Model attributions are important in deep neural networks as they aid practitioners in understanding the models, but recent studies reveal that attributions can be easily perturbed by adding imperceptible noise to the input. The non-differentiable Kendall's rank correlation is a key performance index for attribution protection. In this paper, we first show that the expected Kendall's rank correlation is positively correlated to cosine similarity and then indicate that the direction of attribution is the key to attribution robustness. Based on these findings, we explore the vector space of attribution to explain the shortcomings of attribution defense methods using $\ell_p$ norm and propose integrated gradient regularizer (IGR), which maximizes the cosine similarity between natural and perturbed attributions. Our analysis further exposes that IGR encourages neurons with the same activation states for natural samples and the corresponding perturbed samples. Our experiments on different models and datasets confirm our analysis on attribution protection and demonstrate a decent improvement in adversarial robustness.

Neural networks lack adversarial robustness, i.e., they are vulnerable to adversarial examples that through small perturbations to inputs cause incorrect predictions. Further, trust is undermined when models give miscalibrated predictions, i.e., the predicted probability is not a good indicator of how much we should trust our model. In this paper, we study the connection between adversarial robustness and calibration and find that the inputs for which the model is sensitive to small perturbations (are easily attacked) are more likely to have poorly calibrated predictions. Based on this insight, we examine if calibration can be improved by addressing those adversarially unrobust inputs. To this end, we propose Adversarial Robustness based Adaptive Label Smoothing (AR-AdaLS) that integrates the correlations of adversarial robustness and calibration into training by adaptively softening labels for an example based on how easily it can be attacked by an adversary. We find that our method, taking the adversarial robustness of the in-distribution data into consideration, leads to better calibration over the model even under distributional shifts. In addition, AR-AdaLS can also be applied to an ensemble model to further improve model calibration.

It would be hard to overstate the impact that Peter Thiel has had on the career of Sam Altman. After Altman sold his first startup in 2012, Thiel bankrolled his first venture fund, Hydrazine Capital. Thiel saw Altman as an inveterate optimist who stood at "the absolute epicenter, maybe not of Silicon Valley, but of a Silicon Valley zeitgeist." As Thiel put it, "If you had to look for the one person who represented a millennial tech person, it would be Altman." Each year, Altman would point Thiel toward the most promising startup at Y Combinator–Airbnb in 2012, Stripe in 2013, Zenefits in 2014–and Thiel would swallow hard and invest, even though he sometimes felt like he was being swept up in a hype cycle.

The performance of large language models (LLMs) is significantly affected by the quality and composition of their pre-training data, which is inherently diverse, spanning various domains, sources, and topics. Effectively integrating these heterogeneous data sources is crucial for optimizing LLM performance. Previous research has predominantly concentrated on domain-based data mixing, often neglecting the nuanced topic-level characteristics of the data. To address this gap, we propose a simple yet effective topic-based data mixing strategy that utilizes fine-grained topics generated through our topic modeling method, DataWeave. DataWeave employs a multi-stage clustering process to group semantically similar documents and utilizes LLMs to generate detailed topics, thereby facilitating a more nuanced understanding of dataset composition. Our strategy employs heuristic methods to upsample or downsample specific topics, which significantly enhances LLM performance on downstream tasks, achieving superior results compared to previous, more complex data mixing approaches. Furthermore, we confirm that the topics Science and Relationships are particularly effective, yielding the most substantial performance improvements. We will make our code and datasets publicly available.

Title: Extracting Relationships by Multi-Domain Matching Summary Assuming that a corpus is compiled from many sources belonging to different to domains, of which only a strict subset of domains is suitable to learn how to do prediction in a target domain, this paper proposes a novel approach (called Multiple Domain Matching Network (MDMN)) that aims at learning which domains share strong statistical relationships, and which source domains are best at supporting to learn the target domain prediction tasks. While many approaches to multiple-domain adaptation aim to match the feature-space distribution of *every* source domain to that of the target space, this paper suggests to not only map the distribution between sources and target, but also *within* source domains. The latter allows for identifying subsets of source domains that share a strong statistical relationship. Strengths Paper provides a theoretical analysis that yields a tighter bound on the weighted multi-source discrepancy. Weaknesses Tighter bound on multi-source discrepancy depends on the assumption that source domains that are less relevant for the target domain have lower weights.

Employing Large Language Models (LLMs) for semantic parsing has achieved remarkable success. However, we find existing methods fall short in terms of reliability and efficiency when hallucinations are encountered. In this paper, we address these challenges with a framework called QueryAgent, which solves a question step-by-step and performs step-wise self-correction. We introduce an environmental feedback-based self-correction method called ERASER. Unlike traditional approaches, ERASER leverages rich environmental feedback in the intermediate steps to perform selective and differentiated self-correction only when necessary. Experimental results demonstrate that QueryAgent notably outperforms all previous few-shot methods using only one example on GrailQA and GraphQ by 7.0 and 15.0 F1. Moreover, our approach exhibits superiority in terms of efficiency, including runtime, query overhead, and API invocation costs. By leveraging ERASER, we further improve another baseline (i.e., AgentBench) by approximately 10 points, revealing the strong transferability of our approach.

We're united by a mission: to make the world a safer place. Corvus Insurance uses novel data and artificial intelligence/machine learning to achieve better insights into commercial insurance risk. Our software empowers brokers and policyholders to better predict and prevent complex claims through data-driven tools and Smart Commercial Insurance policies. This allows us to reduce or eliminate the impact of adverse events, creating a safer world for everyone. Drawing inspiration from the intelligent, tool-building corvid family of birds, we are a team of high-flying collaborative builders.

Images of lounging elephants, treefrog embryos and a parasitic fungus erupting from the body of a fly have all won prizes at an ecology photo competition. The doomed fly was captured by evolutionary biologist Roberto García-Roa in the Tambopata National Reserve, Peru, and took the overall win at the second ever BMC Ecology and Evolution Image Competition. The contest aims to showcase the wonder of the natural world and emphasise the growing need to protect it from human activity. Mr García-Roa, from the University of Valencia, Spain, said: 'The image depicts a conquest that has been shaped by thousands of years of evolution. 'The spores of the so-called'zombie' fungus have infiltrated the exoskeleton and mind of the fly and compelled it to migrate to a location that is more favourable for the fungus's growth.

Anyone who has studied the field of science and engineering should have heard the word "control" once. In particular, "automatic control" is often used in the field of science and engineering. The field of control is so deep that one specialized book can be written by itself, but this time, let's take a brief look at "What is control?". We will discuss the difference between automatic control and manual control, the difference between feedback control and feed-forward control, and the relationship between artificial intelligence and edge computing, which have become popular in recent years. What is the definition of control?

A new research paper from Harvard University suggests that OpenAI’s headline-grabbing text-to-image framework DALL-E 2 has notable difficulty in reproducing even infant-level relations between the elements that it composes into synthesized photos, despite the dazzling sophistication of much of its output. The researchers undertook a user study involving 169 crowdsourced participants, who were presented with […]