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/2311.09216v1
Assessing Translation capabilities of Large Language Models involving English and Indian Languages
Vandan Mujadia, Ashok Urlana, Yash Bhaskar, Penumalla Aditya Pavani, Kukkapalli Shravya, Parameswari Krishnamurthy, Dipti Misra Sharma
cs.CL, cs.AI
Generative Large Language Models (LLMs) have achieved remarkable advancements in various NLP tasks. In this work, our aim is to explore the multilingual capabilities of large language models by using machine translation as a task involving English and 22 Indian languages. We first investigate the translation capabiliti...
http://arxiv.org/pdf/2311.09216v1
2023-11-15T18:58:19
2023-11-15T18:58:19
{ "title": [ -0.064453125, -0.0057144165, -0.039916992, -0.021408081, -0.019454956, -0.0038814545, -0.053710938, -0.0027427673, -0.023651123, 0.031311035, -0.0021953583, 0.0040512085, 0.0005722046, -0.032836914, -0.035858154, 0.0054397583, -0.0049171...
http://arxiv.org/abs/2311.09214v1
Mind's Mirror: Distilling Self-Evaluation Capability and Comprehensive Thinking from Large Language Models
Weize Liu, Guocong Li, Kai Zhang, Bang Du, Qiyuan Chen, Xuming Hu, Hongxia Xu, Jintai Chen, Jian Wu
cs.CL
Large language models (LLMs) have achieved remarkable advancements in the field of natural language processing. However, the sheer scale and computational demands of these models present formidable challenges when considering their practical deployment in resource-constrained contexts. While techniques such as chain-of...
http://arxiv.org/pdf/2311.09214v1
2023-11-15T18:56:23
2023-11-15T18:56:23
{ "title": [ -0.023345947, -0.021362305, -0.02482605, -0.03744507, -0.0045700073, -0.011581421, -0.01423645, -0.00409317, -0.0057792664, 0.026794434, 0.0065193176, -0.0062561035, -0.01876831, -0.030181885, -0.055877686, -0.002286911, -0.0066947937, ...
http://arxiv.org/abs/2311.09213v1
GRIM: GRaph-based Interactive narrative visualization for gaMes
Jorge Leandro, Sudha Rao, Michael Xu, Weijia Xu, Nebosja Jojic, Chris Brockett, Bill Dolan
cs.CL
Dialogue-based Role Playing Games (RPGs) require powerful storytelling. The narratives of these may take years to write and typically involve a large creative team. In this work, we demonstrate the potential of large generative text models to assist this process. \textbf{GRIM}, a prototype \textbf{GR}aph-based \textbf{...
http://arxiv.org/pdf/2311.09213v1
2023-11-15T18:55:45
2023-11-15T18:55:45
{ "title": [ -0.018997192, -0.0129852295, -0.016784668, -0.036102295, -0.017303467, 0.000603199, -0.020355225, -0.021972656, -0.014404297, 0.022369385, 0.0006008148, 0.012413025, -0.04058838, -0.028640747, -0.0368042, -0.009010315, -0.0110321045, ...
http://arxiv.org/abs/2311.09212v1
Controllable Text Summarization: Unraveling Challenges, Approaches, and Prospects -- A Survey
Ashok Urlana, Pruthwik Mishra, Tathagato Roy, Rahul Mishra
cs.CL, cs.AI, I.2.7
Generic text summarization approaches often fail to address the specific intent and needs of individual users. Recently, scholarly attention has turned to the development of summarization methods that are more closely tailored and controlled to align with specific objectives and user needs. While a growing corpus of re...
http://arxiv.org/pdf/2311.09212v1
2023-11-15T18:55:43
2023-11-15T18:55:43
{ "title": [ -0.014122009, 0.01108551, -0.0103302, -0.05795288, 0.002363205, 0.020263672, -0.038208008, -0.023254395, -0.003824234, 0.0657959, -0.0072364807, 0.037841797, 0.0029945374, 0.008926392, -0.060638428, 0.015510559, -0.01020813, 0.034118...
http://arxiv.org/abs/2311.09210v1
Chain-of-Note: Enhancing Robustness in Retrieval-Augmented Language Models
Wenhao Yu, Hongming Zhang, Xiaoman Pan, Kaixin Ma, Hongwei Wang, Dong Yu
cs.CL, cs.AI
Retrieval-augmented language models (RALMs) represent a substantial advancement in the capabilities of large language models, notably in reducing factual hallucination by leveraging external knowledge sources. However, the reliability of the retrieved information is not always guaranteed. The retrieval of irrelevant da...
http://arxiv.org/pdf/2311.09210v1
2023-11-15T18:54:53
2023-11-15T18:54:53
{ "title": [ 0.013885498, 0.0040893555, -0.0076293945, -0.058502197, 0.008384705, 0.0041160583, -0.039978027, -0.024887085, 0.016983032, 0.048034668, -0.017654419, 0.021316528, 0.028518677, 0.0031223297, -0.05706787, 0.021453857, -0.022155762, 0....
http://arxiv.org/abs/2311.09206v1
TableLlama: Towards Open Large Generalist Models for Tables
Tianshu Zhang, Xiang Yue, Yifei Li, Huan Sun
cs.CL
Semi-structured tables are ubiquitous. There has been a variety of tasks that aim to automatically interpret, augment, and query tables. Current methods often require pretraining on tables or special model architecture design, are restricted to specific table types, or have simplifying assumptions about tables and task...
http://arxiv.org/pdf/2311.09206v1
2023-11-15T18:47:52
2023-11-15T18:47:52
{ "title": [ -0.012420654, 0.00034999847, -0.053985596, -0.09899902, 0.022979736, 0.016098022, -0.07287598, -0.0017051697, 0.035186768, 0.034454346, -0.011985779, 0.0021305084, 0.0036182404, -0.000688076, -0.05834961, 0.024887085, 0.009239197, 0....
http://arxiv.org/abs/2311.09205v1
When Is Multilinguality a Curse? Language Modeling for 250 High- and Low-Resource Languages
Tyler A. Chang, Catherine Arnett, Zhuowen Tu, Benjamin K. Bergen
cs.CL
Multilingual language models are widely used to extend NLP systems to low-resource languages. However, concrete evidence for the effects of multilinguality on language modeling performance in individual languages remains scarce. Here, we pre-train over 10,000 monolingual and multilingual language models for over 250 la...
http://arxiv.org/pdf/2311.09205v1
2023-11-15T18:47:42
2023-11-15T18:47:42
{ "title": [ -0.034423828, -0.0064697266, -0.058258057, -0.036468506, -0.004310608, -0.02015686, -0.053253174, 0.0011739731, -0.02998352, 0.044555664, -0.017028809, 0.00015354156, 0.02520752, -0.020904541, -0.026672363, -0.0013513565, -0.01576233, ...
http://arxiv.org/abs/2311.09204v1
Fusion-Eval: Integrating Evaluators with LLMs
Lei Shu, Nevan Wichers, Liangchen Luo, Yun Zhu, Yinxiao Liu, Jindong Chen, Lei Meng
cs.CL, cs.AI
Evaluating Large Language Models (LLMs) is a complex task, especially considering the intricacies of natural language understanding and the expectations for high-level reasoning. Traditional evaluations typically lean on human-based, model-based, or automatic-metrics-based paradigms, each with its own advantages and sh...
http://arxiv.org/pdf/2311.09204v1
2023-11-15T18:46:56
2023-11-15T18:46:56
{ "title": [ -0.040222168, 0.008514404, -0.05178833, -0.05859375, -0.002565384, 0.012031555, -0.05065918, -0.019058228, -0.010513306, 0.01272583, -0.035339355, 0.007205963, -0.008399963, -0.03262329, -0.015098572, -0.019042969, -0.024917603, 0.03...
http://arxiv.org/abs/2311.09198v1
Never Lost in the Middle: Improving Large Language Models via Attention Strengthening Question Answering
He Junqing, Pan Kunhao, Dong Xiaoqun, Song Zhuoyang, Liu Yibo, Liang Yuxin, Wang Hao, Sun Qianguo, Zhang Songxin, Xie Zejian, Zhang Jiaxing
cs.CL, cs.AI
While large language models (LLMs) are equipped with longer text input capabilities than before, they are struggling to seek correct information in long contexts. The "lost in the middle" problem challenges most LLMs, referring to the dramatic decline in accuracy when correct information is located in the middle. To ov...
http://arxiv.org/pdf/2311.09198v1
2023-11-15T18:42:44
2023-11-15T18:42:44
{ "title": [ 0.0018405914, 0.00018084049, -0.029464722, -0.07305908, 0.00042939186, 0.0020771027, -0.052093506, 0.011764526, 0.031555176, 0.027648926, -0.026184082, 0.008346558, 0.012390137, 0.004142761, -0.04714966, 0.021911621, -0.00440979, 0.0...
http://arxiv.org/abs/2311.09194v1
Structural Priming Demonstrates Abstract Grammatical Representations in Multilingual Language Models
James A. Michaelov, Catherine Arnett, Tyler A. Chang, Benjamin K. Bergen
cs.CL
Abstract grammatical knowledge - of parts of speech and grammatical patterns - is key to the capacity for linguistic generalization in humans. But how abstract is grammatical knowledge in large language models? In the human literature, compelling evidence for grammatical abstraction comes from structural priming. A sen...
http://arxiv.org/pdf/2311.09194v1
2023-11-15T18:39:56
2023-11-15T18:39:56
{ "title": [ 0.012367249, 0.008659363, 0.0051574707, -0.05166626, -0.030059814, 0.0005221367, -0.066589355, -0.026382446, -0.0090789795, 0.022460938, 0.009216309, 0.0034236908, 0.008071899, -0.055114746, -0.009025574, 0.014587402, -0.009819031, 0...
http://arxiv.org/abs/2311.09193v1
The Role of Chain-of-Thought in Complex Vision-Language Reasoning Task
Yifan Wu, Pengchuan Zhang, Wenhan Xiong, Barlas Oguz, James C. Gee, Yixin Nie
cs.CL, cs.AI, cs.CV
The study explores the effectiveness of the Chain-of-Thought approach, known for its proficiency in language tasks by breaking them down into sub-tasks and intermediate steps, in improving vision-language tasks that demand sophisticated perception and reasoning. We present the "Description then Decision" strategy, whic...
http://arxiv.org/pdf/2311.09193v1
2023-11-15T18:39:21
2023-11-15T18:39:21
{ "title": [ -0.008163452, -0.02947998, -0.024810791, -0.047454834, -0.013473511, -0.026687622, -0.0051727295, 0.011764526, 0.020401001, 0.0033798218, -0.032287598, -0.015022278, -0.001745224, -0.030014038, -0.033233643, -0.018447876, 0.0017757416, ...
http://arxiv.org/abs/2311.09189v1
PsyEval: A Comprehensive Large Language Model Evaluation Benchmark for Mental Health
Haoan Jin, Siyuan Chen, Mengyue Wu, Kenny Q. Zhu
cs.CL
Recently, there has been a growing interest in utilizing large language models (LLMs) in mental health research, with studies showcasing their remarkable capabilities, such as disease detection. However, there is currently a lack of a comprehensive benchmark for evaluating the capability of LLMs in this domain. Therefo...
http://arxiv.org/pdf/2311.09189v1
2023-11-15T18:32:27
2023-11-15T18:32:27
{ "title": [ -0.021774292, -0.0051460266, -0.030838013, -0.07336426, 0.014228821, 0.0060043335, -0.07269287, -0.010482788, 0.00021481514, 0.034179688, 0.01675415, 0.002571106, 0.008811951, -0.038757324, -0.043273926, 0.021774292, -0.010955811, 0....
http://arxiv.org/abs/2311.09188v1
Towards Verifiable Text Generation with Symbolic References
Lucas Torroba Hennigen, Shannon Shen, Aniruddha Nrusimha, Bernhard Gapp, David Sontag, Yoon Kim
cs.CL, cs.AI, cs.LG
Large language models (LLMs) have demonstrated an impressive ability to synthesize plausible and fluent text. However they remain vulnerable to hallucinations, and thus their outputs generally require manual human verification for high-stakes applications, which can be time-consuming and difficult. This paper proposes ...
http://arxiv.org/pdf/2311.09188v1
2023-11-15T18:28:29
2023-11-15T18:28:29
{ "title": [ -0.023223877, -0.009231567, -0.049957275, -0.08843994, 0.004753113, 0.018966675, -0.08129883, 0.009971619, -0.015319824, 0.05368042, 0.0066833496, 0.014404297, 0.008277893, -0.0076942444, -0.051361084, 0.008895874, -0.030685425, 0.04...
http://arxiv.org/abs/2311.09184v1
Benchmarking Generation and Evaluation Capabilities of Large Language Models for Instruction Controllable Summarization
Yixin Liu, Alexander R. Fabbri, Jiawen Chen, Yilun Zhao, Simeng Han, Shafiq Joty, Pengfei Liu, Dragomir Radev, Chien-Sheng Wu, Arman Cohan
cs.CL, cs.LG
While large language models (LLMs) already achieve strong performance on standard generic summarization benchmarks, their performance on more complex summarization task settings is less studied. Therefore, we benchmark LLMs on instruction controllable text summarization, where the model input consists of both a source ...
http://arxiv.org/pdf/2311.09184v1
2023-11-15T18:25:26
2023-11-15T18:25:26
{ "title": [ -0.030197144, -0.010505676, -0.02545166, -0.06762695, -0.018951416, 0.019119263, -0.07946777, -0.00504303, -0.004722595, 0.049438477, -0.0118255615, 0.032409668, -0.021194458, -0.00033521652, -0.059539795, 0.0030574799, -0.0065841675, ...
http://arxiv.org/abs/2311.09182v1
ContraDoc: Understanding Self-Contradictions in Documents with Large Language Models
Jierui Li, Vipul Raheja, Dhruv Kumar
cs.CL
In recent times, large language models (LLMs) have shown impressive performance on various document-level tasks such as document classification, summarization, and question-answering. However, research on understanding their capabilities on the task of self-contradictions in long documents has been very limited. In thi...
http://arxiv.org/pdf/2311.09182v1
2023-11-15T18:23:17
2023-11-15T18:23:17
{ "title": [ -0.036987305, -0.00566864, -0.03564453, -0.050476074, -0.0059165955, 0.0030727386, -0.04815674, -0.0146102905, 0.018630981, 0.012863159, -0.023254395, -0.009765625, 0.00086832047, -0.013259888, -0.055236816, -0.0043563843, -0.02368164, ...
http://arxiv.org/abs/2311.09180v1
PEARL: Personalizing Large Language Model Writing Assistants with Generation-Calibrated Retrievers
Sheshera Mysore, Zhuoran Lu, Mengting Wan, Longqi Yang, Steve Menezes, Tina Baghaee, Emmanuel Barajas Gonzalez, Jennifer Neville, Tara Safavi
cs.CL, cs.HC, cs.IR
Powerful large language models have facilitated the development of writing assistants that promise to significantly improve the quality and efficiency of composition and communication. However, a barrier to effective assistance is the lack of personalization in LLM outputs to the author's communication style and specia...
http://arxiv.org/pdf/2311.09180v1
2023-11-15T18:19:58
2023-11-15T18:19:58
{ "title": [ -0.04925537, -0.015060425, -0.050079346, -0.04449463, -0.03262329, 0.016174316, -0.06341553, 0.016647339, -0.00919342, 0.04727173, -0.023422241, 0.026321411, 0.03100586, -0.0053634644, -0.019729614, 0.010772705, -0.016845703, 0.04714...
http://arxiv.org/abs/2311.09179v1
SiRA: Sparse Mixture of Low Rank Adaptation
Yun Zhu, Nevan Wichers, Chu-Cheng Lin, Xinyi Wang, Tianlong Chen, Lei Shu, Han Lu, Canoee Liu, Liangchen Luo, Jindong Chen, Lei Meng
cs.CL
Parameter Efficient Tuning has been an prominent approach to adapt the Large Language Model to downstream tasks. Most previous works considers adding the dense trainable parameters, where all parameters are used to adapt certain task. We found this less effective empirically using the example of LoRA that introducing m...
http://arxiv.org/pdf/2311.09179v1
2023-11-15T18:15:37
2023-11-15T18:15:37
{ "title": [ -0.04119873, -0.02029419, -0.054901123, -0.032043457, -0.007698059, -0.0020809174, -0.033172607, -0.018035889, 0.010887146, 0.022735596, -0.016433716, 0.0050582886, 0.004131317, -0.028656006, -0.013450623, 0.013015747, 0.019638062, -...
http://arxiv.org/abs/2311.09174v1
AbsPyramid: Benchmarking the Abstraction Ability of Language Models with a Unified Entailment Graph
Zhaowei Wang, Haochen Shi, Weiqi Wang, Tianqing Fang, Hongming Zhang, Sehyun Choi, Xin Liu, Yangqiu Song
cs.CL, cs.AI
Cognitive research indicates that abstraction ability is essential in human intelligence, which remains under-explored in language models. In this paper, we present AbsPyramid, a unified entailment graph of 221K textual descriptions of abstraction knowledge. While existing resources only touch nouns or verbs within sim...
http://arxiv.org/pdf/2311.09174v1
2023-11-15T18:11:23
2023-11-15T18:11:23
{ "title": [ -0.0052833557, -0.015090942, -0.047912598, -0.10412598, 0.014129639, -0.0020313263, -0.07208252, 0.015975952, 0.00617218, 0.04168701, 0.012489319, 0.020095825, 0.0075798035, -0.017501831, -0.04058838, 0.04043579, -0.021697998, 0.0462...
http://arxiv.org/abs/2311.09154v1
CLEAN-EVAL: Clean Evaluation on Contaminated Large Language Models
Wenhong Zhu, Hongkun Hao, Zhiwei He, Yunze Song, Yumeng Zhang, Hanxu Hu, Yiran Wei, Rui Wang, Hongyuan Lu
cs.CL
We are currently in an era of fierce competition among various large language models (LLMs) continuously pushing the boundaries of benchmark performance. However, genuinely assessing the capabilities of these LLMs has become a challenging and critical issue due to potential data contamination, and it wastes dozens of t...
http://arxiv.org/pdf/2311.09154v1
2023-11-15T17:50:30
2023-11-15T17:50:30
{ "title": [ -0.0036678314, 0.0023326874, -0.051605225, -0.107421875, 0.035949707, 0.006893158, -0.067993164, -0.0005450249, 0.0043678284, 0.026260376, -0.0014181137, -0.019744873, -0.0031166077, -0.023544312, -0.053833008, 0.0019426346, -0.031280518...
http://arxiv.org/abs/2311.09149v1
Temporal Knowledge Question Answering via Abstract Reasoning Induction
Ziyang Chen, Dongfang Li, Xiang Zhao, Baotian Hu, Min Zhang
cs.CL, cs.AI, I.2.7
In this paper, we tackle the significant challenge of temporal knowledge reasoning in Large Language Models (LLMs), an area where such models frequently encounter difficulties. These difficulties often result in the generation of misleading or incorrect information, primarily due to their limited capacity to process ev...
http://arxiv.org/pdf/2311.09149v1
2023-11-15T17:46:39
2023-11-15T17:46:39
{ "title": [ -0.009689331, -0.01260376, -0.06414795, -0.09503174, 0.011695862, -0.009994507, -0.0440979, 0.032226562, 0.031280518, 0.015419006, -0.055633545, 0.005432129, -0.035858154, -0.0007596016, -0.061462402, -0.017929077, 0.0028247833, 0.01...
http://arxiv.org/abs/2311.09144v1
Grounding or Guesswork? Large Language Models are Presumptive Grounders
Omar Shaikh, Kristina Gligorić, Ashna Khetan, Matthias Gerstgrasser, Diyi Yang, Dan Jurafsky
cs.CL, cs.HC
Effective conversation requires common ground: a shared understanding between the participants. Common ground, however, does not emerge spontaneously in conversation. Speakers and listeners work together to both identify and construct a shared basis while avoiding misunderstanding. To accomplish grounding, humans rely ...
http://arxiv.org/pdf/2311.09144v1
2023-11-15T17:40:27
2023-11-15T17:40:27
{ "title": [ -0.018997192, -0.007537842, -0.047454834, -0.071899414, 0.002708435, -0.01449585, -0.057769775, 0.0014791489, -0.016738892, 0.0181427, -0.028671265, -0.011909485, -0.0028839111, -0.024551392, -0.040161133, -0.014320374, -0.0048332214, ...
http://arxiv.org/abs/2311.09136v1
RRescue: Ranking LLM Responses to Enhance Reasoning Over Context
Yikun Wang, Rui Zheng, Haoming Li, Qi Zhang, Tao Gui, Fei Liu
cs.CL
Effectively using a given context is paramount for large language models. A context window can include task specifications, retrieved documents, previous conversations, and even model self-reflections, functioning similarly to episodic memory. While efforts are being made to expand the context window, studies indicate ...
http://arxiv.org/pdf/2311.09136v1
2023-11-15T17:27:14
2023-11-15T17:27:14
{ "title": [ -0.023498535, -0.013633728, -0.049224854, -0.04257202, -0.005844116, -0.005493164, -0.044403076, -0.009109497, -0.0023174286, -0.0026817322, -0.022964478, 0.024398804, -0.025604248, -0.03286743, -0.009475708, -0.018585205, -0.019622803, ...
http://arxiv.org/abs/2311.09132v1
Aligning Neural Machine Translation Models: Human Feedback in Training and Inference
Miguel Moura Ramos, Patrick Fernandes, António Farinhas, André F. T. Martins
cs.CL
Reinforcement learning from human feedback (RLHF) is a recent technique to improve the quality of the text generated by a language model, making it closer to what humans would generate. A core ingredient in RLHF's success in aligning and improving large language models (LLMs) is its reward model, trained using human fe...
http://arxiv.org/pdf/2311.09132v1
2023-11-15T17:21:58
2023-11-15T17:21:58
{ "title": [ -0.04458618, 0.0035247803, -0.019241333, -0.029586792, -0.013114929, 0.008430481, -0.04949951, -0.0256958, -0.013076782, 0.043395996, -0.007972717, -0.010894775, 0.013626099, -0.017410278, -0.054260254, 0.0095825195, -0.0014448166, 0...
http://arxiv.org/abs/2311.09130v1
Social Meme-ing: Measuring Linguistic Variation in Memes
Naitian Zhou, David Jurgens, David Bamman
cs.CL
Much work in the space of NLP has used computational methods to explore sociolinguistic variation in text. In this paper, we argue that memes, as multimodal forms of language comprised of visual templates and text, also exhibit meaningful social variation. We construct a computational pipeline to cluster individual ins...
http://arxiv.org/pdf/2311.09130v1
2023-11-15T17:20:20
2023-11-15T17:20:20
{ "title": [ -0.005142212, 0.00283432, -0.037353516, -0.0019760132, 0.016281128, -0.03237915, 0.011375427, -0.029891968, -0.00084257126, 0.044281006, -0.0054969788, 0.046875, -0.011192322, -0.009063721, -0.03729248, 0.027694702, -0.010856628, 0.0...
http://arxiv.org/abs/2311.09122v1
Universal NER: A Gold-Standard Multilingual Named Entity Recognition Benchmark
Stephen Mayhew, Terra Blevins, Shuheng Liu, Marek Šuppa, Hila Gonen, Joseph Marvin Imperial, Börje F. Karlsson, Peiqin Lin, Nikola Ljubešić, LJ Miranda, Barbara Plank, Arij Riabi, Yuval Pinter
cs.CL
We introduce Universal NER (UNER), an open, community-driven project to develop gold-standard NER benchmarks in many languages. The overarching goal of UNER is to provide high-quality, cross-lingually consistent annotations to facilitate and standardize multilingual NER research. UNER v1 contains 18 datasets annotated ...
http://arxiv.org/pdf/2311.09122v1
2023-11-15T17:09:54
2023-11-15T17:09:54
{ "title": [ -0.021392822, -0.009735107, -0.04638672, -0.07611084, 0.0234375, 0.0028896332, -0.044403076, 0.0051612854, -0.001083374, 0.03137207, -0.02218628, 0.03161621, 0.009429932, -0.0057144165, -0.0256958, 0.018295288, -0.011741638, 0.051208...
http://arxiv.org/abs/2311.09117v1
R-Spin: Efficient Speaker and Noise-invariant Representation Learning with Acoustic Pieces
Heng-Jui Chang, James Glass
cs.CL, cs.SD, eess.AS
This paper introduces Robust Spin (R-Spin), a data-efficient self-supervised fine-tuning framework for speaker and noise-invariant speech representations by learning discrete acoustic units with speaker-invariant clustering (Spin). R-Spin resolves Spin's issues and enhances content representations by learning to predic...
http://arxiv.org/pdf/2311.09117v1
2023-11-15T17:07:44
2023-11-15T17:07:44
{ "title": [ 0.012557983, 0.011672974, -0.050079346, -0.07171631, 0.019958496, -0.010864258, -0.034973145, 0.010620117, -0.0146102905, 0.033172607, -0.024475098, 0.018310547, -0.009170532, -0.014221191, -0.028244019, -0.011810303, 0.008377075, 0....
http://arxiv.org/abs/2311.09114v1
Ever: Mitigating Hallucination in Large Language Models through Real-Time Verification and Rectification
Haoqiang Kang, Juntong Ni, Huaxiu Yao
cs.CL, cs.AI, cs.LG
Large Language Models (LLMs) have demonstrated remarkable proficiency in generating fluent text. However, they often encounter the challenge of generating inaccurate or hallucinated content. This issue is common in both non-retrieval-based generation and retrieval-augmented generation approaches, and existing post-hoc ...
http://arxiv.org/pdf/2311.09114v1
2023-11-15T17:04:56
2023-11-15T17:04:56
{ "title": [ -0.013618469, -0.00017797947, -0.035491943, -0.06750488, 0.013435364, 0.0027999878, -0.059661865, -0.008682251, -0.0050582886, 0.04296875, -0.0076026917, 0.0056114197, 0.02330017, -0.008468628, -0.04864502, 0.017623901, -0.018600464, ...
http://arxiv.org/abs/2311.09109v1
Does Pre-trained Language Model Actually Infer Unseen Links in Knowledge Graph Completion?
Yusuke Sakai, Hidetaka Kamigaito, Katsuhiko Hayashi, Taro Watanabe
cs.CL, cs.AI, cs.LG
Knowledge graphs (KGs) consist of links that describe relationships between entities. Due to the difficulty of manually enumerating all relationships between entities, automatically completing them is essential for KGs. Knowledge Graph Completion (KGC) is a task that infers unseen relationships between entities in a KG...
http://arxiv.org/pdf/2311.09109v1
2023-11-15T16:56:49
2023-11-15T16:56:49
{ "title": [ -0.00440979, -0.027145386, -0.024871826, -0.06304932, -0.060150146, -0.0074539185, -0.07171631, 0.013122559, -0.015235901, 0.014533997, -0.02798462, -0.0014190674, 0.0057144165, -0.027420044, 0.006793976, -0.029769897, -0.008972168, ...
http://arxiv.org/abs/2311.09106v1
"We Demand Justice!": Towards Grounding Political Text in Social Context
Rajkumar Pujari, Chengfei Wu, Dan Goldwasser
cs.CL
Social media discourse from US politicians frequently consists of 'seemingly similar language used by opposing sides of the political spectrum'. But often, it translates to starkly contrasting real-world actions. For instance, "We need to keep our students safe from mass shootings" may signal either "arming teachers to...
http://arxiv.org/pdf/2311.09106v1
2023-11-15T16:53:35
2023-11-15T16:53:35
{ "title": [ 0.0007214546, -0.0037345886, -0.025375366, -0.056243896, 0.013244629, -0.019363403, -0.01626587, -0.0052223206, 0.0042762756, 0.020889282, -0.015487671, 0.024215698, -0.00831604, -0.005508423, -0.04449463, 0.0017299652, -0.010932922, ...
http://arxiv.org/abs/2311.09105v1
MAVEN-Arg: Completing the Puzzle of All-in-One Event Understanding Dataset with Event Argument Annotation
Xiaozhi Wang, Hao Peng, Yong Guan, Kaisheng Zeng, Jianhui Chen, Lei Hou, Xu Han, Yankai Lin, Zhiyuan Liu, Ruobing Xie, Jie Zhou, Juanzi Li
cs.CL
Understanding events in texts is a core objective of natural language understanding, which requires detecting event occurrences, extracting event arguments, and analyzing inter-event relationships. However, due to the annotation challenges brought by task complexity, a large-scale dataset covering the full process of e...
http://arxiv.org/pdf/2311.09105v1
2023-11-15T16:52:14
2023-11-15T16:52:14
{ "title": [ 0.007171631, -0.023620605, -0.048461914, -0.12200928, -0.007507324, -0.010643005, -0.071899414, 0.025268555, -0.013671875, 0.024658203, 0.02154541, 0.010795593, 0.013801575, -0.048065186, -0.006816864, 0.018066406, -0.00082302094, 0....
http://arxiv.org/abs/2311.09101v1
Towards A Unified View of Answer Calibration for Multi-Step Reasoning
Shumin Deng, Ningyu Zhang, Nay Oo, Bryan Hooi
cs.CL, cs.AI, cs.IR, cs.LG
Large Language Models (LLMs) employing Chain-of-Thought (CoT) prompting have broadened the scope for improving multi-step reasoning capabilities. Usually, answer calibration strategies such as step-level or path-level calibration play a vital role in multi-step reasoning. While effective, there remains a significant ga...
http://arxiv.org/pdf/2311.09101v1
2023-11-15T16:47:57
2023-11-15T16:47:57
{ "title": [ -0.041931152, -0.0020275116, -0.057647705, -0.076049805, 0.0034637451, 0.006324768, -0.071777344, 0.0063171387, 0.020935059, 0.0054626465, -0.027389526, 0.008392334, -0.0385437, -0.018569946, -0.041015625, 0.0019083023, 0.014732361, ...
http://arxiv.org/abs/2311.09096v1
Defending Large Language Models Against Jailbreaking Attacks Through Goal Prioritization
Zhexin Zhang, Junxiao Yang, Pei Ke, Minlie Huang
cs.CL
Large Language Models (LLMs) continue to advance in their capabilities, yet this progress is accompanied by a growing array of safety risks. While significant attention has been dedicated to exploiting weaknesses in LLMs through jailbreaking attacks, there remains a paucity of exploration into defending against these a...
http://arxiv.org/pdf/2311.09096v1
2023-11-15T16:42:29
2023-11-15T16:42:29
{ "title": [ -0.025634766, -0.008277893, -0.062561035, -0.078552246, 0.0013027191, -0.013847351, -0.05368042, 0.035247803, -0.06616211, 0.028793335, 0.03765869, 0.010765076, -0.0023155212, -0.044067383, -0.005470276, -0.011619568, -0.026367188, 0...
http://arxiv.org/abs/2311.09090v1
Social Bias Probing: Fairness Benchmarking for Language Models
Marta Marchiori Manerba, Karolina Stańczak, Riccardo Guidotti, Isabelle Augenstein
cs.CL
Large language models have been shown to encode a variety of social biases, which carries the risk of downstream harms. While the impact of these biases has been recognized, prior methods for bias evaluation have been limited to binary association tests on small datasets, offering a constrained view of the nature of so...
http://arxiv.org/pdf/2311.09090v1
2023-11-15T16:35:59
2023-11-15T16:35:59
{ "title": [ -0.03503418, -0.001947403, -0.041748047, -0.048309326, 0.0014543533, -0.009407043, -0.04727173, -0.007106781, 0.0121154785, 0.043395996, -0.030853271, 0.007911682, 0.025543213, -0.007507324, -0.0592041, 0.016799927, -0.03173828, 0.05...
http://arxiv.org/abs/2311.09086v1
The Uli Dataset: An Exercise in Experience Led Annotation of oGBV
Arnav Arora, Maha Jinadoss, Cheshta Arora, Denny George, Brindaalakshmi, Haseena Dawood Khan, Kirti Rawat, Div, Ritash, Seema Mathur, Shivani Yadav, Shehla Rashid Shora, Rie Raut, Sumit Pawar, Apurva Paithane, Sonia, Vivek, Dharini Priscilla, Khairunnisha, Grace Banu, Ambika Tandon, Rishav Thakker, Rahul Dev Korr...
cs.CL, cs.AI, cs.SI
Online gender based violence has grown concomitantly with adoption of the internet and social media. Its effects are worse in the Global majority where many users use social media in languages other than English. The scale and volume of conversations on the internet has necessitated the need for automated detection of ...
http://arxiv.org/pdf/2311.09086v1
2023-11-15T16:30:44
2023-11-15T16:30:44
{ "title": [ 0.010131836, -0.002588272, -0.03375244, -0.08886719, -0.013557434, -0.0016746521, -0.082214355, -0.020614624, -0.01134491, 0.040130615, -0.025817871, 0.0029125214, 0.005203247, -0.042053223, -0.032958984, 0.0070343018, 0.02104187, 0....
http://arxiv.org/abs/2311.09071v1
How Multilingual is Multilingual LLM?
Fei Yuan, Shuai Yuan, Zhiyong Wu, Lei Li
cs.CL, cs.AI
Large Language Models (LLMs), trained predominantly on extensive English data, often exhibit limitations when applied to other languages. Current research is primarily focused on enhancing the multilingual capabilities of these models by employing various tuning strategies. Despite their effectiveness in certain langua...
http://arxiv.org/pdf/2311.09071v1
2023-11-15T16:13:14
2023-11-15T16:13:14
{ "title": [ -0.06939697, -0.004852295, -0.048919678, -0.0019340515, -0.053649902, 0.023117065, -0.09991455, 0.0040016174, -0.042877197, 0.026550293, 0.019927979, 0.0061569214, -0.030929565, -0.046173096, 0.036468506, -0.0051460266, -0.011062622, ...
http://arxiv.org/abs/2311.09069v1
How Well Do Large Language Models Truly Ground?
Hyunji Lee, Sejune Joo, Chaeeun Kim, Joel Jang, Doyoung Kim, Kyoung-Woon On, Minjoon Seo
cs.CL, cs.AI
Reliance on the inherent knowledge of Large Language Models (LLMs) can cause issues such as hallucinations, lack of control, and difficulties in integrating variable knowledge. To mitigate this, LLMs can be probed to generate responses by grounding on external context, often given as input (knowledge-augmented models)....
http://arxiv.org/pdf/2311.09069v1
2023-11-15T16:11:27
2023-11-15T16:11:27
{ "title": [ -0.0066833496, -0.001326561, -0.02293396, -0.059783936, -0.038726807, -0.0000731349, -0.0680542, 0.0030727386, 0.0001718998, 0.015686035, -0.039520264, -0.035858154, 0.014015198, -0.033569336, -0.011070251, -0.009811401, -0.000035643578,...
http://arxiv.org/abs/2311.09066v1
Identifying Self-Disclosures of Use, Misuse and Addiction in Community-based Social Media Posts
Chenghao Yang, Tuhin Chakrabarty, Karli R Hochstatter, Melissa N Slavin, Nabila El-Bassel, Smaranda Muresan
cs.CL
In the last decade, the United States has lost more than 500,000 people from an overdose involving prescription and illicit opioids (https://www.cdc.gov/drugoverdose/epidemic/index.html) making it a national public health emergency (USDHHS, 2017). To more effectively prevent unintentional opioid overdoses, medical prac...
http://arxiv.org/pdf/2311.09066v1
2023-11-15T16:05:55
2023-11-15T16:05:55
{ "title": [ 0.04284668, 0.029129028, -0.032470703, -0.058441162, 0.060272217, -0.020950317, 0.01637268, -0.00932312, -0.04309082, 0.024414062, -0.02319336, 0.05117798, -0.036193848, -0.010971069, -0.014053345, -0.017227173, 0.0023784637, 0.04748...
http://arxiv.org/abs/2311.09060v1
Do Localization Methods Actually Localize Memorized Data in LLMs?
Ting-Yun Chang, Jesse Thomason, Robin Jia
cs.CL
Large language models (LLMs) can memorize many pretrained sequences verbatim. This paper studies if we can locate a small set of neurons in LLMs responsible for memorizing a given sequence. While the concept of localization is often mentioned in prior work, methods for localization have never been systematically and di...
http://arxiv.org/pdf/2311.09060v1
2023-11-15T15:52:40
2023-11-15T15:52:40
{ "title": [ -0.022827148, 0.0062294006, -0.004108429, 0.033477783, -0.03753662, 0.001203537, -0.0071258545, -0.05529785, -0.025100708, 0.03555298, -0.029785156, -0.0093307495, -0.006225586, -0.020492554, 0.00982666, -0.039093018, 0.013465881, 0....
http://arxiv.org/abs/2311.09053v1
Assessing Knowledge Editing in Language Models via Relation Perspective
Yifan Wei, Xiaoyan Yu, Huanhuan Ma, Fangyu Lei, Yixuan Weng, Ran Song, Kang Liu
cs.CL, cs.AI
Knowledge Editing (KE) for modifying factual knowledge in Large Language Models (LLMs) has been receiving increasing attention. However, existing knowledge editing methods are entity-centric, and it is unclear whether this approach is suitable for a relation-centric perspective. To address this gap, this paper construc...
http://arxiv.org/pdf/2311.09053v1
2023-11-15T15:44:42
2023-11-15T15:44:42
{ "title": [ -0.05319214, -0.022872925, -0.06427002, -0.06524658, -0.021743774, 0.005218506, -0.06500244, 0.036895752, -0.0052261353, 0.026153564, -0.014259338, -0.031402588, 0.011398315, -0.021850586, -0.060821533, -0.0158844, -0.036010742, 0.02...
http://arxiv.org/abs/2311.09048v1
GRASP: A novel benchmark for evaluating language GRounding And Situated Physics understanding in multimodal language models
Serwan Jassim, Mario Holubar, Annika Richter, Cornelius Wolff, Xenia Ohmer, Elia Bruni
cs.CL
This paper presents GRASP, a novel benchmark to evaluate the language grounding and physical understanding capabilities of video-based multimodal large language models (LLMs). This evaluation is accomplished via a two-tier approach leveraging Unity simulations. The initial level tests for language grounding by assessin...
http://arxiv.org/pdf/2311.09048v1
2023-11-15T15:38:28
2023-11-15T15:38:28
{ "title": [ -0.01852417, 0.01159668, -0.03604126, -0.05871582, -0.011985779, -0.012237549, -0.070007324, -0.007106781, -0.014625549, 0.03665161, 0.015434265, 0.020721436, 0.014465332, -0.02609253, -0.027770996, 0.014129639, -0.024154663, 0.03286...
http://arxiv.org/abs/2311.09033v1
MELA: Multilingual Evaluation of Linguistic Acceptability
Ziyin Zhang, Yikang Liu, Weifang Huang, Junyu Mao, Rui Wang, Hai Hu
cs.CL, cs.AI
Recent benchmarks for Large Language Models (LLMs) have mostly focused on application-driven tasks such as complex reasoning and code generation, and this has led to a scarcity in purely linguistic evaluation of LLMs. Against this background, we introduce Multilingual Evaluation of Linguistic Acceptability -- MELA, the...
http://arxiv.org/pdf/2311.09033v1
2023-11-15T15:25:28
2023-11-15T15:25:28
{ "title": [ -0.02848816, -0.00049734116, -0.021728516, -0.027511597, 0.014961243, 0.007926941, -0.026123047, -0.022476196, 0.01309967, 0.019042969, -0.005882263, -0.0112838745, -0.00010114908, -0.028900146, -0.0513916, 0.008796692, -0.033172607, ...
http://arxiv.org/abs/2311.09022v1
Exploring the Potential of Large Language Models in Computational Argumentation
Guizhen Chen, Liying Cheng, Luu Anh Tuan, Lidong Bing
cs.CL
Computational argumentation has become an essential tool in various fields, including artificial intelligence, law, and public policy. It is an emerging research field in natural language processing (NLP) that attracts increasing attention. Research on computational argumentation mainly involves two types of tasks: arg...
http://arxiv.org/pdf/2311.09022v1
2023-11-15T15:12:15
2023-11-15T15:12:15
{ "title": [ 0.000667572, -0.020080566, -0.028778076, -0.074035645, -0.007091522, -0.0037593842, -0.0446167, 0.009521484, -0.013046265, 0.021530151, 0.003004074, -0.01423645, -0.0045814514, -0.029937744, -0.032470703, -0.00894165, -0.014122009, 0...
http://arxiv.org/abs/2311.09008v1
End-to-end Task-oriented Dialogue: A Survey of Tasks, Methods, and Future Directions
Libo Qin, Wenbo Pan, Qiguang Chen, Lizi Liao, Zhou Yu, Yue Zhang, Wanxiang Che, Min Li
cs.CL
End-to-end task-oriented dialogue (EToD) can directly generate responses in an end-to-end fashion without modular training, which attracts escalating popularity. The advancement of deep neural networks, especially the successful use of large pre-trained models, has further led to significant progress in EToD research i...
http://arxiv.org/pdf/2311.09008v1
2023-11-15T14:50:16
2023-11-15T14:50:16
{ "title": [ -0.020507812, -0.009780884, -0.040985107, -0.08154297, 0.035064697, 0.0052757263, -0.03463745, 0.011077881, -0.0030384064, 0.03540039, 0.027069092, -0.008766174, 0.00021719933, -0.045043945, -0.046966553, 0.017227173, -0.009895325, 0...
http://arxiv.org/abs/2311.09006v1
Data Similarity is Not Enough to Explain Language Model Performance
Gregory Yauney, Emily Reif, David Mimno
cs.CL, cs.LG
Large language models achieve high performance on many but not all downstream tasks. The interaction between pretraining data and task data is commonly assumed to determine this variance: a task with data that is more similar to a model's pretraining data is assumed to be easier for that model. We test whether distribu...
http://arxiv.org/pdf/2311.09006v1
2023-11-15T14:48:08
2023-11-15T14:48:08
{ "title": [ -0.035461426, 0.010520935, -0.023727417, -0.023117065, -0.044128418, -0.004928589, -0.034088135, -0.009017944, -0.006549835, 0.03793335, -0.027816772, -0.039794922, -0.0012769699, -0.03918457, -0.041809082, -0.018676758, -0.0008955002, ...
http://arxiv.org/abs/2311.09000v1
Factcheck-GPT: End-to-End Fine-Grained Document-Level Fact-Checking and Correction of LLM Output
Yuxia Wang, Revanth Gangi Reddy, Zain Muhammad Mujahid, Arnav Arora, Aleksandr Rubashevskii, Jiahui Geng, Osama Mohammed Afzal, Liangming Pan, Nadav Borenstein, Aditya Pillai, Isabelle Augenstein, Iryna Gurevych, Preslav Nakov
cs.CL
The increased use of large language models (LLMs) across a variety of real-world applications calls for mechanisms to verify the factual accuracy of their outputs. In this work, we present a holistic end-to-end solution for annotating the factuality of LLM-generated responses, which encompasses a multi-stage annotation...
http://arxiv.org/pdf/2311.09000v1
2023-11-15T14:41:57
2023-11-15T14:41:57
{ "title": [ -0.018035889, -0.014190674, -0.053253174, -0.064208984, -0.030090332, -0.000075399876, -0.075683594, 0.016677856, -0.028045654, 0.022949219, -0.037017822, 0.020217896, 0.02331543, -0.041656494, 0.012359619, -0.034484863, -0.027282715, ...
http://arxiv.org/abs/2311.08993v1
When does In-context Learning Fall Short and Why? A Study on Specification-Heavy Tasks
Hao Peng, Xiaozhi Wang, Jianhui Chen, Weikai Li, Yunjia Qi, Zimu Wang, Zhili Wu, Kaisheng Zeng, Bin Xu, Lei Hou, Juanzi Li
cs.CL, cs.AI
In-context learning (ICL) has become the default method for using large language models (LLMs), making the exploration of its limitations and understanding the underlying causes crucial. In this paper, we find that ICL falls short of handling specification-heavy tasks, which are tasks with complicated and extensive tas...
http://arxiv.org/pdf/2311.08993v1
2023-11-15T14:26:30
2023-11-15T14:26:30
{ "title": [ 0.006538391, -0.0126571655, -0.041137695, -0.038879395, 0.013916016, -0.03591919, 0.007167816, -0.010375977, -0.007160187, 0.021194458, -0.03842163, 0.02330017, -0.029373169, -0.0002808571, -0.034210205, -0.03186035, -0.0035152435, 0...
http://arxiv.org/abs/2311.08982v1
SentAlign: Accurate and Scalable Sentence Alignment
Steinþór Steingrímsson, Hrafn Loftsson, Andy Way
cs.CL
We present SentAlign, an accurate sentence alignment tool designed to handle very large parallel document pairs. Given user-defined parameters, the alignment algorithm evaluates all possible alignment paths in fairly large documents of thousands of sentences and uses a divide-and-conquer approach to align documents con...
http://arxiv.org/pdf/2311.08982v1
2023-11-15T14:15:41
2023-11-15T14:15:41
{ "title": [ -0.03277588, -0.0084991455, -0.033050537, -0.051086426, 0.0034446716, 0.010124207, -0.048034668, -0.014205933, 0.011230469, 0.052764893, 0.00012266636, 0.009269714, 0.024627686, -0.0066871643, -0.058502197, 0.023361206, -0.011726379, ...
http://arxiv.org/abs/2311.08981v1
Speculative Contrastive Decoding
Hongyi Yuan, Keming Lu, Fei Huang, Zheng Yuan, Chang Zhou
cs.CL
Large language models (LLMs) have shown extraordinary performance in various language tasks, but high computational requirements hinder their widespread deployment. Speculative decoding, which uses amateur models to predict the generation of expert models, has been proposed as a way to accelerate LLM inference. However...
http://arxiv.org/pdf/2311.08981v1
2023-11-15T14:15:30
2023-11-15T14:15:30
{ "title": [ -0.0056877136, -0.0021018982, -0.009841919, -0.029388428, -0.048065186, -0.010467529, -0.031463623, -0.023376465, 0.028778076, 0.059051514, -0.0309906, -0.037353516, 0.019729614, -0.030471802, -0.04788208, 0.0127334595, 0.005683899, ...
http://arxiv.org/abs/2311.08968v1
Identifying Linear Relational Concepts in Large Language Models
David Chanin, Anthony Hunter, Oana-Maria Camburu
cs.CL, cs.AI
Transformer language models (LMs) have been shown to represent concepts as directions in the latent space of hidden activations. However, for any given human-interpretable concept, how can we find its direction in the latent space? We present a technique called linear relational concepts (LRC) for finding concept direc...
http://arxiv.org/pdf/2311.08968v1
2023-11-15T14:01:41
2023-11-15T14:01:41
{ "title": [ -0.014846802, -0.0138549805, -0.023529053, -0.06713867, -0.014595032, 0.0046157837, -0.07281494, -0.007369995, 0.0063209534, 0.022064209, -0.008346558, 0.0015087128, 0.0015535355, -0.016479492, -0.040222168, 0.005622864, -0.02168274, ...
http://arxiv.org/abs/2311.08966v1
Improving Large-scale Deep Biasing with Phoneme Features and Text-only Data in Streaming Transducer
Jin Qiu, Lu Huang, Boyu Li, Jun Zhang, Lu Lu, Zejun Ma
cs.CL, cs.SD, eess.AS
Deep biasing for the Transducer can improve the recognition performance of rare words or contextual entities, which is essential in practical applications, especially for streaming Automatic Speech Recognition (ASR). However, deep biasing with large-scale rare words remains challenging, as the performance drops signifi...
http://arxiv.org/pdf/2311.08966v1
2023-11-15T13:53:28
2023-11-15T13:53:28
{ "title": [ 0.0067977905, 0.0060195923, -0.04623413, -0.05117798, 0.010848999, -0.023239136, -0.015899658, -0.0017490387, 0.030761719, 0.038909912, -0.044647217, -0.00046658516, 0.023330688, -0.010322571, -0.033599854, 0.008460999, -0.0060310364, ...
http://arxiv.org/abs/2311.08941v1
Reasoning over Description Logic-based Contexts with Transformers
Angelos Poulis, Eleni Tsalapati, Manolis Koubarakis
cs.CL, cs.AI
One way that the current state of the art measures the reasoning ability of transformer-based models is by evaluating accuracy in downstream tasks like logical question answering or proof generation over synthetic contexts expressed in natural language. However, most of the contexts used are in practice very simple; in...
http://arxiv.org/pdf/2311.08941v1
2023-11-15T13:23:24
2023-11-15T13:23:24
{ "title": [ -0.00020420551, -0.007949829, -0.042663574, -0.10998535, 0.03164673, 0.013168335, -0.049987793, 0.0039253235, -0.0011138916, 0.02331543, -0.05697632, 0.025222778, 0.00374794, -0.00447464, -0.003047943, -0.0018424988, -0.009239197, 0....
http://arxiv.org/abs/2311.08921v1
Self-Improving for Zero-Shot Named Entity Recognition with Large Language Models
Tingyu Xie, Qi Li, Yan Zhang, Zuozhu Liu, Hongwei Wang
cs.CL
Exploring the application of powerful large language models (LLMs) on the fundamental named entity recognition (NER) task has drawn much attention recently. This work aims to investigate the possibilities of pushing the boundary of zero-shot NER with LLM via a training-free self-improving strategy. We propose a self-im...
http://arxiv.org/pdf/2311.08921v1
2023-11-15T12:47:52
2023-11-15T12:47:52
{ "title": [ -0.003189087, -0.0019273758, -0.044006348, -0.070373535, 0.017333984, -0.008407593, -0.03390503, 0.018692017, 0.018829346, 0.025848389, -0.033325195, 0.012588501, -0.0023021698, 0.000012457371, -0.038330078, 0.012748718, 0.0020141602, ...
http://arxiv.org/abs/2311.08896v1
HELLaMA: LLaMA-based Table to Text Generation by Highlighting the Important Evidence
Junyi Bian, Xiaolei Qin, Wuhe Zou, Mengzuo Huang, Weidong Zhang
cs.CL
Large models have demonstrated significant progress across various domains, particularly in tasks related to text generation. In the domain of Table to Text, many Large Language Model (LLM)-based methods currently resort to modifying prompts to invoke public APIs, incurring potential costs and information leaks. With t...
http://arxiv.org/pdf/2311.08896v1
2023-11-15T12:02:52
2023-11-15T12:02:52
{ "title": [ -0.03652954, -0.0027332306, -0.05279541, -0.04019165, -0.01473999, 0.004432678, -0.054107666, -0.017074585, -0.0078125, 0.052947998, -0.026901245, 0.04272461, 0.022705078, -0.005882263, -0.02217102, 0.014801025, -0.010894775, 0.04275...
http://arxiv.org/abs/2311.08894v1
Combining Transfer Learning with In-context Learning using Blackbox LLMs for Zero-shot Knowledge Base Question Answering
Mayur Patidar, Avinash Singh, Riya Sawhney, Indrajit Bhattacharya, Mausam
cs.CL, cs.AI
We address the zero-shot transfer learning setting for the knowledge base question answering (KBQA) problem, where a large volume of labeled training data is available for the source domain, but no such labeled examples are available for the target domain. Transfer learning for KBQA makes use of large volumes of unlabe...
http://arxiv.org/pdf/2311.08894v1
2023-11-15T11:56:56
2023-11-15T11:56:56
{ "title": [ -0.0015354156, 0.005264282, -0.049346924, -0.06097412, -0.0048217773, -0.016540527, -0.028427124, 0.015060425, 0.021377563, 0.038635254, -0.045043945, 0.030349731, -0.017700195, 0.003993988, -0.035736084, -0.009529114, 0.010627747, 0...
http://arxiv.org/abs/2311.08890v1
Large Language Models are legal but they are not: Making the case for a powerful LegalLLM
Thanmay Jayakumar, Fauzan Farooqui, Luqman Farooqui
cs.CL
Realizing the recent advances in Natural Language Processing (NLP) to the legal sector poses challenging problems such as extremely long sequence lengths, specialized vocabulary that is usually only understood by legal professionals, and high amounts of data imbalance. The recent surge of Large Language Models (LLMs) h...
http://arxiv.org/pdf/2311.08890v1
2023-11-15T11:50:10
2023-11-15T11:50:10
{ "title": [ -0.018829346, -0.012451172, -0.044128418, -0.06210327, -0.008399963, -0.017990112, -0.03552246, -0.0004968643, -0.027816772, 0.022018433, -0.039093018, 0.011566162, -0.008407593, -0.022109985, -0.02394104, -0.018875122, -0.0022201538, ...
http://arxiv.org/abs/2311.08886v1
CLIMB: Curriculum Learning for Infant-inspired Model Building
Richard Diehl Martinez, Zebulon Goriely, Hope McGovern, Christopher Davis, Andrew Caines, Paula Buttery, Lisa Beinborn
cs.CL
We describe our team's contribution to the STRICT-SMALL track of the BabyLM Challenge. The challenge requires training a language model from scratch using only a relatively small training dataset of ten million words. We experiment with three variants of cognitively-motivated curriculum learning and analyze their effec...
http://arxiv.org/pdf/2311.08886v1
2023-11-15T11:48:16
2023-11-15T11:48:16
{ "title": [ -0.0725708, -0.01473999, -0.06262207, -0.019897461, -0.019989014, -0.00894165, -0.03366089, -0.0014429092, -0.028427124, -0.0059127808, -0.026138306, 0.031402588, -0.04611206, -0.017150879, 0.00932312, -0.02998352, -0.0033130646, 0.0...
http://arxiv.org/abs/2311.08883v1
Enabling Large Language Models to Learn from Rules
Wenkai Yang, Yankai Lin, Jie Zhou, Jirong Wen
cs.CL
Large language models (LLMs) have shown incredible performance in completing various real-world tasks. The current knowledge learning paradigm of LLMs is mainly based on learning from examples, in which LLMs learn the internal rule implicitly from a certain number of supervised examples. However, the learning paradigm ...
http://arxiv.org/pdf/2311.08883v1
2023-11-15T11:42:41
2023-11-15T11:42:41
{ "title": [ -0.018096924, -0.020599365, -0.049743652, -0.078552246, 0.0007586479, 0.004272461, -0.056243896, 0.028320312, 0.001329422, 0.013946533, -0.043792725, 0.0011720657, -0.008872986, -0.008369446, -0.029846191, -0.0073432922, -0.009712219, ...
http://arxiv.org/abs/2311.08877v1
Llamas Know What GPTs Don't Show: Surrogate Models for Confidence Estimation
Vaishnavi Shrivastava, Percy Liang, Ananya Kumar
cs.CL, cs.LG
To maintain user trust, large language models (LLMs) should signal low confidence on examples where they are incorrect, instead of misleading the user. The standard approach of estimating confidence is to use the softmax probabilities of these models, but as of November 2023, state-of-the-art LLMs such as GPT-4 and Cla...
http://arxiv.org/pdf/2311.08877v1
2023-11-15T11:27:44
2023-11-15T11:27:44
{ "title": [ -0.032409668, -0.0047340393, -0.011436462, -0.0055656433, -0.021774292, -0.0017642975, -0.033294678, -0.06573486, -0.033447266, 0.00957489, -0.021087646, -0.004463196, -0.04534912, -0.05493164, -0.020233154, -0.03060913, -0.0004506111, ...
http://arxiv.org/abs/2311.08849v1
OFA: A Framework of Initializing Unseen Subword Embeddings for Efficient Large-scale Multilingual Continued Pretraining
Yihong Liu, Peiqin Lin, Mingyang Wang, Hinrich Schütze
cs.CL
Pretraining multilingual language models from scratch requires considerable computational resources and substantial training data. Therefore, a more efficient method is to adapt existing pretrained language models (PLMs) to new languages via vocabulary extension and continued pretraining. However, this method usually r...
http://arxiv.org/pdf/2311.08849v1
2023-11-15T10:40:45
2023-11-15T10:40:45
{ "title": [ 0.006980896, -0.008666992, 0.017044067, -0.07086182, -0.020401001, 0.02230835, -0.05783081, -0.014915466, -0.0048561096, 0.027053833, -0.02909851, 0.029327393, 0.028930664, -0.017364502, 0.0024471283, 0.011268616, -0.01550293, 0.0472...
http://arxiv.org/abs/2311.08844v1
Violet: A Vision-Language Model for Arabic Image Captioning with Gemini Decoder
Abdelrahman Mohamed, Fakhraddin Alwajih, El Moatez Billah Nagoudi, Alcides Alcoba Inciarte, Muhammad Abdul-Mageed
cs.CV, cs.CL
Although image captioning has a vast array of applications, it has not reached its full potential in languages other than English. Arabic, for instance, although the native language of more than 400 million people, remains largely underrepresented in this area. This is due to the lack of labeled data and powerful Arabi...
http://arxiv.org/pdf/2311.08844v1
2023-11-15T10:34:14
2023-11-15T10:34:14
{ "title": [ -0.02494812, -0.026824951, -0.0143966675, -0.048187256, -0.026031494, -0.00687027, -0.03765869, -0.00008583069, 0.019989014, 0.05001831, -0.0016746521, -0.011688232, 0.001499176, -0.029937744, -0.075683594, 0.0006957054, 0.013389587, ...
http://arxiv.org/abs/2311.08838v1
Disinformation Capabilities of Large Language Models
Ivan Vykopal, Matúš Pikuliak, Ivan Srba, Robert Moro, Dominik Macko, Maria Bielikova
cs.CL
Automated disinformation generation is often listed as one of the risks of large language models (LLMs). The theoretical ability to flood the information space with disinformation content might have dramatic consequences for democratic societies around the world. This paper presents a comprehensive study of the disinfo...
http://arxiv.org/pdf/2311.08838v1
2023-11-15T10:25:30
2023-11-15T10:25:30
{ "title": [ 0.0049057007, 0.0031471252, -0.03515625, -0.06652832, 0.0073661804, -0.010719299, -0.022125244, -0.0036029816, -0.0055160522, 0.023910522, -0.018066406, -0.0034694672, 0.016189575, -0.028442383, -0.01789856, 0.010032654, -0.026672363, ...
http://arxiv.org/abs/2311.08836v1
Evaluating Gender Bias in the Translation of Gender-Neutral Languages into English
Spencer Rarrick, Ranjita Naik, Sundar Poudel, Vishal Chowdhary
cs.CL, cs.AI
Machine Translation (MT) continues to improve in quality and adoption, yet the inadvertent perpetuation of gender bias remains a significant concern. Despite numerous studies into gender bias in translations from gender-neutral languages such as Turkish into more strongly gendered languages like English, there are no b...
http://arxiv.org/pdf/2311.08836v1
2023-11-15T10:25:14
2023-11-15T10:25:14
{ "title": [ -0.053253174, 0.002571106, -0.045013428, 0.00522995, -0.015930176, -0.028121948, -0.0038070679, -0.014099121, -0.025009155, 0.06451416, -0.013160706, 0.0028457642, 0.04498291, -0.01940918, -0.037902832, 0.003791809, -0.028656006, 0.0...
http://arxiv.org/abs/2311.08817v1
MAP's not dead yet: Uncovering true language model modes by conditioning away degeneracy
Davis Yoshida, Kartik Goyal, Kevin Gimpel
cs.CL, cs.AI, cs.LG
It has been widely observed that exact or approximate MAP (mode-seeking) decoding from natural language generation (NLG) models consistently leads to degenerate outputs (Stahlberg and Byrne, 2019, Holtzman et al., 2019). This has generally been attributed to either a fundamental inadequacy of modes in models or weaknes...
http://arxiv.org/pdf/2311.08817v1
2023-11-15T09:38:53
2023-11-15T09:38:53
{ "title": [ -0.012008667, -0.00667572, -0.021224976, -0.05014038, 0.004306793, 0.0014686584, -0.03540039, -0.005619049, -0.00381279, 0.023345947, 0.0045204163, -0.0051574707, 0.006832123, -0.031951904, -0.04449463, 0.013679504, -0.012420654, 0.0...
http://arxiv.org/abs/2311.08803v1
StrategyLLM: Large Language Models as Strategy Generators, Executors, Optimizers, and Evaluators for Problem Solving
Chang Gao, Haiyun Jiang, Deng Cai, Shuming Shi, Wai Lam
cs.CL
Most existing chain-of-thought (CoT) prompting methods suffer from the issues of generalizability and consistency, as they often rely on instance-specific solutions that may not be applicable to other cases and lack task-level consistency in their reasoning steps. To address these limitations, we propose a comprehensiv...
http://arxiv.org/pdf/2311.08803v1
2023-11-15T09:18:09
2023-11-15T09:18:09
{ "title": [ -0.027709961, -0.035949707, -0.015426636, -0.03543091, -0.049804688, 0.014823914, -0.06500244, 0.0060653687, -0.032592773, 0.025848389, 0.0104522705, 0.005466461, -0.0034542084, -0.039978027, 0.0061302185, -0.0102005005, -0.018051147, ...
http://arxiv.org/abs/2311.08793v1
German FinBERT: A German Pre-trained Language Model
Moritz Scherrmann
cs.CL, stat.ML
This study presents German FinBERT, a novel pre-trained German language model tailored for financial textual data. The model is trained through a comprehensive pre-training process, leveraging a substantial corpus comprising financial reports, ad-hoc announcements and news related to German companies. The corpus size i...
http://arxiv.org/pdf/2311.08793v1
2023-11-15T09:07:29
2023-11-15T09:07:29
{ "title": [ -0.031799316, -0.0018348694, -0.036621094, -0.034484863, -0.0115356445, 0.0042762756, -0.052825928, -0.016647339, -0.013687134, 0.044403076, 0.016723633, -0.006034851, 0.050567627, -0.048187256, -0.021575928, 0.027008057, -0.020004272, ...
http://arxiv.org/abs/2311.08788v1
X-Eval: Generalizable Multi-aspect Text Evaluation via Augmented Instruction Tuning with Auxiliary Evaluation Aspects
Minqian Liu, Ying Shen, Zhiyang Xu, Yixin Cao, Eunah Cho, Vaibhav Kumar, Reza Ghanadan, Lifu Huang
cs.CL, cs.AI, cs.LG
Natural Language Generation (NLG) typically involves evaluating the generated text in various aspects (e.g., consistency and naturalness) to obtain a comprehensive assessment. However, multi-aspect evaluation remains challenging as it may require the evaluator to generalize to any given evaluation aspect even if it's a...
http://arxiv.org/pdf/2311.08788v1
2023-11-15T09:01:55
2023-11-15T09:01:55
{ "title": [ -0.038726807, -0.0034122467, -0.027679443, -0.07409668, -0.0058021545, 0.01979065, -0.07086182, -0.0017967224, -0.0006656647, 0.03656006, -0.00096940994, 0.010360718, -0.009864807, -0.01525116, -0.051513672, 0.0034618378, -0.03567505, ...
http://arxiv.org/abs/2311.08768v1
Three Conjectures on Unexpectedeness
Giovanni Sileno, Jean-Louis Dessalles
cs.AI, cs.CL, cs.IT, cs.SY, eess.SY, math.IT
Unexpectedness is a central concept in Simplicity Theory, a theory of cognition relating various inferential processes to the computation of Kolmogorov complexities, rather than probabilities. Its predictive power has been confirmed by several experiments with human subjects, yet its theoretical basis remains largely u...
http://arxiv.org/pdf/2311.08768v1
2023-11-15T08:24:41
2023-11-15T08:24:41
{ "title": [ -0.00015223026, 0.03237915, -0.028747559, -0.013557434, -0.017547607, -0.018814087, -0.043945312, -0.038238525, 0.0075187683, 0.034118652, -0.02268982, -0.0018596649, -0.0014772415, -0.009101868, -0.026931763, -0.006084442, 0.0211792, ...
http://arxiv.org/abs/2311.08756v1
Accelerating Toeplitz Neural Network with Constant-time Inference Complexity
Zhen Qin, Yiran Zhong
cs.CL
Toeplitz Neural Networks (TNNs) have exhibited outstanding performance in various sequence modeling tasks. They outperform commonly used Transformer-based models while benefiting from log-linear space-time complexities. On the other hand, State Space Models (SSMs) achieve lower performance than TNNs in language modelin...
http://arxiv.org/pdf/2311.08756v1
2023-11-15T07:50:57
2023-11-15T07:50:57
{ "title": [ -0.020568848, -0.00030994415, -0.02885437, -0.037384033, 0.011520386, -0.017364502, 0.00008249283, -0.012756348, 0.012458801, 0.023101807, -0.020401001, -0.0076904297, -0.011703491, -0.027191162, -0.041290283, -0.001791954, 0.013534546, ...
http://arxiv.org/abs/2311.08734v1
Thread of Thought Unraveling Chaotic Contexts
Yucheng Zhou, Xiubo Geng, Tao Shen, Chongyang Tao, Guodong Long, Jian-Guang Lou, Jianbing Shen
cs.CL
Large Language Models (LLMs) have ushered in a transformative era in the field of natural language processing, excelling in tasks related to text comprehension and generation. Nevertheless, they encounter difficulties when confronted with chaotic contexts (e.g., distractors rather than long irrelevant context), leading...
http://arxiv.org/pdf/2311.08734v1
2023-11-15T06:54:44
2023-11-15T06:54:44
{ "title": [ 0.046325684, -0.012802124, 0.029251099, 0.009429932, -0.045166016, -0.010848999, 0.019210815, -0.044067383, 0.00819397, 0.02192688, -0.05429077, 0.01146698, 0.027740479, -0.021697998, 0.01789856, 0.004737854, 0.04309082, 0.0052833557...
http://arxiv.org/abs/2311.08732v1
Enhancing Emergency Decision-making with Knowledge Graphs and Large Language Models
Minze Chen, Zhenxiang Tao, Weitong Tang, Tingxin Qin, Rui Yang, Chunli Zhu
cs.CL
Emergency management urgently requires comprehensive knowledge while having a high possibility to go beyond individuals' cognitive scope. Therefore, artificial intelligence(AI) supported decision-making under that circumstance is of vital importance. Recent emerging large language models (LLM) provide a new direction f...
http://arxiv.org/pdf/2311.08732v1
2023-11-15T06:48:50
2023-11-15T06:48:50
{ "title": [ -0.013465881, 0.0028266907, -0.06549072, -0.07293701, 0.012969971, -0.0044937134, -0.0574646, 0.0095825195, -0.02218628, 0.026275635, -0.027404785, 0.0028457642, -0.0057258606, -0.015090942, -0.017074585, -0.020553589, -0.0023555756, ...
http://arxiv.org/abs/2311.08726v1
Uncertainty Estimation on Sequential Labeling via Uncertainty Transmission
Jianfeng He, Linlin Yu, Shuo Lei, Chang-Tien Lu, Feng Chen
cs.CL
Sequential labeling is a task predicting labels for each token in a sequence, such as Named Entity Recognition (NER). NER tasks aim to extract entities and predict their labels given a text, which is important in information extraction. Although previous works have shown great progress in improving NER performance, unc...
http://arxiv.org/pdf/2311.08726v1
2023-11-15T06:36:29
2023-11-15T06:36:29
{ "title": [ -0.0524292, 0.0017156601, -0.048828125, -0.06604004, -0.020248413, 0.00737381, -0.07550049, -0.015510559, 0.003250122, -0.013053894, -0.010673523, 0.008010864, -0.044067383, -0.057128906, -0.0022888184, -0.01272583, 0.019546509, 0.01...
http://arxiv.org/abs/2311.08724v1
Method for Text Entity Linking in Power Distribution Scheduling Oriented to Power Distribution Network Knowledge Graph
Xiang Li, Che Wang, Bing Li, Hao Chen, Sizhe Li
cs.CL, cs.LG
The proposed method for linking entities in power distribution dispatch texts to a power distribution network knowledge graph is based on a deep understanding of these networks. This method leverages the unique features of entities in both the power distribution network's knowledge graph and the dispatch texts, focusin...
http://arxiv.org/pdf/2311.08724v1
2023-11-15T06:35:01
2023-11-15T06:35:01
{ "title": [ -0.008262634, 0.0024929047, -0.04257202, -0.010894775, -0.0025997162, -0.023635864, 0.007701874, 0.003232956, -0.044067383, 0.041778564, -0.012878418, 0.022079468, 0.03363037, -0.051605225, 0.052368164, -0.009902954, 0.0020713806, 0....
http://arxiv.org/abs/2311.08723v1
Token Prediction as Implicit Classification to Identify LLM-Generated Text
Yutian Chen, Hao Kang, Vivian Zhai, Liangze Li, Rita Singh, Bhiksha Raj
cs.CL
This paper introduces a novel approach for identifying the possible large language models (LLMs) involved in text generation. Instead of adding an additional classification layer to a base LM, we reframe the classification task as a next-token prediction task and directly fine-tune the base LM to perform it. We utilize...
http://arxiv.org/pdf/2311.08723v1
2023-11-15T06:33:52
2023-11-15T06:33:52
{ "title": [ -0.04067993, -0.0037631989, -0.04534912, -0.039794922, -0.034423828, 0.014343262, -0.06713867, -0.005493164, -0.017028809, 0.0056762695, -0.02444458, 0.027740479, -0.027038574, -0.020080566, -0.018539429, -0.020324707, -0.040771484, ...
http://arxiv.org/abs/2311.08719v1
Think-in-Memory: Recalling and Post-thinking Enable LLMs with Long-Term Memory
Lei Liu, Xiaoyan Yang, Yue Shen, Binbin Hu, Zhiqiang Zhang, Jinjie Gu, Guannan Zhang
cs.CL
Memory-augmented Large Language Models (LLMs) have demonstrated remarkable performance in long-term human-machine interactions, which basically relies on iterative recalling and reasoning of history to generate high-quality responses. However, such repeated recall-reason steps easily produce biased thoughts, \textit{i....
http://arxiv.org/pdf/2311.08719v1
2023-11-15T06:08:35
2023-11-15T06:08:35
{ "title": [ 0.019622803, -0.005771637, -0.025177002, -0.032440186, -0.020736694, -0.033325195, -0.022445679, -0.007293701, -0.009429932, 0.016342163, -0.020858765, 0.022354126, -0.015220642, -0.046417236, 0.0021209717, -0.022567749, 0.0011415482, ...
http://arxiv.org/abs/2311.08718v1
Decomposing Uncertainty for Large Language Models through Input Clarification Ensembling
Bairu Hou, Yujian Liu, Kaizhi Qian, Jacob Andreas, Shiyu Chang, Yang Zhang
cs.CL
Uncertainty decomposition refers to the task of decomposing the total uncertainty of a model into data (aleatoric) uncertainty, resulting from the inherent complexity or ambiguity of the data, and model (epistemic) uncertainty, resulting from the lack of knowledge in the model. Performing uncertainty decomposition for ...
http://arxiv.org/pdf/2311.08718v1
2023-11-15T05:58:35
2023-11-15T05:58:35
{ "title": [ -0.047668457, -0.015090942, -0.04916382, -0.0670166, -0.018066406, 0.008377075, -0.09118652, 0.0044021606, 0.00843811, 0.006160736, -0.01600647, 0.012741089, -0.014076233, -0.018478394, -0.03378296, 0.0022640228, -0.011184692, 0.0266...
http://arxiv.org/abs/2311.08711v1
PLUG: Leveraging Pivot Language in Cross-Lingual Instruction Tuning
Zhihan Zhang, Dong-Ho Lee, Yuwei Fang, Wenhao Yu, Mengzhao Jia, Meng Jiang, Francesco Barbieri
cs.CL
Instruction tuning has remarkably advanced large language models (LLMs) in understanding and responding to diverse human instructions. Despite the success in high-resource languages, its application in lower-resource ones faces challenges due to the imbalanced foundational abilities of LLMs across different languages, ...
http://arxiv.org/pdf/2311.08711v1
2023-11-15T05:28:07
2023-11-15T05:28:07
{ "title": [ -0.052246094, -0.033935547, -0.011627197, -0.061676025, -0.039489746, 0.0184021, -0.076538086, 0.007293701, -0.01902771, 0.036010742, -0.018463135, 0.015098572, -0.0030384064, -0.020202637, -0.010032654, -0.0062828064, -0.00497818, 0...
http://arxiv.org/abs/2311.08705v1
Evaluating Robustness of Dialogue Summarization Models in the Presence of Naturally Occurring Variations
Ankita Gupta, Chulaka Gunasekara, Hui Wan, Jatin Ganhotra, Sachindra Joshi, Marina Danilevsky
cs.CL
Dialogue summarization task involves summarizing long conversations while preserving the most salient information. Real-life dialogues often involve naturally occurring variations (e.g., repetitions, hesitations) and existing dialogue summarization models suffer from performance drop on such conversations. In this stud...
http://arxiv.org/pdf/2311.08705v1
2023-11-15T05:11:43
2023-11-15T05:11:43
{ "title": [ -0.0032291412, 0.013549805, -0.033935547, -0.06689453, 0.0031547546, -0.0004556179, -0.0769043, -0.01889038, 0.007080078, 0.040740967, 0.0020713806, 0.019042969, -0.026428223, -0.0035076141, -0.07946777, 0.0056877136, -0.017578125, 0...
http://arxiv.org/abs/2311.08704v1
Can Large Language Models Follow Concept Annotation Guidelines? A Case Study on Scientific and Financial Domains
Marcio Fonseca, Shay B. Cohen
cs.CL, cs.AI
Although large language models (LLMs) exhibit remarkable capacity to leverage in-context demonstrations, it is still unclear to what extent they can learn new concepts or facts from ground-truth labels. To address this question, we examine the capacity of instruction-tuned LLMs to follow in-context concept guidelines f...
http://arxiv.org/pdf/2311.08704v1
2023-11-15T05:11:26
2023-11-15T05:11:26
{ "title": [ -0.021255493, -0.010627747, -0.019134521, -0.04940796, -0.026016235, 0.0062217712, -0.064086914, -0.013290405, 0.02406311, 0.040893555, -0.007293701, -0.009559631, 0.021240234, -0.00920105, -0.059814453, 0.018981934, -0.024246216, 0....
http://arxiv.org/abs/2311.08702v1
Debate Helps Supervise Unreliable Experts
Julian Michael, Salsabila Mahdi, David Rein, Jackson Petty, Julien Dirani, Vishakh Padmakumar, Samuel R. Bowman
cs.AI, cs.CL, I.2.0
As AI systems are used to answer more difficult questions and potentially help create new knowledge, judging the truthfulness of their outputs becomes more difficult and more important. How can we supervise unreliable experts, which have access to the truth but may not accurately report it, to give answers that are sys...
http://arxiv.org/pdf/2311.08702v1
2023-11-15T05:05:40
2023-11-15T05:05:40
{ "title": [ 0.08746338, -0.011459351, 0.03845215, -0.058166504, -0.025405884, -0.0041923523, 0.0039749146, -0.012710571, -0.060791016, 0.037384033, 0.021499634, -0.0107421875, -0.030715942, -0.04135132, 0.02720642, -0.02571106, 0.018569946, 0.00...
http://arxiv.org/abs/2311.08695v1
Attribute Diversity Determines the Systematicity Gap in VQA
Ian Berlot-Attwell, A. Michael Carrell, Kumar Krishna Agrawal, Yash Sharma, Naomi Saphra
cs.LG, cs.CL, cs.CV
The degree to which neural networks can generalize to new combinations of familiar concepts, and the conditions under which they are able to do so, has long been an open question. In this work, we study the systematicity gap in visual question answering: the performance difference between reasoning on previously seen a...
http://arxiv.org/pdf/2311.08695v1
2023-11-15T04:50:30
2023-11-15T04:50:30
{ "title": [ -0.017547607, 0.04168701, -0.087402344, -0.020217896, -0.0006070137, -0.054473877, -0.032592773, -0.03286743, -0.012023926, 0.024353027, 0.014640808, -0.053344727, 0.007041931, -0.07122803, -0.0340271, -0.014259338, 0.0039634705, 0.0...
http://arxiv.org/abs/2311.08692v1
Routing to the Expert: Efficient Reward-guided Ensemble of Large Language Models
Keming Lu, Hongyi Yuan, Runji Lin, Junyang Lin, Zheng Yuan, Chang Zhou, Jingren Zhou
cs.CL, cs.LG
The complementary potential of Large Language Models (LLM) assumes off-the-shelf LLMs have heterogeneous expertise in a wide range of domains and tasks so that an ensemble of LLMs can achieve consistently better performance. Existing ensemble methods for LLMs mainly focus on reward model ranking of outputs, leading to ...
http://arxiv.org/pdf/2311.08692v1
2023-11-15T04:40:43
2023-11-15T04:40:43
{ "title": [ -0.011482239, -0.013023376, -0.03668213, -0.07409668, 0.008560181, 0.005016327, -0.055267334, 0.019454956, 0.003479004, 0.031143188, -0.013893127, 0.0146484375, -0.0038642883, -0.0078048706, -0.047668457, 0.015129089, -0.010391235, 0...
http://arxiv.org/abs/2311.08687v1
An Eye on Clinical BERT: Investigating Language Model Generalization for Diabetic Eye Disease Phenotyping
Keith Harrigian, Tina Tang, Anthony Gonzales, Cindy X. Cai, Mark Dredze
cs.CL, cs.AI, cs.LG
Diabetic eye disease is a major cause of blindness worldwide. The ability to monitor relevant clinical trajectories and detect lapses in care is critical to managing the disease and preventing blindness. Alas, much of the information necessary to support these goals is found only in the free text of the electronic medi...
http://arxiv.org/pdf/2311.08687v1
2023-11-15T04:30:20
2023-11-15T04:30:20
{ "title": [ -0.01612854, -0.032165527, -0.03741455, -0.059051514, -0.030151367, 0.0075531006, -0.06304932, 0.015327454, -0.0019369125, 0.049957275, 0.011146545, 0.0016593933, 0.015213013, -0.024505615, -0.0513916, 0.00409317, -0.010009766, 0.012...
http://arxiv.org/abs/2311.08685v1
Safer-Instruct: Aligning Language Models with Automated Preference Data
Taiwei Shi, Kai Chen, Jieyu Zhao
cs.CL, cs.AI
Reinforcement Learning from Human Feedback (RLHF) is a vital strategy for enhancing model safety in language models. However, annotating preference data for RLHF is a resource-intensive and creativity-demanding process, while automatic generation methods face limitations in data diversity and quality. In response, we p...
http://arxiv.org/pdf/2311.08685v1
2023-11-15T04:22:22
2023-11-15T04:22:22
{ "title": [ -0.026138306, -0.01146698, -0.049926758, -0.059143066, -0.0033073425, 0.0008459091, -0.048858643, 0.0019569397, 0.0013427734, 0.024246216, -0.03543091, 0.005596161, 0.0016117096, -0.009101868, -0.035949707, -0.00068569183, -0.022628784, ...
http://arxiv.org/abs/2311.08669v1
Understanding Calibration for Multilingual Question Answering Models
Yahan Yang, Soham Dan, Dan Roth, Insup Lee
cs.CL, cs.LG
Multilingual pre-trained language models are incredibly effective at Question Answering (QA), a core task in Natural Language Understanding, achieving high accuracies on several multilingual benchmarks. However, little is known about how well they are calibrated. In this paper, we study the calibration properties of se...
http://arxiv.org/pdf/2311.08669v1
2023-11-15T03:29:02
2023-11-15T03:29:02
{ "title": [ -0.040130615, -0.004322052, -0.02507019, -0.045166016, -0.0140686035, 0.0124435425, -0.060180664, -0.014770508, 0.010009766, 0.04147339, -0.029144287, 0.0010213852, 0.0012140274, -0.00035309792, -0.07128906, 0.0009832382, -0.007270813, ...
http://arxiv.org/abs/2311.08666v1
It Takes Two to Negotiate: Modeling Social Exchange in Online Multiplayer Games
Kokil Jaidka, Hansin Ahuja, Lynnette Ng
cs.CL, cs.GT, cs.LG
Online games are dynamic environments where players interact with each other, which offers a rich setting for understanding how players negotiate their way through the game to an ultimate victory. This work studies online player interactions during the turn-based strategy game, Diplomacy. We annotated a dataset of over...
http://arxiv.org/pdf/2311.08666v1
2023-11-15T03:21:04
2023-11-15T03:21:04
{ "title": [ 0.006828308, 0.015403748, 0.0039024353, -0.007621765, 0.014595032, -0.026046753, 0.03173828, -0.04650879, -0.02432251, 0.023620605, 0.0076026917, 0.0052719116, -0.039154053, -0.030838013, -0.04257202, -0.016479492, -0.027297974, 0.01...
http://arxiv.org/abs/2311.08662v1
Multi-Set Inoculation: Assessing Model Robustness Across Multiple Challenge Sets
Vatsal Gupta, Pranshu Pandya, Tushar Kataria, Vivek Gupta, Dan Roth
cs.CL, cs.AI, cs.IR
Language models, given their black-box nature, often exhibit sensitivity to input perturbations, leading to trust issues due to hallucinations. To bolster trust, it's essential to understand these models' failure modes and devise strategies to enhance their performance. In this study, we propose a framework to study th...
http://arxiv.org/pdf/2311.08662v1
2023-11-15T02:59:10
2023-11-15T02:59:10
{ "title": [ 0.007133484, 0.01713562, -0.052886963, -0.055267334, 0.023773193, -0.014442444, -0.055236816, -0.041412354, -0.010627747, 0.015655518, -0.010719299, 0.020111084, -0.0357666, -0.03152466, -0.035980225, -0.011161804, -0.010505676, 0.01...
http://arxiv.org/abs/2311.08648v1
Explore Spurious Correlations at the Concept Level in Language Models for Text Classification
Yuhang Zhou, Paiheng Xu, Xiaoyu Liu, Bang An, Wei Ai, Furong Huang
cs.CL, cs.AI
Language models (LMs) have gained great achievement in various NLP tasks for both fine-tuning and in-context learning (ICL) methods. Despite its outstanding performance, evidence shows that spurious correlations caused by imbalanced label distributions in training data (or exemplars in ICL) lead to robustness issues. H...
http://arxiv.org/pdf/2311.08648v1
2023-11-15T01:58:54
2023-11-15T01:58:54
{ "title": [ 0.002368927, -0.0031661987, -0.035858154, -0.0075187683, -0.03479004, -0.025436401, 0.0059814453, -0.01133728, -0.00072050095, 0.03643799, -0.041900635, 0.010925293, 0.012252808, -0.0060310364, -0.027862549, -0.024108887, -0.02658081, ...
http://arxiv.org/abs/2311.08640v1
Multistage Collaborative Knowledge Distillation from Large Language Models
Jiachen Zhao, Wenlong Zhao, Andrew Drozdov, Benjamin Rozonoyer, Md Arafat Sultan, Jay-Yoon Lee, Mohit Iyyer, Andrew McCallum
cs.CL, cs.LG
We study semi-supervised sequence prediction tasks where labeled data are too scarce to effectively finetune a model and at the same time few-shot prompting of a large language model (LLM) has suboptimal performance. This happens when a task, such as parsing, is expensive to annotate and also unfamiliar to a pretrained...
http://arxiv.org/pdf/2311.08640v1
2023-11-15T01:28:28
2023-11-15T01:28:28
{ "title": [ -0.041503906, -0.014816284, -0.08129883, -0.09857178, 0.0046806335, 0.002565384, -0.10168457, 0.026626587, 0.0385437, 0.025009155, -0.026473999, -0.020980835, -0.018051147, -0.01108551, -0.05307007, -0.017044067, -0.005420685, 0.0199...
http://arxiv.org/abs/2311.08637v1
Formal Proofs as Structured Explanations: Proposing Several Tasks on Explainable Natural Language Inference
Lasha Abzianidze
cs.CL, 68T50, I.2.7
In this position paper, we propose a way of exploiting formal proofs to put forward several explainable natural language inference (NLI) tasks. The formal proofs will be produced by a reliable and high-performing logic-based NLI system. Taking advantage of the in-depth information available in the generated formal proo...
http://arxiv.org/pdf/2311.08637v1
2023-11-15T01:24:09
2023-11-15T01:24:09
{ "title": [ -0.020370483, -0.021499634, -0.026687622, -0.11480713, 0.0005288124, 0.023605347, -0.06640625, 0.025817871, -0.020523071, -0.023880005, -0.030654907, 0.016860962, -0.032470703, -0.016464233, 0.0126953125, -0.016036987, -0.016937256, ...
http://arxiv.org/abs/2311.08623v1
DEED: Dynamic Early Exit on Decoder for Accelerating Encoder-Decoder Transformer Models
Peng Tang, Pengkai Zhu, Tian Li, Srikar Appalaraju, Vijay Mahadevan, R. Manmatha
cs.CV, cs.CL, cs.LG
Encoder-decoder transformer models have achieved great success on various vision-language (VL) tasks, but they suffer from high inference latency. Typically, the decoder takes up most of the latency because of the auto-regressive decoding. To accelerate the inference, we propose an approach of performing Dynamic Early ...
http://arxiv.org/pdf/2311.08623v1
2023-11-15T01:01:02
2023-11-15T01:01:02
{ "title": [ -0.027542114, 0.012008667, -0.04257202, 0.021575928, -0.02142334, -0.032043457, 0.005619049, -0.029251099, 0.0181427, 0.059509277, -0.00054216385, -0.0027370453, 0.06036377, -0.043304443, -0.016098022, 0.03439331, 0.030166626, 0.0440...
http://arxiv.org/abs/2311.08622v1
Multiple-Question Multiple-Answer Text-VQA
Peng Tang, Srikar Appalaraju, R. Manmatha, Yusheng Xie, Vijay Mahadevan
cs.CV, cs.CL, cs.LG
We present Multiple-Question Multiple-Answer (MQMA), a novel approach to do text-VQA in encoder-decoder transformer models. The text-VQA task requires a model to answer a question by understanding multi-modal content: text (typically from OCR) and an associated image. To the best of our knowledge, almost all previous a...
http://arxiv.org/pdf/2311.08622v1
2023-11-15T01:00:02
2023-11-15T01:00:02
{ "title": [ -0.010169983, -0.014259338, -0.05545044, -0.037750244, -0.034606934, -0.031585693, -0.0071868896, 0.029129028, -0.031433105, 0.039886475, 0.002380371, 0.0022563934, -0.017333984, -0.040985107, -0.0021018982, -0.0446167, 0.01763916, -...
http://arxiv.org/abs/2311.08620v1
Toucan: Token-Aware Character Level Language Modeling
William Fleshman, Benjamin Van Durme
cs.CL, cs.LG
Character-level language models obviate the need for separately trained tokenizers, but efficiency suffers from longer sequence lengths. Learning to combine character representations into tokens has made training these models more efficient, but they still require decoding characters individually. We propose Toucan, an...
http://arxiv.org/pdf/2311.08620v1
2023-11-15T00:57:51
2023-11-15T00:57:51
{ "title": [ -0.0317688, -0.021835327, -0.032196045, -0.06933594, -0.011146545, 0.019638062, -0.05847168, 0.022918701, 0.021270752, 0.031585693, -0.0003399849, 0.005104065, 0.02861023, -0.022262573, -0.03326416, 0.026916504, -0.018981934, 0.02903...
http://arxiv.org/abs/2311.08614v1
XplainLLM: A QA Explanation Dataset for Understanding LLM Decision-Making
Zichen Chen, Jianda Chen, Mitali Gaidhani, Ambuj Singh, Misha Sra
cs.CL, cs.AI
Large Language Models (LLMs) have recently made impressive strides in natural language understanding tasks. Despite their remarkable performance, understanding their decision-making process remains a big challenge. In this paper, we look into bringing some transparency to this process by introducing a new explanation d...
http://arxiv.org/pdf/2311.08614v1
2023-11-15T00:34:28
2023-11-15T00:34:28
{ "title": [ -0.06335449, -0.025604248, -0.05899048, -0.026062012, -0.041625977, -0.008796692, -0.06939697, 0.008651733, -0.0034332275, -0.0051651, -0.039031982, 0.012428284, -0.019226074, -0.026428223, 0.0042762756, -0.01828003, -0.034179688, 0....
http://arxiv.org/abs/2311.08607v1
Towards Generalizable SER: Soft Labeling and Data Augmentation for Modeling Temporal Emotion Shifts in Large-Scale Multilingual Speech
Mohamed Osman, Tamer Nadeem, Ghada Khoriba
cs.CL, cs.LG, eess.AS
Recognizing emotions in spoken communication is crucial for advanced human-machine interaction. Current emotion detection methodologies often display biases when applied cross-corpus. To address this, our study amalgamates 16 diverse datasets, resulting in 375 hours of data across languages like English, Chinese, and J...
http://arxiv.org/pdf/2311.08607v1
2023-11-15T00:09:21
2023-11-15T00:09:21
{ "title": [ 0.013893127, -0.004257202, -0.040924072, -0.09429932, -0.0042762756, -0.005176544, -0.061279297, 0.037597656, 0.013198853, 0.039886475, -0.018753052, -0.0025787354, 0.018951416, -0.024414062, -0.040893555, -0.0015001297, 0.0064964294, ...
http://arxiv.org/abs/2311.08605v1
Navigating the Ocean of Biases: Political Bias Attribution in Language Models via Causal Structures
David F. Jenny, Yann Billeter, Mrinmaya Sachan, Bernhard Schölkopf, Zhijing Jin
cs.CL, cs.AI, cs.CY, cs.SI
The rapid advancement of Large Language Models (LLMs) has sparked intense debate regarding their ability to perceive and interpret complex socio-political landscapes. In this study, we undertake an exploration of decision-making processes and inherent biases within LLMs, exemplified by ChatGPT, specifically contextuali...
http://arxiv.org/pdf/2311.08605v1
2023-11-15T00:02:25
2023-11-15T00:02:25
{ "title": [ -0.007331848, -0.009567261, -0.027862549, -0.052612305, -0.014953613, -0.010688782, -0.04107666, -0.018600464, -0.022827148, 0.017623901, -0.0178833, -0.004486084, -0.0067596436, -0.018463135, -0.025222778, -0.011352539, -0.03729248, ...
http://arxiv.org/abs/2311.08598v1
DALA: A Distribution-Aware LoRA-Based Adversarial Attack against Pre-trained Language Models
Yibo Wang, Xiangjue Dong, James Caverlee, Philip S. Yu
cs.CL
Pre-trained language models (PLMs) that achieve success in applications are susceptible to adversarial attack methods that are capable of generating adversarial examples with minor perturbations. Although recent attack methods can achieve a relatively high attack success rate (ASR), our observation shows that the gener...
http://arxiv.org/pdf/2311.08598v1
2023-11-14T23:43:47
2023-11-14T23:43:47
{ "title": [ -0.033233643, -0.014099121, -0.035491943, -0.037384033, -0.010986328, 0.0013599396, -0.024795532, -0.0013389587, -0.043518066, 0.036193848, 0.03427124, 0.007522583, -0.0101623535, -0.038635254, -0.020492554, 0.0018053055, -0.011001587, ...
http://arxiv.org/abs/2311.08596v1
Are You Sure? Challenging LLMs Leads to Performance Drops in The FlipFlop Experiment
Philippe Laban, Lidiya Murakhovs'ka, Caiming Xiong, Chien-Sheng Wu
cs.CL
The interactive nature of Large Language Models (LLMs) theoretically allows models to refine and improve their answers, yet systematic analysis of the multi-turn behavior of LLMs remains limited. In this paper, we propose the FlipFlop experiment: in the first round of the conversation, an LLM responds to a prompt conta...
http://arxiv.org/pdf/2311.08596v1
2023-11-14T23:40:22
2023-11-14T23:40:22
{ "title": [ -0.035736084, 0.010429382, -0.041656494, 0.037231445, -0.015167236, -0.020812988, -0.014480591, -0.026260376, -0.03463745, 0.01852417, -0.01222229, 0.0012521744, 0.019943237, -0.041900635, 0.026290894, -0.035095215, 0.0032672882, 0.0...
http://arxiv.org/abs/2311.08593v1
ACID: Abstractive, Content-Based IDs for Document Retrieval with Language Models
Haoxin Li, Phillip Keung, Daniel Cheng, Jungo Kasai, Noah A. Smith
cs.CL, cs.IR
Generative retrieval (Wang et al., 2022; Tay et al., 2022) is a new approach for end-to-end document retrieval that directly generates document identifiers given an input query. Techniques for designing effective, high-quality document IDs remain largely unexplored. We introduce ACID, in which each document's ID is com...
http://arxiv.org/pdf/2311.08593v1
2023-11-14T23:28:36
2023-11-14T23:28:36
{ "title": [ -0.018157959, -0.017501831, -0.028030396, -0.049804688, 0.006187439, 0.0022010803, -0.0385437, -0.0050468445, 0.018676758, 0.038513184, -0.02670288, 0.0037193298, 0.01876831, -0.0034255981, -0.05166626, 0.0052375793, -0.017425537, 0....
http://arxiv.org/abs/2311.08592v1
AART: AI-Assisted Red-Teaming with Diverse Data Generation for New LLM-powered Applications
Bhaktipriya Radharapu, Kevin Robinson, Lora Aroyo, Preethi Lahoti
cs.SE, cs.AI, cs.CL
Adversarial testing of large language models (LLMs) is crucial for their safe and responsible deployment. We introduce a novel approach for automated generation of adversarial evaluation datasets to test the safety of LLM generations on new downstream applications. We call it AI-assisted Red-Teaming (AART) - an automat...
http://arxiv.org/pdf/2311.08592v1
2023-11-14T23:28:23
2023-11-14T23:28:23
{ "title": [ -0.07128906, -0.012496948, -0.062042236, -0.050567627, -0.048095703, 0.017150879, -0.09320068, -0.0044174194, -0.018066406, 0.015625, -0.005443573, 0.026412964, -0.012588501, -0.048706055, 0.015731812, -0.009361267, -0.006385803, -0....
http://arxiv.org/abs/2311.08590v1
PEMA: Plug-in External Memory Adaptation for Language Models
HyunJin Kim, Young Jin Kim, JinYeong Bak
cs.CL
Pre-trained language models (PLMs) have demonstrated impressive performance across various downstream NLP tasks. Nevertheless, the resource requirements of pre-training large language models in terms of memory and training compute pose significant challenges. Furthermore, due to the substantial resources required, many...
http://arxiv.org/pdf/2311.08590v1
2023-11-14T23:20:51
2023-11-14T23:20:51
{ "title": [ -0.038391113, -0.02809143, -0.032958984, -0.06756592, -0.013679504, 0.01687622, -0.066589355, 0.016586304, 0.0051994324, 0.04071045, -0.009223938, 0.0018539429, 0.009300232, -0.015045166, -0.048980713, 0.023803711, -0.0064582825, 0.0...