Four days after the attempt on his life, the voice of Donald Trump booms from the speakers in Montego Bay, Jamaica: "If they needed an assassin, they should have sent for Bodyguard … about to commit a quadruple murder at Sumfest in Montego Bay." The audience are taken by surprise, having been primed for a reggae riddim to drop, and laugh. The Bodyguard crew have just taken to the stage at Sumfest Global Sound Clash, a musical gladiatorial contest where sound systems battle against one another with creative mixing, hyped-up MCs and exclusive – often incendiary – recordings featuring star guests and in-jokes. AI vocalists such as this fake Trump, however, are sending shockwaves through a decades-old musical tradition in which authenticity and originality are paramount, and sound systems pay premium rates to artists to get vocals for clashes. "AI is going to mash up the industry," says Fabian Anderson, a dub agent who liaises between artists and sound systems to secure those exclusive tracks.

Dataset distillation aims to compress a training dataset by creating a small number of informative synthetic samples such that neural networks trained on them perform as well as those trained on the original training dataset. Current text dataset distillation methods create each synthetic sample as a sequence of word embeddings instead of a text to apply gradient-based optimization; however, such embedding-level distilled datasets cannot be used for training other models whose word embedding weights are different from the model used for distillation. To address this issue, we propose a novel text dataset distillation approach, called Distilling dataset into Language Model (DiLM), which trains a language model to generate informative synthetic training samples as text data, instead of directly optimizing synthetic samples. We evaluated DiLM on various text classification datasets and showed that distilled synthetic datasets from DiLM outperform those from current coreset selection methods. DiLM achieved remarkable generalization performance in training different types of models and in-context learning of large language models. Our code will be available at https://github.com/arumaekawa/DiLM.

Code provides a general syntactic structure to build complex programs and perform precise computations when paired with a code interpreter - we hypothesize that language models (LMs) can leverage code-writing to improve Chain of Thought reasoning not only for logic and arithmetic tasks, but also for semantic ones (and in particular, those that are a mix of both). For example, consider prompting an LM to write code that counts the number of times it detects sarcasm in an essay: the LM may struggle to write an implementation for "detect_sarcasm(string)" that can be executed by the interpreter (handling the edge cases would be insurmountable). However, LMs may still produce a valid solution if they not only write code, but also selectively "emulate" the interpreter by generating the expected output of "detect_sarcasm(string)" and other lines of code that cannot be executed. In this work, we propose Chain of Code (CoC), a simple yet surprisingly effective extension that improves LM code-driven reasoning. The key idea is to encourage LMs to format semantic sub-tasks in a program as flexible pseudocode that the interpreter can explicitly catch undefined behaviors and hand off to simulate with an LM (as an "LMulator"). Experiments demonstrate that Chain of Code outperforms Chain of Thought and other baselines across a variety of benchmarks; on BIG-Bench Hard, Chain of Code achieves 84%, a gain of 12% over Chain of Thought. CoC scales well with large and small models alike, and broadens the scope of reasoning questions that LMs can correctly answer by "thinking in code". Project webpage: https://chain-of-code.github.io.

In this paper, we address the hallucination problem commonly found in natural language generation tasks. Language models often generate fluent and convincing content but can lack consistency with the provided source, resulting in potential inaccuracies. We propose a new decoding method called Fidelity-Enriched Contrastive Search (FECS), which augments the contrastive search framework with context-aware regularization terms. FECS promotes tokens that are semantically similar to the provided source while penalizing repetitiveness in the generated text. We demonstrate its effectiveness across two tasks prone to hallucination: abstractive summarization and dialogue generation. Results show that FECS consistently enhances faithfulness across various language model sizes while maintaining output diversity comparable to well-performing decoding algorithms.

The task of triplet extraction aims to extract pairs of entities and their corresponding relations from unstructured text. Most existing methods train an extraction model on training data involving specific target relations, and are incapable of extracting new relations that were not observed at training time. Generalizing the model to unseen relations typically requires fine-tuning on synthetic training data which is often noisy and unreliable. We show that by reducing triplet extraction to a template infilling task over a pre-trained language model (LM), we can equip the extraction model with zero-shot learning capabilities and eliminate the need for additional training data. We propose a novel framework, ZETT (ZEro-shot Triplet extraction by Template infilling), that aligns the task objective to the pre-training objective of generative transformers to generalize to unseen relations. Experiments on FewRel and Wiki-ZSL datasets demonstrate that ZETT shows consistent and stable performance, outperforming previous state-of-the-art methods, even when using automatically generated templates. https://github.com/megagonlabs/zett/

One popular technique to solve temporal planning problems consists in decoupling the causal decisions, demanding them to heuristic search, from temporal decisions, demanding them to a simple temporal network (STN) solver. In this architecture, one needs to check the consistency of a series of STNs that are related one another, therefore having methods to incrementally re-use previous computations and that avoid expensive memory duplication is of paramount importance. In this paper, we describe in detail how STNs are used in temporal planning, we identify a clear interface to support this use-case and we present an efficient data-structure implementing this interface that is both time- and memory-efficient. We show that our data structure, called \deltastn, is superior to other state-of-the-art approaches on temporal planning sequences of problems.

Investors and workers in the ever -expanding local business process outsourcing (BPO) sector are being warned about the dangers of the evolving development of artificial intelligence (AI) within the scientific community and the effects it can have on the industry. According to Science, Energy and Technology Minister Dr Andrew Wheatley who was speaking at last week's inaugural Symposium and Exposition in Montego Bay, St James, the direction of the continued innovation within the technological sector towards AI will have profound effects on the BPO sector. "Some persons may interpret it as being troubling developments, but I want for us to look at it more from the perspective that it provides an opportunity for us to be more innovative," said Wheatley. "These developments, I am sure, will have profound effects on our way of life, on our way of doing business, on our way of interacting with each other. It will also have profound effects on the BPO sector. In fact, it has already started."

Use code WTBDBP16 to get 15% off the price of registration for 2 Day and Combo passes. Imaging, Vision and Learning based on Optimization and PDEs, August 29 – September 1, Bergen, Norway.

International networks, although highly efficient, may produce surprising threshold effects that shift costs to geographically distant locations. International utility, transportation, and information networks facilitate the efficient flow of information, energy, goods and people. These networks exhibit a scale-free network structure with a few large “hubs”. Yet their efficiency belies their lack of robustness. Because such networks transcend national boundaries, furthermore, disruptions to the network in one geographic region may have profound economic and national security costs for countries in another region. To illustrate how complex networks may transmit costs among countries, this paper builds an agent-based model (ABM) of the international air transportation system. The ABM employs a genetic algorithm to identify “small” disruptions that produce cascading network failures. The study makes two contributions. First, it demonstrates how some complex networks evolve into network structures that trade off robustness for efficiency. Second, it illustrates how researchers can combine agent-based modeling, evolutionary computation, and network analysis to simulate differing failure modes for global networks. This convergence of simulation methodologies characterizes the emerging field of computational social science.

Model transformations operate on models conforming to precisely defined metamodels. Consequently, it often seems relatively easy to chain them: the output of a transformation may be given as input to a second one if metamodels match. However, this simple rule has some obvious limitations. For instance, a transformation may only use a subset of a metamodel. Therefore, chaining transformations appropriately requires more information. We present here an approach that automatically discovers more detailed information about actual chaining constraints by statically analyzing transformations. The objective is to provide developers who decide to chain transformations with more data on which to base their choices. This approach has been successfully applied to the case of a library of endogenous transformations. They all have the same source and target metamodel but have some hidden chaining constraints. In such a case, the simple metamodel matching rule given above does not provide any useful information.