Recent research efforts have explored the potential of leveraging natural language inference (NLI) techniques to enhance relation extraction (RE). In this vein, we introduce MetaEntail-RE, a novel adaptation method that harnesses NLI principles to enhance RE performance. Our approach follows past works by verbalizing relation classes into class-indicative hypotheses, aligning a traditionally multi-class classification task to one of textual entailment. We introduce three key enhancements: (1) Instead of labeling non-entailed premise-hypothesis pairs with the uninformative "neutral" entailment label, we introduce meta-class analysis, which provides additional context by analyzing overarching meta relationships between classes when assigning entailment labels; (2) Feasible hypothesis filtering, which removes unlikely hypotheses from consideration based on pairs of entity types; and (3) Group-based prediction selection, which further improves performance by selecting highly confident predictions. MetaEntail-RE is conceptually simple and empirically powerful, yielding significant improvements over conventional relation extraction techniques and other NLI formulations. Our experimental results underscore the versatility of MetaEntail-RE, demonstrating performance gains across both biomedical and general domains.

Constructing simulation scenes that are both visually and physically realistic is a problem of practical interest in domains ranging from robotics to computer vision. This problem has become even more relevant as researchers wielding large data-hungry learning methods seek new sources of training data for physical decision-making systems. However, building simulation models is often still done by hand. A graphic designer and a simulation engineer work with predefined assets to construct rich scenes with realistic dynamic and kinematic properties. While this may scale to small numbers of scenes, to achieve the generalization properties that are required for data-driven robotic control, we require a pipeline that is able to synthesize large numbers of realistic scenes, complete with 'natural' kinematic and dynamic structures. To attack this problem, we develop models for inferring structure and generating simulation scenes from natural images, allowing for scalable scene generation from web-scale datasets. To train these image-to-simulation models, we show how controllable text-to-image generative models can be used in generating paired training data that allows for modeling of the inverse problem, mapping from realistic images back to complete scene models. We show how this paradigm allows us to build large datasets of scenes in simulation with semantic and physical realism. We present an integrated end-to-end pipeline that generates simulation scenes complete with articulated kinematic and dynamic structures from real-world images and use these for training robotic control policies. We then robustly deploy in the real world for tasks like articulated object manipulation. In doing so, our work provides both a pipeline for large-scale generation of simulation environments and an integrated system for training robust robotic control policies in the resulting environments.

Large language models (LLMs) like transformers have impressive in-context learning (ICL) capabilities; they can generate predictions for new queries based on input-output sequences in prompts without parameter updates. While many theories have attempted to explain ICL, they often focus on structured training data similar to ICL tasks, such as regression. In practice, however, these models are trained in an unsupervised manner on unstructured text data, which bears little resemblance to ICL tasks. To this end, we investigate how ICL emerges from unsupervised training on unstructured data. The key observation is that ICL can arise simply by modeling co-occurrence information using classical language models like continuous bag of words (CBOW), which we theoretically prove and empirically validate. Furthermore, we establish the necessity of positional information and noise structure to generalize ICL to unseen data. Finally, we present instances where ICL fails and provide theoretical explanations; they suggest that the ICL ability of LLMs to identify certain tasks can be sensitive to the structure of the training data.

This paper introduces the sgboost package in R, which implements sparse-group boosting for modeling high-dimensional data with natural groupings in covariates. Sparse-group boosting offers a flexible approach for both group and individual variable selection, reducing overfitting and enhancing model interpretability. The package uses regularization techniques based on the degrees of freedom of individual and group base-learners, and is designed to be used in conjunction with the mboost package. Through comparisons with existing methods and demonstration of its unique functionalities, this paper provides a practical guide on utilizing sparse-group boosting in R, accompanied by code examples to facilitate its application in various research domains. Overall, this paper serves as a valuable resource for researchers and practitioners seeking to use sparse-group boosting for efficient and interpretable high-dimensional data analysis.

The new artificial intelligence features Google announced just weeks ago are finally breaking through to the mainstream--albeit not in the manner Google might prefer. As you may have gleaned from recent coverage and chatter (or even experienced yourself), the autogenerated A.I. Overviews now sitting atop so many Google search results are giving answers that … well, to call them incorrect is true but doesn't quite nail it. Try surreal and ridiculous and potentially dangerous instead. Since their rollout, A.I. Overviews have told users to smoke cigarettes while pregnant, add glue to their home-baked pizza, sprinkle used antifreeze on their lawns, and boil mint in order to cure their appendicitis. To address the erroneous answers to both straightforward and jokey queries, Google appears to be addressing each incident one by one and tweaking the relevant Overviews accordingly. Still, the broken top-of-Google answers may even be spilling over into the search engine's other features, like its automatic calculator: One U.S.–based user found, posting a screenshot to X, that Google's tech couldn't even scan that the unit cm stands for centimeter, reading the measure as a whole meter.

If you use Facebook, WhatsApp or Instagram, you've probably noticed a new character pop up answering search queries or eagerly offering tidbits of information in your feeds, with varying degrees of accuracy. It's Meta AI, and it's here to help, at least according to Meta Platforms' CEO Mark Zuckerberg, who calls it "the most intelligent AI assistant that you can freely use." The chatbot can recommend local restaurants, offer more information on something you see in a Facebook post, search for airline flights or generate images in the blink of an eye. If you're chatting with friends to plan a night out, you can invite it into your group conversation by typing @MetaAI, then ask it to recommend, say, cocktail bars. Meta's AI tool has been integrated into chat boxes and search bars throughout the tech giant's platforms. The assistant appears, for example, at the top of your chat list on Messenger.

I'm a reporter on the WIRED Politics desk, and I'm taking over for Makena this week to talk about politicians rising from the dead in India and the rapper Eminem endorsing opposition parties in South Africa. These things haven't really happened, obviously, but deepfakes created by generative AI have made it seem like they have. Already, we're seeing how politicians, campaigns, and regular people are using generative AI in elections. And this is only the beginning. So today, WIRED is launching a project to track it, all over the world.

In India and Indonesia, dead leaders are rising to throw their support behind their political successors; rapper Eminem is endorsing opposition parties in South Africa; and in the United States, President Biden is telling voters in New Hampshire to stay home. All of these things "happened"–but none of them are real. The generative AI revolution is here, and it's coming for your elections. Welcome to the future, welcome to 2024. For the very first time, the widespread availability of generative AI is going to clash head-on with political campaigns and elections.

Large language models (LLMs) often hallucinate and lack the ability to provide attribution for their generations. Semi-parametric LMs, such as kNN-LM, approach these limitations by refining the output of an LM for a given prompt using its nearest neighbor matches in a non-parametric data store. However, these models often exhibit slow inference speeds and produce non-fluent texts. In this paper, we introduce Nearest Neighbor Speculative Decoding (NEST), a novel semi-parametric language modeling approach that is capable of incorporating real-world text spans of arbitrary length into the LM generations and providing attribution to their sources. NEST performs token-level retrieval at each inference step to compute a semi-parametric mixture distribution and identify promising span continuations in a corpus. It then uses an approximate speculative decoding procedure that accepts a prefix of the retrieved span or generates a new token. NEST significantly enhances the generation quality and attribution rate of the base LM across a variety of knowledge-intensive tasks, surpassing the conventional kNN-LM method and performing competitively with in-context retrieval augmentation. In addition, NEST substantially improves the generation speed, achieving a 1.8x speedup in inference time when applied to Llama-2-Chat 70B.

World modelling, i.e. building a representation of the rules that govern the world so as to predict its evolution, is an essential ability for any agent interacting with the physical world. Recent applications of the Transformer architecture to the problem of world modelling from video input show notable improvements in sample efficiency. However, existing approaches tend to work only at the image level thus disregarding that the environment is composed of objects interacting with each other. In this paper, we propose an architecture combining Transformers for world modelling with the slot-attention paradigm, an approach for learning representations of objects appearing in a scene. We describe the resulting neural architecture and report experimental results showing an improvement over the existing solutions in terms of sample efficiency and a reduction of the variation of the performance over the training examples. The code for our architecture and experiments is available at https://github.com/torchipeppo/transformers-and-slot-encoding-for-wm