.06262207, 0.00034356117, 0.007167816, 0.0284729, -0.02053833, -0.059936523, 0.032836914, -0.017181396, -0.00592041, -0....
http://arxiv.org/abs/2305.14964v2
Detecting Multidimensional Political Incivility on Social Media
Sagi Pendzel, Nir Lotan, Alon Zoizner, Einat Minkov
cs.CL
The rise of social media has been argued to intensify uncivil and hostile online political discourse. Yet, to date, there is a lack of clarity on what incivility means in the political sphere. In this work, we utilize a multidimensional perspective of political incivility, developed in the fields of political science a...
http://arxiv.org/pdf/2305.14964v2
2023-11-14T22:41:40
2023-05-24T09:57:12
{ "title": [ 0.03955078, 0.019104004, -0.04434204, -0.064819336, 0.038757324, -0.038848877, 0.014808655, -0.002614975, -0.029342651, 0.026657104, -0.032287598, 0.027160645, -0.007881165, 0.0013828278, -0.025756836, -0.02243042, -0.022476196, 0.03...
http://arxiv.org/abs/2305.14963v1
PESCO: Prompt-enhanced Self Contrastive Learning for Zero-shot Text Classification
Yau-Shian Wang, Ta-Chung Chi, Ruohong Zhang, Yiming Yang
cs.CL
We present PESCO, a novel contrastive learning framework that substantially improves the performance of zero-shot text classification. We formulate text classification as a neural text matching problem where each document is treated as a query, and the system learns the mapping from each query to the relevant class lab...
http://arxiv.org/pdf/2305.14963v1
2023-05-24T09:57:06
2023-05-24T09:57:06
{ "title": [ 0.012573242, -0.000641346, -0.023544312, -0.042266846, -0.004032135, -0.027633667, 0.0004274845, -0.011238098, 0.034423828, 0.022476196, -0.020980835, 0.019805908, -0.022628784, -0.012283325, -0.05529785, 0.0028800964, -0.014419556, ...
http://arxiv.org/abs/2305.14956v3
Editing Common Sense in Transformers
Anshita Gupta, Debanjan Mondal, Akshay Krishna Sheshadri, Wenlong Zhao, Xiang Lorraine Li, Sarah Wiegreffe, Niket Tandon
cs.CL
Editing model parameters directly in Transformers makes updating open-source transformer-based models possible without re-training (Meng et al., 2023). However, these editing methods have only been evaluated on statements about encyclopedic knowledge with a single correct answer. Commonsense knowledge with multiple cor...
http://arxiv.org/pdf/2305.14956v3
2023-10-26T15:38:38
2023-05-24T09:50:54
{ "title": [ 0.028900146, -0.021697998, -0.013168335, 0.0032196045, -0.046325684, -0.012786865, 0.0066070557, -0.0028896332, -0.029327393, 0.054901123, -0.025390625, -0.011398315, 0.014076233, -0.0236969, 0.00818634, -0.014122009, 0.009841919, 0....
http://arxiv.org/abs/2305.14950v2
Adversarial Demonstration Attacks on Large Language Models
Jiongxiao Wang, Zichen Liu, Keun Hee Park, Zhuojun Jiang, Zhaoheng Zheng, Zhuofeng Wu, Muhao Chen, Chaowei Xiao
cs.CL, cs.AI, cs.CR, cs.LG
With the emergence of more powerful large language models (LLMs), such as ChatGPT and GPT-4, in-context learning (ICL) has gained significant prominence in leveraging these models for specific tasks by utilizing data-label pairs as precondition prompts. While incorporating demonstrations can greatly enhance the perform...
http://arxiv.org/pdf/2305.14950v2
2023-10-14T05:03:53
2023-05-24T09:40:56
{ "title": [ -0.01737976, 0.0010051727, -0.029067993, -0.06542969, 0.016830444, 0.004070282, -0.045654297, 0.0012331009, -0.025970459, 0.02368164, 0.0040512085, 0.01071167, -0.015533447, -0.022140503, -0.02330017, -0.0058898926, -0.01322937, 0.04...
http://arxiv.org/abs/2305.14949v2
Cross-lingual Data Augmentation for Document-grounded Dialog Systems in Low Resource Languages
Qi Gou, Zehua Xia, Wenzhe Du
cs.CL
This paper proposes a framework to address the issue of data scarcity in Document-Grounded Dialogue Systems(DGDS). Our model leverages high-resource languages to enhance the capability of dialogue generation in low-resource languages. Specifically, We present a novel pipeline CLEM (Cross-Lingual Enhanced Model) includi...
http://arxiv.org/pdf/2305.14949v2
2023-09-20T07:39:19
2023-05-24T09:40:52
{ "title": [ -0.010940552, -0.012084961, -0.054382324, -0.08508301, 0.0022907257, -0.009124756, -0.07952881, 0.018737793, 0.024368286, 0.044952393, -0.022781372, 0.005958557, 0.020935059, -0.010406494, -0.05319214, 0.004688263, -0.0074310303, 0.0...
http://arxiv.org/abs/2305.14947v2
How Predictable Are Large Language Model Capabilities? A Case Study on BIG-bench
Qinyuan Ye, Harvey Yiyun Fu, Xiang Ren, Robin Jia
cs.CL, cs.LG
We investigate the predictability of large language model (LLM) capabilities: given records of past experiments using different model families, numbers of parameters, tasks, and numbers of in-context examples, can we accurately predict LLM performance on new experiment configurations? Answering this question has practi...
http://arxiv.org/pdf/2305.14947v2
2023-10-31T17:27:07
2023-05-24T09:35:34
{ "title": [ -0.035125732, -0.021575928, -0.041107178, -0.073791504, -0.010093689, 0.009284973, -0.07232666, 0.008506775, -0.002456665, 0.03390503, -0.011222839, 0.0084991455, 0.00390625, -0.016326904, -0.040527344, 0.0040130615, -0.0039253235, 0...
http://arxiv.org/abs/2305.14938v1
Do LLMs Understand Social Knowledge? Evaluating the Sociability of Large Language Models with SocKET Benchmark
Minje Choi, Jiaxin Pei, Sagar Kumar, Chang Shu, David Jurgens
cs.CL, cs.AI
Large language models (LLMs) have been shown to perform well at a variety of syntactic, discourse, and reasoning tasks. While LLMs are increasingly deployed in many forms including conversational agents that interact with humans, we lack a grounded benchmark to measure how well LLMs understand \textit{social} language....
http://arxiv.org/pdf/2305.14938v1
2023-05-24T09:21:06
2023-05-24T09:21:06
{ "title": [ -0.058898926, -0.009338379, -0.022598267, -0.039489746, -0.03479004, -0.0022144318, -0.07098389, -0.0020141602, 0.0062942505, 0.015319824, -0.04925537, 0.013343811, -0.01537323, -0.0032653809, -0.032470703, -0.022994995, -0.013038635, ...
http://arxiv.org/abs/2305.14937v1
A Fair and In-Depth Evaluation of Existing End-to-End Entity Linking Systems
Hannah Bast, Matthias Hertel, Natalie Prange
cs.CL
Existing evaluations of entity linking systems often say little about how the system is going to perform for a particular application. There are four fundamental reasons for this: many benchmarks focus on named entities; it is hard to define which other entities to include; there are ambiguities in entity recognition a...
http://arxiv.org/pdf/2305.14937v1
2023-05-24T09:20:15
2023-05-24T09:20:15
{ "title": [ 0.04663086, 0.0062789917, -0.026382446, -0.0803833, 0.030792236, -0.037719727, 0.002243042, 0.019836426, -0.007499695, 0.03225708, -0.0028686523, -0.015197754, 0.04626465, -0.042144775, 0.010482788, 0.022384644, -0.02204895, 0.041961...