We construct an instruction dataset for the large language model (LLM) in the Japanese finance domain. Domain adaptation of language models, including LLMs, is receiving more attention as language models become more popular. This study demonstrates the effectiveness of domain adaptation through instruction tuning. To achieve this, we propose an instruction tuning data in Japanese called JaFIn, the Japanese Financial Instruction Dataset. JaFIn is manually constructed based on multiple data sources, including Japanese government websites, which provide extensive financial knowledge. We then utilize JaFIn to apply instruction tuning for several LLMs, demonstrating that our models specialized in finance have better domain adaptability than the original models. The financial-specialized LLMs created were evaluated using a quantitative Japanese financial benchmark and qualitative response comparisons, showing improved performance over the originals.

Fox News Flash top headlines are here. Check out what's clicking on Foxnews.com. New Jersey Democratic Gov. Phil Murphy on Tuesday announced a series of new measures he wants the newly expanded Democrat-led Legislature to adopt, including allowing 16-year-olds to vote in school board elections, reducing medical debt, expanding affordable housing and launching an artificial intelligence "moonshot." Murphy delivered his sixth state of the state address before a joint legislative session in the ornate Assembly chamber where Democrats picked up six seats in the November election. Murphy also reiterated calls he's made since his reelection in 2021 to further ease property taxes and expand free pre-K, among the measures that he says make the state "stronger and fairer."

Abstractive summarization systems aim to produce more coherent and concise summaries than their extractive counterparts. Popular neural models have achieved impressive results for single-document summarization, yet their outputs are often incoherent and unfaithful to the input. In this paper, we introduce SENECA, a novel System for ENtity-drivEn Coherent Abstractive summarization framework that leverages entity information to generate informative and coherent abstracts. Our framework takes a two-step approach: (1) an entity-aware content selection module first identifies salient sentences from the input, then (2) an abstract generation module conducts cross-sentence information compression and abstraction to generate the final summary, which is trained with rewards to promote coherence, conciseness, and clarity. The two components are further connected using reinforcement learning. Automatic evaluation shows that our model significantly outperforms previous state-of-the-art on ROUGE and our proposed coherence measures on New York Times and CNN/Daily Mail datasets. Human judges further rate our system summaries as more informative and coherent than those by popular summarization models.

"So there's no sales tax revenue because there's no sales," says Joseph Henchman, vice president of state projects at the Tax Foundation. Cities and states get about 30 percent of their revenue from property taxes, 20 percent from sales tax, and another 20 from individual income taxes. "If revenues drop by a third"--the projected impact of automation--Henchman says, "that means services need to be cut back by a third, either through trying to be more focused or efficient with the services we do provide, or by actually having to pare back what government does." A robot tax isn't going to save jobs, but the idea is that it could help cushion the impact of mass automation by funding a universal basic income.