Keyphrases are the essential topical phrases that summarize a document. Keyphrase generation is a long-standing NLP task for automatically generating keyphrases for a given document. While the task has been comprehensively explored in the past via various models, only a few works perform some preliminary analysis of Large Language Models (LLMs) for the task. Given the impact of LLMs in the field of NLP, it is important to conduct a more thorough examination of their potential for keyphrase generation. In this paper, we attempt to meet this demand with our research agenda. Specifically, we focus on the zero-shot capabilities of open-source instruction-tuned LLMs (Phi-3, Llama-3) and the closed-source GPT-4o for this task. We systematically investigate the effect of providing task-relevant specialized instructions in the prompt. Moreover, we design task-specific counterparts to self-consistency-style strategies for LLMs and show significant benefits from our proposals over the baselines.

Recent work in interpretability shows that large language models (LLMs) can be adapted for new tasks in a learning-free way: it is possible to intervene on LLM representations to elicit desired behaviors for alignment. For instance, adding certain bias vectors to the outputs of certain attention heads is reported to boost the truthfulness of models. In this work, we show that localized fine-tuning serves as an effective alternative to such representation intervention methods. We introduce a framework called Localized Fine-Tuning on LLM Representations (LoFiT), which identifies a subset of attention heads that are most important for learning a specific task, then trains offset vectors to add to the model's hidden representations at those selected heads. LoFiT localizes to a sparse set of heads (3%) and learns the offset vectors from limited training data, comparable to the settings used for representation intervention. For truthfulness and reasoning tasks, we find that LoFiT's intervention vectors are more effective for LLM adaptation than vectors from representation intervention methods such as Inference-time Intervention. We also find that the localization step is important: selecting a task-specific set of attention heads can lead to higher performance than intervening on heads selected for a different task. Finally, for the tasks we study, LoFiT achieves comparable performance to other parameter-efficient fine-tuning methods such as LoRA, despite modifying 20x-200x fewer parameters than these methods.

Stanley Black & Decker is best known as the manufacturer of tools for home improvement projects, but it also makes products the average consumer seldom notices, like fasteners to keep car parts secure and the electronic doors typically used at retail stores. Me, Myself, and AI podcast hosts Sam Ransbotham and Shervin Khodabandeh sat down with Mark Maybury, the company's first chief technology officer, to learn how artificial intelligence factors into this 179-year-old brand's story. As Stanley Black & Decker's CTO, Mark Maybury manages a team across the company's businesses and functions, advises on technological threats and opportunities, and provides access to all elements of the global technology ecosystem. Previously, Maybury spent 27 years at The Mitre Corporation, where he held a variety of strategic technology roles, including vice president of intelligence portfolios and chief security officer. Before joining Mitre, he was an officer in the U.S. Air Force, where he also served as chief scientist from 2010 to 2013. Maybury is on the Defense Science Board and the Connecticut Science Center Board and served on the Air Force Scientific Advisory Board and the Homeland Security Science and Technology Advisory Committee for several years. He is a fellow in IEEE and the Association for the Advancement of Artificial Intelligence. Maybury has a doctorate degree in AI from Cambridge University. During their conversation, Mark described how categorizing the technology-infused innovation projects he leads across the company into six levels, ranging from incremental improvements to radical innovations, helps Stanley Black & Decker balance its product development portfolio. He also shared some insights for organizations thinking about responsible AI guidelines and discussed how Stanley Black & Decker is increasing its focus on sustainability. If you're enjoying the Me, Myself, and AI podcast, continue the conversation with us on LinkedIn.

FILE - In this May 13, 2018, file photo, MLB umpire Joe West, right, talks with a player in the ninth inning during a baseball game between the Arizona Diamondbacks and the Washington Nationals in Phoenix. West, who has umpired more than 5,000 big league games, said the 2016 TrackMan computer system test was far from perfect. NEW YORK – Get ready for strikes by robots. Computers will be used for ball/strike calls starting April 25 in the independent Atlantic League, where the distance between home and first will be shortened by 3 inches. The ground between the mound and home plate will lengthen by 2 feet for the second half of the season beginning July 12.

Get ready for strikes by robots. Computers will be used for ball/strike calls starting April 25 in the independent Atlantic League, where the distance between home and first will be shortened by 3 inches. The ground between the mound and home plate will lengthen by 2 feet for the second half of the season beginning July 12. The 60-foot-6-inch distance between the front of the pitching rubber and the back point of home plate has been standard since 1893, but Major League Baseball reached a three-year deal to experiment in the Atlantic League, an eight-team circuit that occasionally produces big leaguers. Infield defensive shifts will be limited.

For some positions that Siemens AG SIEGY 1.43% needs to fill, there may be a universe of fewer than 2,000 qualified people in the U.S., said Michael Brown, vice president of talent acquisition in the Americas for the German industrial conglomerate that makes everything from gas turbines to mammography machines. "The question is how many of those are looking for a job?" Finding the right potential candidates on sites like LinkedIn isn't easy because "they're tired of being found." Siemens has 377,000 employees world-wide and about 50,000 in the U.S. At the moment, it has about 1,500 open jobs across America, most of which require some software or science-related background. Employers are handicapped by several factors, data show and recruiters say: Cutting-edge skills are evolving faster than universities can train people, the supply of talented young workers entering these fields isn't satisfying the huge demand for them, and mobility--a worker's willingness to uproot their life for a job in a new place--has declined. The odds of luring rare, coveted candidates away from their current job or city are long, Mr. Brown said.

Ensuring the consistency and completeness of Semantic Web ontologies is practically impossible, because of their scale and highly dynamic nature. Many web applications, therefore, must deal with vague, incomplete and even inconsistent knowledge. Rules were shown to be very effective in processing such knowledge, and future web services are expected to depend heavily on them. RuleML, which is the earliest effort to define a normalized markup for representing and exchanging rules on the web, is currently limited to Horn rules. Significant research efforts are underway to extend RuleML with more flexible representation and reasoning capabilities. This paper presents an extension of the current rule format intended to accommodate uncertain and/or inconsistent knowledge, and shows how one truth maintenance logic can be adapted and extended to support such rules.