Search

Transformers dominate NLP and IR; but their inference inefficiencies and challenges in extrapolating to longer contexts have sparked interest in alternative model architectures. Among these, state space models (SSMs) like Mamba offer promising advantages, particularly $O(1)$ time complexity in inference. Despite their potential, SSMs' effectiveness at text reranking -- a task requiring fine-grained query-document interaction and long-context understanding -- remains underexplored. This study benchmarks SSM-based architectures (specifically, Mamba-1 and Mamba-2) against transformer-based models across various scales, architectures, and pre-training objectives, focusing on performance and efficiency in text reranking tasks. We find that (1) Mamba architectures achieve competitive text ranking performance, comparable to transformer-based models of similar size; (2) they are less efficient in training and inference compared to transformers with flash attention; and (3) Mamba-2 outperforms Mamba-1 in both performance and efficiency. These results underscore the potential of state space models as a transformer alternative and highlight areas for improvement in future IR applications., Comment: The first two authors contributed equally, order decided randomly

Increasingly, model compression techniques enable large language models (LLMs) to be deployed in real-world applications. As a result of this momentum towards local deployment, compressed LLMs will interact with a large population. Prior work on compression typically prioritize preserving perplexity, which is directly analogous to training loss. The impact of compression method on other critical aspects of model behavior\, -- \,particularly safety\, -- \,requires systematic assessment. To this end, we investigate the impact of model compression along four dimensions: (1) degeneration harm, i.e., bias and toxicity in generation; (2) representational harm, i.e., biases in discriminative tasks; (3) dialect bias; and(4) language modeling and downstream task performance. We examine a wide spectrum of LLM compression techniques, including unstructured pruning, semi-structured pruning, and quantization. Our analysis reveals that compression can lead to unexpected consequences. Although compression may unintentionally alleviate LLMs' degeneration harm, it can still exacerbate representational harm. Furthermore, increasing compression produces a divergent impact on different protected groups. Finally, different compression methods have drastically different safety impacts: for example, quantization mostly preserves bias while pruning degrades quickly. Our findings underscore the importance of integrating safety assessments into the development of compressed LLMs to ensure their reliability across real-world applications.\footnote{Our implementation and results are available here: \url{https://github.com/zhichaoxu-shufe/Beyond-Perplexity-Compression-Safety-Eval}}, Comment: Findings of EMNLP 2024

The $\texttt{Hi-COLA}$ code is an efficient dark matter simulation suite that flexibly handles the Horndeski family of modified gravity models. In this work we extend the scope of $\texttt{Hi-COLA}$ to accommodate Horndeski theories with K-mouflage screening, allowing for the computation of matter power spectra in the non-linear regime in these models. We explore the boost of the dark matter power spectrum relative to GR-$\Lambda$CDM in K-mouflage gravity, and also discuss how large-scale structure computations change between the Einstein and Jordan frames. A dissection of the relative contributions of the modified background, linear growth, fifth force, and the conformal factor (a new inclusion to $\texttt{Hi-COLA}$) to the boost factor is presented. The ability of $\texttt{Hi-COLA}$ to run with general Horndeski models and multiple screening mechanisms makes it an ideal tool for testing gravity with upcoming galaxy survey data., Comment: 31 pages, 10 figures, 1 table. Comments welcome. The Hi-COLA code can be found at https://github.com/Hi-COLACode/Hi-COLA

Testing gravity and the concordance model of cosmology, $\Lambda$CDM, at large scales is a key goal of this decade's largest galaxy surveys. Here we present a comparative study of dark matter power spectrum predictions from different numerical codes in the context of three popular theories of gravity that induce scale-independent modifications to the linear growth of structure: nDGP, Cubic Galileon and K-mouflage. In particular, we compare the predictions from full $N$-body simulations, two $N$-body codes with approximate time integration schemes, a parametrised modified $N$-body implementation and the analytic halo model reaction approach. We find the modification to the $\Lambda$CDM spectrum is in $2\%$ agreement for $z\leq1$ and $k\leq 1~h/{\rm Mpc}$ over all gravitational models and codes, in accordance with many previous studies, indicating these modelling approaches are robust enough to be used in forthcoming survey analyses under appropriate scale cuts. We further make public the new code implementations presented, specifically the halo model reaction K-mouflage implementation and the relativistic Cubic Galileon implementation., Comment: 20 pages, 4 figures, 4 tables

The increasing size of transformer-based models in NLP makes the question of compressing them important. In this work, we present a comprehensive analysis of factorization based model compression techniques. Specifically, we focus on comparing straightforward low-rank factorization against the recently introduced Monarch factorization, which exhibits impressive performance preservation on the GLUE benchmark. To mitigate stability issues associated with low-rank factorization of the matrices in pre-trained transformers, we introduce a staged factorization approach wherein layers are factorized one by one instead of being factorized simultaneously. Through this strategy we significantly enhance the stability and reliability of the compression process. Further, we introduce a simple block-wise low-rank factorization method, which has a close relationship to Monarch factorization. Our experiments lead to the surprising conclusion that straightforward low-rank factorization consistently outperforms Monarch factorization across both different compression ratios and six different text classification tasks.

To completely understand a document, the use of textual information is not enough. Understanding visual cues, such as layouts and charts, is also required. While the current state-of-the-art approaches for document understanding (both OCR-based and OCR-free) work well, a thorough analysis of their capabilities and limitations has not yet been performed. Therefore, in this work, we addresses the limitation of current VisualQA models when applied to charts and plots. To investigate shortcomings of the state-of-the-art models, we conduct a comprehensive behavioral analysis, using ChartQA as a case study. Our findings indicate that existing models particularly underperform in answering questions related to the chart's structural and visual context, as well as numerical information. To address these issues, we propose three simple pre-training tasks that enforce the existing model in terms of both structural-visual knowledge, as well as its understanding of numerical questions. We evaluate our pre-trained model (called MatCha-v2) on three chart datasets - both extractive and abstractive question datasets - and observe that it achieves an average improvement of 1.7% over the baseline model., Comment: Accepted in the BlackboxNLP workshop at EMNLP 2024

Do larger and more performant models resolve NLP's longstanding robustness issues? We investigate this question using over 20 models of different sizes spanning different architectural choices and pretraining objectives. We conduct evaluations using (a) out-of-domain and challenge test sets, (b) behavioral testing with CheckLists, (c) contrast sets, and (d) adversarial inputs. Our analysis reveals that not all out-of-domain tests provide insight into robustness. Evaluating with CheckLists and contrast sets shows significant gaps in model performance; merely scaling models does not make them adequately robust. Finally, we point out that current approaches for adversarial evaluations of models are themselves problematic: they can be easily thwarted, and in their current forms, do not represent a sufficiently deep probe of model robustness. We conclude that not only is the question of robustness in NLP as yet unresolved, but even some of the approaches to measure robustness need to be reassessed., Comment: To appear at NAACL 24 - main conference. The code is available at: https://github.com/utahnlp/scaling_robustness/

Identifying intents from dialogue utterances forms an integral component of task-oriented dialogue systems. Intent-related tasks are typically formulated either as a classification task, where the utterances are classified into predefined categories or as a clustering task when new and previously unknown intent categories need to be discovered from these utterances. Further, the intent classification may be modeled in a multiclass (MC) or multilabel (ML) setup. While typically these tasks are modeled as separate tasks, we propose IntenDD, a unified approach leveraging a shared utterance encoding backbone. IntenDD uses an entirely unsupervised contrastive learning strategy for representation learning, where pseudo-labels for the unlabeled utterances are generated based on their lexical features. Additionally, we introduce a two-step post-processing setup for the classification tasks using modified adsorption. Here, first, the residuals in the training data are propagated followed by smoothing the labels both modeled in a transductive setting. Through extensive evaluations on various benchmark datasets, we find that our approach consistently outperforms competitive baselines across all three tasks. On average, IntenDD reports percentage improvements of 2.32%, 1.26%, and 1.52% in their respective metrics for few-shot MC, few-shot ML, and the intent discovery tasks respectively., Comment: EMNLP 2023 Findings

Clean-label (CL) attack is a form of data poisoning attack where an adversary modifies only the textual input of the training data, without requiring access to the labeling function. CL attacks are relatively unexplored in NLP, as compared to label flipping (LF) attacks, where the latter additionally requires access to the labeling function as well. While CL attacks are more resilient to data sanitization and manual relabeling methods than LF attacks, they often demand as high as ten times the poisoning budget than LF attacks. In this work, we first introduce an Adversarial Clean Label attack which can adversarially perturb in-class training examples for poisoning the training set. We then show that an adversary can significantly bring down the data requirements for a CL attack, using the aforementioned approach, to as low as 20% of the data otherwise required. We then systematically benchmark and analyze a number of defense methods, for both LF and CL attacks, some previously employed solely for LF attacks in the textual domain and others adapted from computer vision. We find that text-specific defenses greatly vary in their effectiveness depending on their properties., Comment: RepL4NLP 2023 at ACL 2023

Can language models transform inputs to protect text classifiers against adversarial attacks? In this work, we present ATINTER, a model that intercepts and learns to rewrite adversarial inputs to make them non-adversarial for a downstream text classifier. Our experiments on four datasets and five attack mechanisms reveal that ATINTER is effective at providing better adversarial robustness than existing defense approaches, without compromising task accuracy. For example, on sentiment classification using the SST-2 dataset, our method improves the adversarial accuracy over the best existing defense approach by more than 4% with a smaller decrease in task accuracy (0.5% vs 2.5%). Moreover, we show that ATINTER generalizes across multiple downstream tasks and classifiers without having to explicitly retrain it for those settings. Specifically, we find that when ATINTER is trained to remove adversarial perturbations for the sentiment classification task on the SST-2 dataset, it even transfers to a semantically different task of news classification (on AGNews) and improves the adversarial robustness by more than 10%., Comment: Accepted to ACL 2023

We release S\={a}mayik, a dataset of around 53,000 parallel English-Sanskrit sentences, written in contemporary prose. Sanskrit is a classical language still in sustenance and has a rich documented heritage. However, due to the limited availability of digitized content, it still remains a low-resource language. Existing Sanskrit corpora, whether monolingual or bilingual, have predominantly focused on poetry and offer limited coverage of contemporary written materials. S\={a}mayik is curated from a diverse range of domains, including language instruction material, textual teaching pedagogy, and online tutorials, among others. It stands out as a unique resource that specifically caters to the contemporary usage of Sanskrit, with a primary emphasis on prose writing. Translation models trained on our dataset demonstrate statistically significant improvements when translating out-of-domain contemporary corpora, outperforming models trained on older classical-era poetry datasets. Finally, we also release benchmark models by adapting four multilingual pre-trained models, three of them have not been previously exposed to Sanskrit for translating between English and Sanskrit while one of them is multi-lingual pre-trained translation model including English and Sanskrit. The dataset and source code is present at https://github.com/ayushbits/saamayik., Comment: LREC-COLING, 2024

We introduce $\texttt{Hi-COLA}$, a code designed to run fast, approximate $\textit{N}$-body simulations of non-linear structure formation in reduced Horndeski gravity. Given an input Lagrangian, $\texttt{Hi-COLA}$ dynamically constructs the appropriate field equations and consistently solves for the cosmological background, linear growth, and screened fifth force of that theory. Hence $\texttt{Hi-COLA}$ is a general, adaptable, and useful tool that allows the mildly non-linear regime of many Horndeski theories to be investigated for the first time, at low computational cost. In this work, we first describe the screening approximations and simulation setup of $\texttt{Hi-COLA}$ for theories with Vainshtein screening. We validate the code against traditional $\textit{N}$-body simulations for cubic Galileon gravity, finding $2.5\%$ agreement up to $k_{\rm max}=1.2~h/{\rm Mpc}$. To demonstrate the flexibility of $\texttt{Hi-COLA}$, we additionally run the first simulations of an extended shift-symmetric gravity theory. We use the consistency and modularity of $\texttt{Hi-COLA}$ to dissect how the modified background, linear growth, and screened fifth force all contribute to departures from $\Lambda$CDM in the non-linear matter power spectrum. $\texttt{Hi-COLA}$ can be found at https://github.com/Hi-COLACode/Hi-COLA ., Comment: 51 pages, 20 figures, 5 tables, comments welcome. Updated to match the version published in JCAP

Symbolic knowledge can provide crucial inductive bias for training neural models, especially in low data regimes. A successful strategy for incorporating such knowledge involves relaxing logical statements into sub-differentiable losses for optimization. In this paper, we study the question of how best to relax logical expressions that represent labeled examples and knowledge about a problem; we focus on sub-differentiable t-norm relaxations of logic. We present theoretical and empirical criteria for characterizing which relaxation would perform best in various scenarios. In our theoretical study driven by the goal of preserving tautologies, the Lukasiewicz t-norm performs best. However, in our empirical analysis on the text chunking and digit recognition tasks, the product t-norm achieves best predictive performance. We analyze this apparent discrepancy, and conclude with a list of best practices for defining loss functions via logic., Comment: IJCAI 2021 paper (Extended Version)

In this work, we introduce X-FACT: the largest publicly available multilingual dataset for factual verification of naturally existing real-world claims. The dataset contains short statements in 25 languages and is labeled for veracity by expert fact-checkers. The dataset includes a multilingual evaluation benchmark that measures both out-of-domain generalization, and zero-shot capabilities of the multilingual models. Using state-of-the-art multilingual transformer-based models, we develop several automated fact-checking models that, along with textual claims, make use of additional metadata and evidence from news stories retrieved using a search engine. Empirically, our best model attains an F-score of around 40%, suggesting that our dataset is a challenging benchmark for evaluation of multilingual fact-checking models., Comment: ACL 2021; For data and code, see https://github.com/utahnlp/x-fact/

Neural dependency parsing has achieved remarkable performance for many domains and languages. The bottleneck of massive labeled data limits the effectiveness of these approaches for low resource languages. In this work, we focus on dependency parsing for morphological rich languages (MRLs) in a low-resource setting. Although morphological information is essential for the dependency parsing task, the morphological disambiguation and lack of powerful analyzers pose challenges to get this information for MRLs. To address these challenges, we propose simple auxiliary tasks for pretraining. We perform experiments on 10 MRLs in low-resource settings to measure the efficacy of our proposed pretraining method and observe an average absolute gain of 2 points (UAS) and 3.6 points (LAS). Code and data available at: https://github.com/jivnesh/LCM, Comment: 6 pages, The work is accepted at EACL-SRW, 2021, Kyiv, Ukraine Typos corrected in Section 3.2

In this paper, we study the response of large models from the BERT family to incoherent inputs that should confuse any model that claims to understand natural language. We define simple heuristics to construct such examples. Our experiments show that state-of-the-art models consistently fail to recognize them as ill-formed, and instead produce high confidence predictions on them. As a consequence of this phenomenon, models trained on sentences with randomly permuted word order perform close to state-of-the-art models. To alleviate these issues, we show that if models are explicitly trained to recognize invalid inputs, they can be robust to such attacks without a drop in performance., Comment: Accepted at AAAI 2021, Camera Ready Version

Neural sequence labelling approaches have achieved state of the art results in morphological tagging. We evaluate the efficacy of four standard sequence labelling models on Sanskrit, a morphologically rich, fusional Indian language. As its label space can theoretically contain more than 40,000 labels, systems that explicitly model the internal structure of a label are more suited for the task, because of their ability to generalise to labels not seen during training. We find that although some neural models perform better than others, one of the common causes for error for all of these models is mispredictions due to syncretism., Comment: Accepted to SIGMORPHON Workshop at ACL 2020

Data-driven approaches for dependency parsing have been of great interest in Natural Language Processing for the past couple of decades. However, Sanskrit still lacks a robust purely data-driven dependency parser, probably with an exception to Krishna (2019). This can primarily be attributed to the lack of availability of task-specific labelled data and the morphologically rich nature of the language. In this work, we evaluate four different data-driven machine learning models, originally proposed for different languages, and compare their performances on Sanskrit data. We experiment with 2 graph based and 2 transition based parsers. We compare the performance of each of the models in a low-resource setting, with 1,500 sentences for training. Further, since our focus is on the learning power of each of the models, we do not incorporate any Sanskrit specific features explicitly into the models, and rather use the default settings in each of the paper for obtaining the feature functions. In this work, we analyse the performance of the parsers using both an in-domain and an out-of-domain test dataset. We also investigate the impact of word ordering in which the sentences are provided as input to these systems, by parsing verses and their corresponding prose order (anvaya) sentences., Comment: submitted to WSC 2021