@inproceedings{10204341,

title = {Positive-Augmented Contrastive Learning for Image and Video Captioning Evaluation},

author = {Sara Sarto and Manuele Barraco and Marcella Cornia and Lorenzo Baraldi and Rita Cucchiara},

doi = {10.1109/CVPR52729.2023.00668},

issn = {2575-7075},

year  = {2023},

date = {2023-06-01},

booktitle = {2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},

pages = {6914-6924},

abstract = {The CLIP model has been recently proven to be very effective for a variety of cross-modal tasks, including the evaluation of captions generated from vision-and-language architectures. In this paper, we propose a new recipe for a contrastive-based evaluation metric for image captioning, namely Positive-Augmented Contrastive learning Score (PAC-S), that in a novel way unifies the learning of a contrastive visual-semantic space with the addition of generated images and text on curated data. Experiments spanning several datasets demonstrate that our new metric achieves the highest correlation with human judgments on both images and videos, outperforming existing referencebased metrics like CIDEr and SPICE and reference-free metrics like CLIP-Score. Finally, we test the system-level correlation of the proposed metric when considering popular image captioning approaches, and assess the impact of employing different cross-modal features. Our source code and trained models are publicly available at: https://github.com/aimagelab/pacscore.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{pmlr-v206-cella23a,

title = {Multi-task Representation Learning with Stochastic Linear Bandits},

author = {Leonardo Cella and Karim Lounici and Grégoire Pacreau and Massimiliano Pontil},

editor = {Francisco Ruiz and Jennifer Dy and Jan-Willem Meent},

url = {https://proceedings.mlr.press/v206/cella23a.html},

year  = {2023},

date = {2023-04-01},

booktitle = {Proceedings of The 26th International Conference on Artificial Intelligence and Statistics},

volume = {206},

pages = {4822–4847},

publisher = {PMLR},

series = {Proceedings of Machine Learning Research},

abstract = {We study the problem of transfer-learning in the setting of stochastic linear contextual bandit tasks. We consider that a low dimensional linear representation is shared across the tasks, and study the benefit of learning the tasks jointly. Following recent results to design Lasso stochastic bandit policies, we propose an efficient greedy policy based on trace norm regularization. It implicitly learns a low dimensional representation by encouraging the matrix formed by the task regression vectors to be of low rank. Unlike previous work in the literature, our policy does not need to know the rank of the underlying matrix, nor does it requires the covariance of the arms distribution to be invertible. We derive an upper bound on the multi-task regret of our policy, which is, up to logarithmic factors, of order $TsqrtrN+sqrtrNTd$, where $T$ is the number of tasks, $r$ the rank, $d$ the number of variables and $N$ the number of rounds per task. We show the benefit of our strategy over an independent task learning baseline, which has a worse regret of order $TsqrtdN$. We also argue that our policy is minimax optimal and, when $Tgeq d$, has a multi-task regret which is comparable to the regret of an oracle policy which knows the true underlying representation.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{pmlr-v206-qiu23a,

title = {Iterative Teaching by Data Hallucination},

author = {Zeju Qiu and Weiyang Liu and Tim Z. Xiao and Zhen Liu and Umang Bhatt and Yucen Luo and Adrian Weller and Bernhard Sch""olkopf},

editor = {Francisco Ruiz and Jennifer Dy and Jan-Willem Meent},

url = {https://proceedings.mlr.press/v206/qiu23a.html},

year  = {2023},

date = {2023-04-01},

booktitle = {Proceedings of The 26th International Conference on Artificial Intelligence and Statistics},

volume = {206},

pages = {9892–9913},

publisher = {PMLR},

series = {Proceedings of Machine Learning Research},

abstract = {We consider the problem of iterative machine teaching, where a teacher sequentially provides examples based on the status of a learner under a discrete input space (i.e., a pool of finite samples), which greatly limits the teacher’s capability. To address this issue, we study iterative teaching under a continuous input space where the input example (i.e., image) can be either generated by solving an optimization problem or drawn directly from a continuous distribution. Specifically, we propose data hallucination teaching (DHT) where the teacher can generate input data intelligently based on labels, the learner’s status and the target concept. We study a number of challenging teaching setups (e.g., linear/neural learners in omniscient and black-box settings). Extensive empirical results verify the effectiveness of DHT.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{pmlr-v206-quinzan23a,

title = {Fast Feature Selection with Fairness Constraints},

author = {Francesco Quinzan and Rajiv Khanna and Moshik Hershcovitch and Sarel Cohen and Daniel Waddington and Tobias Friedrich and Michael W. Mahoney},

editor = {Francisco Ruiz and Jennifer Dy and Jan-Willem Meent},

url = {https://proceedings.mlr.press/v206/quinzan23a.html},

year  = {2023},

date = {2023-04-01},

booktitle = {Proceedings of The 26th International Conference on Artificial Intelligence and Statistics},

volume = {206},

pages = {7800–7823},

publisher = {PMLR},

series = {Proceedings of Machine Learning Research},

abstract = {We study the fundamental problem of selecting optimal features for model construction. This problem is computationally challenging on large datasets, even with the use of greedy algorithm variants. To address this challenge, we extend the adaptive query model, recently proposed for the greedy forward selection for submodular functions, to the faster paradigm of Orthogonal Matching Pursuit for non-submodular functions. The proposed algorithm achieves exponentially fast parallel run time in the adaptive query model, scaling much better than prior work. Furthermore, our extension allows the use of downward-closed constraints, which can be used to encode certain fairness criteria into the feature selection process. We prove strong approximation guarantees for the algorithm based on standard assumptions. These guarantees are applicable to many parametric models, including Generalized Linear Models. Finally, we demonstrate empirically that the proposed algorithm competes favorably with state-of-the-art techniques for feature selection, on real-world and synthetic datasets.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{nokey,

title = {TAPS: Connecting Certified and Adversarial Training},

author = {Yuhao Mao and Mark Niklas Müller and Marc Fischer and Martin Vechev},

url = {https://arxiv.org/abs/2305.04574},

year  = {2023},

date = {2023-03-08},

publisher = {NeurIPS},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{kwiatkowska2023a,

title = {When to trust AI: advances and challenges for certification of neural networks},

author = {M Kwiatkowska and X Zhang},

year  = {2023},

date = {2023-01-01},

pages = {25-37},

publisher = {Polish Information Processing Society},

series = {Annals of Computer Science and Information Systems},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.5555/3618408.3619857,

title = {Trading-off payments and accuracy in online classification with paid stochastic experts},

author = {Dirk Van Der Hoeven and Ciara Pike-Burke and Hao Qiu and Nicolò Cesa-Bianchi},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the 40th International Conference on Machine Learning},

publisher = {JMLR.org},

address = {Honolulu, Hawaii, USA},

series = {ICML'23},

abstract = {We investigate online classification with paid stochastic experts. Here, before making their prediction, each expert must be paid. The amount that we pay each expert directly influences the accuracy of their prediction through some unknown Lipschitz ""productivity"" function. In each round, the learner must decide how much to pay each expert and then make a prediction. They incur a cost equal to a weighted sum of the prediction error and upfront payments for all experts. We introduce an online learning algorithm whose total cost after T rounds exceeds that of a predictor which knows the productivity of all experts in advance by at most O(K2(ln T)√T) where K is the number of experts. In order to achieve this result, we combine Lipschitz bandits and online classification with surrogate losses. These tools allow us to improve upon the bound of order T2/3 one would obtain in the standard Lipschitz bandit setting. Our algorithm is empirically evaluated on synthetic data.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.1609/aaai.v37i10.26392,

title = {Towards robust metrics for concept representation evaluation},

author = {Mateo Espinosa Zarlenga and Pietro Barbiero and Zohreh Shams and Dmitry Kazhdan and Umang Bhatt and Adrian Weller and Mateja Jamnik},

url = {https://doi.org/10.1609/aaai.v37i10.26392},

doi = {10.1609/aaai.v37i10.26392},

isbn = {978-1-57735-880-0},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the Thirty-Seventh AAAI Conference on Artificial Intelligence and Thirty-Fifth Conference on Innovative Applications of Artificial Intelligence and Thirteenth Symposium on Educational Advances in Artificial Intelligence},

publisher = {AAAI Press},

series = {AAAI'23/IAAI'23/EAAI'23},

abstract = {Recent work on interpretability has focused on concept-based explanations, where deep learning models are explained in terms of high-level units of information, referred to as concepts. Concept learning models, however, have been shown to be prone to encoding impurities in their representations, failing to fully capture meaningful features of their inputs. While concept learning lacks metrics to measure such phenomena, the field of disentanglement learning has explored the related notion of underlying factors of variation in the data, with plenty of metrics to measure the purity of such factors. In this paper, we show that such metrics are not appropriate for concept learning and propose novel metrics for evaluating the purity of concept representations in both approaches. We show the advantage of these metrics over existing ones and demonstrate their utility in evaluating the robustness of concept representations and interventions performed on them. In addition, we show their utility for benchmarking state-of-the-art methods from both families and find that, contrary to common assumptions, supervision alone may not be sufficient for pure concept representations.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.1609/aaai.v37i7.25986,

title = {Towards more robust interpretation via local gradient alignment},

author = {Sunghwan Joo and Seokhyeon Jeong and Juyeon Heo and Adrian Weller and Taesup Moon},

url = {https://doi.org/10.1609/aaai.v37i7.25986},

doi = {10.1609/aaai.v37i7.25986},

isbn = {978-1-57735-880-0},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the Thirty-Seventh AAAI Conference on Artificial Intelligence and Thirty-Fifth Conference on Innovative Applications of Artificial Intelligence and Thirteenth Symposium on Educational Advances in Artificial Intelligence},

publisher = {AAAI Press},

series = {AAAI'23/IAAI'23/EAAI'23},

abstract = {Neural network interpretation methods, particularly feature attribution methods, are known to be fragile with respect to adversarial input perturbations. To address this, several methods for enhancing the local smoothness of the gradient while training have been proposed for attaining robust feature attributions. However, the lack of considering the normalization of the attributions, which is essential in their visualizations, has been an obstacle to understanding and improving the robustness of feature attribution methods. In this paper, we provide new insights by taking such normalization into account. First, we show that for every non-negative homogeneous neural network, a naive ℓ2-robust criterion for gradients is not normalization invariant, which means that two functions with the same normalized gradient can have different values. Second, we formulate a normalization invariant cosine distance-based criterion and derive its upper bound, which gives insight for why simply minimizing the Hessian norm at the input, as has been done in previous work, is not sufficient for attaining robust feature attribution. Finally, we propose to combine both ℓ2 and cosine distance-based criteria as regularization terms to leverage the advantages of both in aligning the local gradient. As a result, we experimentally show that models trained with our method produce much more robust interpretations on CIFAR-10 and ImageNet-100 without significantly hurting the accuracy, compared to the recent baselines. To the best of our knowledge, this is the first work to verify the robustness of interpretation on a larger-scale dataset beyond CIFAR-10, thanks to the computational efficiency of our method.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.1145/3593013.3594099,

title = {The Progression of Disparities within the Criminal Justice System: Differential Enforcement and Risk Assessment Instruments},

author = {Miri Zilka and Riccardo Fogliato and Jiri Hron and Bradley Butcher and Carolyn Ashurst and Adrian Weller},

url = {https://doi.org/10.1145/3593013.3594099},

doi = {10.1145/3593013.3594099},

isbn = {9798400701924},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency},

pages = {1553–1569},

publisher = {Association for Computing Machinery},

address = {Chicago, IL, USA},

series = {FAccT '23},

abstract = {Algorithmic risk assessment instruments (RAIs) increasingly inform decision-making in criminal justice. RAIs largely rely on arrest records as a proxy for underlying crime. Problematically, the extent to which arrests reflect overall offending can vary with the person’s characteristics. We examine how the disconnect between crime and arrest rates impacts RAIs and their evaluation. Our main contribution is a method for quantifying this bias via estimation of the amount of unobserved offenses associated with particular demographics. These unobserved offenses are then used to augment real-world arrest records to create part real, part synthetic crime records. Using this data, we estimate that four currently deployed RAIs assign 0.5–2.8 percentage points higher risk scores to Black individuals than to White individuals with a similar arrest record, but the gap grows to 4.5–11.0 percentage points when we match on the semi-synthetic crime record. We conclude by discussing the potential risks around the use of RAIs, highlighting how they may exacerbate existing inequalities if the underlying disparities of the criminal justice system are not taken into account. In light of our findings, we provide recommendations to improve the development and evaluation of such tools.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.1609/aaai.v37i2.25328,

title = {Text-DIAE: a self-supervised degradation invariant autoencoder for text recognition and document enhancement},

author = {Mohamed Ali Souibgui and Sanket Biswas and Andres Mafla and Ali Furkan Biten and Alicia Fornés and Yousri Kessentini and Josep Lladós and Lluis Gomez and Dimosthenis Karatzas},

url = {https://doi.org/10.1609/aaai.v37i2.25328},

doi = {10.1609/aaai.v37i2.25328},

isbn = {978-1-57735-880-0},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the Thirty-Seventh AAAI Conference on Artificial Intelligence and Thirty-Fifth Conference on Innovative Applications of Artificial Intelligence and Thirteenth Symposium on Educational Advances in Artificial Intelligence},

publisher = {AAAI Press},

series = {AAAI'23/IAAI'23/EAAI'23},

abstract = {In this paper, we propose a Text-Degradation Invariant Auto Encoder (Text-DIAE), a self-supervised model designed to tackle two tasks, text recognition (handwritten or scene-text) and document image enhancement. We employ a transformer-based architecture that incorporates three pretext tasks as learning objectives to be optimized during pre-training without the usage of labelled data. Each of the pretext objectives is tailored for the final downstream tasks. We conduct several ablation experiments that confirm the design choice of the selected pretext tasks. Importantly, the proposed model does not exhibit limitations of previous state-of-the-art methods based on contrastive losses, while at the same time requiring substantially fewer data samples to converge. Finally, we demonstrate that our method surpasses the state-of-the-art in existing supervised and self-supervised settings in handwritten and scene text recognition and document image enhancement. Our code and trained models will be made publicly available.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.1609/aaai.v37i2.25285,

title = {Show, interpret and tell: entity-aware contextualised image captioning in wikipedia},

author = {Khanh Nguyen and Ali Furkan Biten and Andres Mafla and Lluis Gomez and Dimosthenis Karatzas},

url = {https://doi.org/10.1609/aaai.v37i2.25285},

doi = {10.1609/aaai.v37i2.25285},

isbn = {978-1-57735-880-0},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the Thirty-Seventh AAAI Conference on Artificial Intelligence and Thirty-Fifth Conference on Innovative Applications of Artificial Intelligence and Thirteenth Symposium on Educational Advances in Artificial Intelligence},

publisher = {AAAI Press},

series = {AAAI'23/IAAI'23/EAAI'23},

abstract = {Humans exploit prior knowledge to describe images, and are able to adapt their explanation to specific contextual information, even to the extent of inventing plausible explanations when contextual information and images do not match. In this work, we propose the novel task of captioning Wikipedia images by integrating contextual knowledge. Specifically, we produce models that jointly reason over Wikipedia articles, Wikimedia images and their associated descriptions to produce contextualized captions. Particularly, a similar Wikimedia image can be used to illustrate different articles, and the produced caption needs to be adapted to a specific context, therefore allowing us to explore the limits of a model to adjust captions to different contextual information. A particular challenging task in this domain is dealing with out-of-dictionary words and Named Entities. To address this, we propose a pre-training objective, Masked Named Entity Modeling (MNEM), and show that this pretext task yields an improvement compared to baseline models. Furthermore, we verify that a model pre-trained with the MNEM objective in Wikipedia generalizes well to a News Captioning dataset. Additionally, we define two different test splits according to the difficulty of the captioning task. We offer insights on the role and the importance of each modality and highlight the limitations of our model.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.1145/3605764.3623985,

title = {Not What You've Signed Up For: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection},

author = {Kai Greshake and Sahar Abdelnabi and Shailesh Mishra and Christoph Endres and Thorsten Holz and Mario Fritz},

url = {https://doi.org/10.1145/3605764.3623985},

doi = {10.1145/3605764.3623985},

isbn = {9798400702600},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the 16th ACM Workshop on Artificial Intelligence and Security},

pages = {79–90},

publisher = {Association for Computing Machinery},

address = {Copenhagen, Denmark},

series = {AISec '23},

abstract = {Large Language Models (LLMs) are increasingly being integrated into applications, with versatile functionalities that can be easily modulated via natural language prompts. So far, it was assumed that the user is directly prompting the LLM. But, what if it is not the user prompting? We show that LLM-Integrated Applications blur the line between data and instructions and reveal several new attack vectors, using Indirect Prompt Injection, that enable adversaries to remotely (i.e., without a direct interface) exploit LLM-integrated applications by strategically injecting prompts into data likely to be retrieved at inference time. We derive a comprehensive taxonomy from a computer security perspective to broadly investigate impacts and vulnerabilities, including data theft, worming, information ecosystem contamination, and other novel security risks. We then demonstrate the practical viability of our attacks against both real-world systems, such as Bing Chat and code-completion engines, and GPT-4 synthetic applications. We show how processing retrieved prompts can act as arbitrary code execution, manipulate the application's functionality, and control how and if other APIs are called. Despite the increasing reliance on LLMs, effective mitigations of these emerging threats are lacking. By raising awareness of these vulnerabilities, we aim to promote the safe and responsible deployment of these powerful models and the development of robust defenses that protect users from potential attacks.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{NEURIPS2023_15d15045,

title = {Multitask Learning with No Regret: from Improved Confidence Bounds to Active Learning},

author = {Pier Giuseppe Sessa and Pierre Laforgue and Nicolò Cesa-Bianchi and Andreas Krause},

editor = {A. Oh and T. Naumann and A. Globerson and K. Saenko and M. Hardt and S. Levine},

url = {https://proceedings.neurips.cc/paper_files/paper/2023/file/15d15045f93b44d933a260b249608d43-Paper-Conference.pdf},

year  = {2023},

date = {2023-01-01},

booktitle = {Advances in Neural Information Processing Systems},

volume = {36},

pages = {6770–6781},

publisher = {Curran Associates, Inc.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.5555/3666122.3666482,

title = {Multiplication-free transformer training via piecewise affine operations},

author = {Atli Kosson and Martin Jaggi},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the 37th International Conference on Neural Information Processing Systems},

publisher = {Curran Associates Inc.},

address = {New Orleans, LA, USA},

series = {NIPS '23},

abstract = {Multiplications are responsible for most of the computational cost involved in neural network training and inference. Recent research has thus looked for ways to reduce the cost associated with them. Inspired by Mogami (2020), we replace multiplication with a cheap piecewise affine approximation that is achieved by adding the bit representation of the floating point numbers together as integers. We show that transformers can be trained with the resulting modified matrix multiplications on both vision and language tasks with little to no performance impact, and without changes to the training hyperparameters. We further replace all non-linearities in the networks making them fully and jointly piecewise affine in both inputs and weights. Finally, we show that we can eliminate all multiplications in the entire training process, including operations in the forward pass, backward pass and optimizer update, demonstrating the first successful training of modern neural network architectures in a fully multiplication-free fashion.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.1145/3617694.3623249,

title = {Media Coverage of Predictive Policing: Bias, Police Engagement, and the Future of Transparency},

author = {Harry Camilleri and Carolyn Ashurst and Nithya Jaisankar and Adrian Weller and Miri Zilka},

url = {https://doi.org/10.1145/3617694.3623249},

doi = {10.1145/3617694.3623249},

isbn = {9798400703812},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the 3rd ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization},

publisher = {Association for Computing Machinery},

address = {Boston, MA, USA},

series = {EAAMO '23},

abstract = {The last decade has seen a wave of technological innovation by police forces. While this development has attracted significant scholarly attention, little is known about the accompanying media coverage, which can play a significant role in public understanding and perception of such technologies. In this article, we examine how the UK press has reported on two specific types of police technology, individual risk assessment, and predictive area mapping, in the 2012–2021 period. We find that (i) press coverage peaked in 2018, before dropping sharply, (ii) press reports have grown significantly more negative, (iii) concerns about bias have become predominant, and (iv) some evidence that the police have become more reticent in their dealings with the press. We hypothesise that this has inadvertently contributed to a reduction in transparency, which risks further erosion of efforts to identify and mitigate potential harms. In light of these findings, we propose potential paths towards improving transparency.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.1145/3581783.3612706,

title = {Let's ViCE! Mimicking Human Cognitive Behavior in Image Generation Evaluation},

author = {Federico Betti and Jacopo Staiano and Lorenzo Baraldi and Lorenzo Baraldi and Rita Cucchiara and Nicu Sebe},

url = {https://doi.org/10.1145/3581783.3612706},

doi = {10.1145/3581783.3612706},

isbn = {9798400701085},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the 31st ACM International Conference on Multimedia},

pages = {9306–9312},

publisher = {Association for Computing Machinery},

address = {Ottawa ON, Canada},

series = {MM '23},

abstract = {Research in Image Generation has recently made significant progress, particularly boosted by the introduction of VisionLanguage models which are able to produce high-quality visual content based on textual inputs. Despite ongoing advancements in terms of generation quality and realism, no methodical frameworks have been defined yet to quantitatively measure the quality of the generated content and the adherence with the prompted requests: so far, only humanbased evaluations have been adopted for quality satisfaction and for comparing different generative methods. We introduce a novel automated method for Visual Concept Evaluation (ViCE), i.e. to assess consistency between a generated/edited image and the corresponding prompt/instructions, with a process inspired by the human cognitive behaviour. ViCE combines the strengths of Large Language Models (LLMs) and Visual Question Answering (VQA) into a unified pipeline, aiming to replicate the human cognitive process in quality assessment. This method outlines visual concepts, formulates image-specific verification questions, utilizes the Q&A system to investigate the image, and scores the combined outcome. Although this brave new hypothesis of mimicking humans in the image evaluation process is in its preliminary assessment stage, results are promising and open the door to a new form of automatic evaluation which could have significant impact as the image generation or the image target editing tasks become more and more sophisticated.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{inproceedingsd,

title = {Improving Fairness via Intrinsic Plasticity in Echo State Networks},

author = {Andrea Ceni and Davide Bacciu and Valerio Caro and Claudio Gallicchio and Luca Oneto},

doi = {10.14428/esann/2023.ES2023-90},

year  = {2023},

date = {2023-01-01},

pages = {205-210},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.1145/3600211.3604692,

title = {Human Uncertainty in Concept-Based AI Systems},

author = {Katherine Maeve Collins and Matthew Barker and Mateo Espinosa Zarlenga and Naveen Raman and Umang Bhatt and Mateja Jamnik and Ilia Sucholutsky and Adrian Weller and Krishnamurthy Dvijotham},

url = {https://doi.org/10.1145/3600211.3604692},

doi = {10.1145/3600211.3604692},

isbn = {9798400702310},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the 2023 AAAI/ACM Conference on AI, Ethics, and Society},

pages = {869–889},

publisher = {Association for Computing Machinery},

address = {Montréal, QC, Canada},

series = {AIES '23},

abstract = {Placing a human in the loop may help abate the risks of deploying AI systems in safety-critical settings (e.g., a clinician working with a medical AI system). However, mitigating risks arising from human error and uncertainty within such human-AI interactions is an important and understudied issue. In this work, we study human uncertainty in the context of concept-based models, a family of AI systems that enable human feedback via concept interventions where an expert intervenes on human-interpretable concepts relevant to the task. Prior work in this space often assumes that humans are oracles who are always certain and correct. Yet, real-world decision-making by humans is prone to occasional mistakes and uncertainty. We study how existing concept-based models deal with uncertain interventions from humans using two novel datasets: UMNIST, a visual dataset with controlled simulated uncertainty based on the MNIST dataset, and CUB-S, a relabeling of the popular CUB concept dataset with rich, densely-annotated soft labels from humans. We show that training with uncertain concept labels may help mitigate weaknesses of concept-based systems when handling uncertain interventions. These results allow us to identify several open challenges, which we argue can be tackled through future multidisciplinary research on building interactive uncertainty-aware systems. To facilitate further research, we release a new elicitation platform, UElic, to collect uncertain feedback from humans in collaborative prediction tasks.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.1145/3593013.3594033,

title = {Harms from Increasingly Agentic Algorithmic Systems},

author = {Alan Chan and Rebecca Salganik and Alva Markelius and Chris Pang and Nitarshan Rajkumar and Dmitrii Krasheninnikov and Lauro Langosco and Zhonghao He and Yawen Duan and Micah Carroll and Michelle Lin and Alex Mayhew and Katherine Collins and Maryam Molamohammadi and John Burden and Wanru Zhao and Shalaleh Rismani and Konstantinos Voudouris and Umang Bhatt and Adrian Weller and David Krueger and Tegan Maharaj},

url = {https://doi.org/10.1145/3593013.3594033},

doi = {10.1145/3593013.3594033},

isbn = {9798400701924},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency},

pages = {651–666},

publisher = {Association for Computing Machinery},

address = {Chicago, IL, USA},

series = {FAccT '23},

abstract = {Research in Fairness, Accountability, Transparency, and Ethics (FATE)1 has established many sources and forms of algorithmic harm, in domains as diverse as health care, finance, policing, and recommendations. Much work remains to be done to mitigate the serious harms of these systems, particularly those disproportionately affecting marginalized communities. Despite these ongoing harms, new systems are being developed and deployed, typically without strong regulatory barriers, threatening the perpetuation of the same harms and the creation of novel ones. In response, the FATE community has emphasized the importance of anticipating harms, rather than just responding to them. Anticipation of harms is especially important given the rapid pace of developments in machine learning (ML). Our work focuses on the anticipation of harms from increasingly agentic systems. Rather than providing a definition of agency as a binary property, we identify 4 key characteristics which, particularly in combination, tend to increase the agency of a given algorithmic system: underspecification, directness of impact, goal-directedness, and long-term planning. We also discuss important harms which arise from increasing agency – notably, these include systemic and/or long-range impacts, often on marginalized or unconsidered stakeholders. We emphasize that recognizing agency of algorithmic systems does not absolve or shift the human responsibility for algorithmic harms. Rather, we use the term agency to highlight the increasingly evident fact that ML systems are not fully under human control. Our work explores increasingly agentic algorithmic systems in three parts. First, we explain the notion of an increase in agency for algorithmic systems in the context of diverse perspectives on agency across disciplines. Second, we argue for the need to anticipate harms from increasingly agentic systems. Third, we discuss important harms from increasingly agentic systems and ways forward for addressing them. We conclude by reflecting on implications of our work for anticipating algorithmic harms from emerging systems.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10341369,

title = {Generating Scenarios from High-Level Specifications for Object Rearrangement Tasks},

author = {Sanne Waveren and Christian Pek and Iolanda Leite and Jana Tumova and Danica Kragic},

doi = {10.1109/IROS55552.2023.10341369},

year  = {2023},

date = {2023-01-01},

booktitle = {2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},

pages = {11420-11427},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.1145/3627106.3627172,

title = {From Attachments to SEO: Click Here to Learn More about Clickbait PDFs!},

author = {Giada Stivala and Sahar Abdelnabi and Andrea Mengascini and Mariano Graziano and Mario Fritz and Giancarlo Pellegrino},

url = {https://doi.org/10.1145/3627106.3627172},

doi = {10.1145/3627106.3627172},

isbn = {9798400708862},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the 39th Annual Computer Security Applications Conference},

pages = {14–28},

publisher = {Association for Computing Machinery},

address = {Austin, TX, USA},

series = {ACSAC '23},

abstract = {Clickbait PDFs are PDF documents that do not embed malware but trick victims into visiting malicious web pages leading to attacks like password theft or drive-by download. While recent reports indicate a surge of clickbait PDFs, prior works have largely neglected this new threat, considering PDFs only as accessories of email phishing campaigns. This paper investigates the landscape of clickbait PDFs and presents the first systematic and comprehensive study of this phenomenon. Starting from a real-world dataset, we identify 44 clickbait PDF clusters via clustering and characterize them by looking at their volumetric, temporal, and visual features. Among these, we identify three large clusters covering 89% of the dataset, exhibiting significantly different volumetric and temporal properties compared to classical email phishing, and relying on web UI elements as visual baits. Finally, we look at the distribution vectors and show that clickbait PDFs are not only distributed via attachments but also via Search Engine Optimization attacks, placing clickbait PDFs outside the email distribution ecosystem. Clickbait PDFs seem to be a lurking threat, not subjected to any form of content-based filtering or detection: AV scoring systems, like VirusTotal, rank them considerably low, creating a blind spot for organizations. While URL blocklists can help to prevent victims from visiting the attack web pages, we observe that they have a limited coverage.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.5555/3666122.3668735,

title = {Fast attention over long sequences with dynamic sparse flash attention},

author = {Matteo Pagliardini and Daniele Paliotta and Martin Jaggi and François Fleuret},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the 37th International Conference on Neural Information Processing Systems},

publisher = {Curran Associates Inc.},

address = {New Orleans, LA, USA},

series = {NIPS '23},

abstract = {Transformer-based language models have found many diverse applications requiring them to process sequences of increasing length. For these applications, the causal self-attention—which is the only component scaling quadratically w.r.t. the sequence length—becomes a central concern. While many works have proposed schemes to sparsify the attention patterns and reduce the computational overhead of self-attention, those are often limited by implementation concerns and end up imposing a simple and static structure over the attention matrix. Conversely, implementing more dynamic sparse attention often results in runtimes significantly slower than computing the full attention using the Flash implementation from Dao et al. (2022). We extend FlashAttention to accommodate a large class of attention sparsity patterns that, in particular, encompass key/query dropping and hashing-based attention. This leads to implementations with no computational complexity overhead and a multi-fold runtime speedup on top of FlashAttention. Even with relatively low degrees of sparsity, our method improves visibly upon FlashAttention as the sequence length increases. Without sacrificing perplexity, we increase the training speed of a transformer language model by 2.0 × and 3.3 × for sequences of respectively 8k and 16k tokens.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.1145/3617694.3623234,

title = {Fairness Without Demographic Data: A Survey of Approaches},

author = {Carolyn Ashurst and Adrian Weller},

url = {https://doi.org/10.1145/3617694.3623234},

doi = {10.1145/3617694.3623234},

isbn = {9798400703812},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the 3rd ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization},

publisher = {Association for Computing Machinery},

address = {Boston, MA, USA},

series = {EAAMO '23},

abstract = {Detecting, measuring and mitigating various measures of unfairness are core aims of algorithmic fairness research. However, the most prominent approaches require access to individual level demographic information, such as sex or race. In practice, such information is often inaccurate, incomplete or entirely absent. A wealth of techniques have been proposed in response, such as proxies, trusted third parties and cryptographic solutions. These vary greatly in their assumptions, aims and costs, making it challenging to ascertain which methods might be appropriate for a given setting. We aim to clarify the landscape by providing an overview of the different proposals, surfacing the key dimensions in which they differ, and briefly discussing associated benefits, limitations and ethical considerations. We hope this will be of benefit to researchers, practitioners and regulators attempting to navigate this space.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.5555/3620237.3620613,

title = {Fact-saboteurs: a taxonomy of evidence manipulation attacks against fact-verification systems},

author = {Sahar Abdelnabi and Mario Fritz},

isbn = {978-1-939133-37-3},

year  = {2023},

date = {2023-01-01},

booktitle = {Proceedings of the 32nd USENIX Conference on Security Symposium},

publisher = {USENIX Association},

address = {Anaheim, CA, USA},

series = {SEC '23},

abstract = {Mis- and disinformation are a substantial global threat to our security and safety. To cope with the scale of online misinformation, researchers have been working on automating fact-checking by retrieving and verifying against relevant evidence. However, despite many advances, a comprehensive evaluation of the possible attack vectors against such systems is still lacking. Particularly, the automated fact-verification process might be vulnerable to the exact disinformation campaigns it is trying to combat. In this work, we assume an adversary that automatically tampers with the online evidence in order to disrupt the fact-checking model via camouflaging the relevant evidence or planting a misleading one. We first propose an exploratory taxonomy that spans these two targets and the different threat model dimensions. Guided by this, we design and propose several potential attack methods. We show that it is possible to subtly modify claim-salient snippets in the evidence and generate diverse and claim-aligned evidence. Thus, we highly degrade the fact-checking performance under many different permutations of the taxonomy's dimensions. The attacks are also robust against post-hoc modifications of the claim. Our analysis further hints at potential limitations in models' inference when faced with contradicting evidence. We emphasize that these attacks can have harmful implications on the inspectable and human-in-the-loop usage scenarios of such models, and we conclude by discussing challenges and directions for future defenses.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}