The decompose-then-verify strategy for verification of Large Language Model (LLM) generations decomposes claims that are then independently verified. Decontextualization augments text (claims) to ensure it can be verified outside of the original context, enabling reliable verification. While decomposition and decontextualization have been explored independently, their interactions in a complete system have not been investigated. Their conflicting purposes can create tensions: decomposition isolates atomic facts while decontextualization inserts relevant information. Furthermore, a decontextualized subclaim presents a challenge to the verification step: what part of the augmented text should be verified as it now contains multiple atomic facts? We conduct an evaluation of different decomposition, decontextualization, and verification strategies and find that the choice of strategy matters in the resulting factuality scores. Additionally, we introduce DnDScore, a decontextualization aware verification method which validates subclaims in the context of contextual information.

The number of voice actors in a rabid panic over AI in the industry is reaching a head, with social media brimming with daily posts on the topic, despite very little real world evidence of synthetic voices impacting the bottom line of working pros, or even amateurs for that matter. There's a supposition among the masses that because the technology is improving, its ascension is inevitable, and that by definition it will supplant human voice actors to a highly disruptive degree. It's easy to get caught up in the terror, but worst-case scenarios….heck, Now, there's no question that numerous companies and platforms want AI voiceover to be an Earth-shattering thing. And, inevitably, we are going to start seeing even well-known casting platforms offer AI voices against or alongside their human talent. Many voice actors are busy creating their own voice clones which they expect to make available through their websites, casting platforms, or through the platforms of the companies creating these artificial voices for them.

Large language models (LLMs) have shown impressive results while requiring little or no direct supervision. Further, there is mounting evidence that LLMs may have potential in information-seeking scenarios. We believe the ability of an LLM to attribute the text that it generates is likely to be crucial in this setting. We formulate and study Attributed QA as a key first step in the development of attributed LLMs. We propose a reproducible evaluation framework for the task and benchmark a broad set of architectures. We take human annotations as a gold standard and show that a correlated automatic metric is suitable for development. Our experimental work gives concrete answers to two key questions (How to measure attribution?, and How well do current state-of-the-art methods perform on attribution?), and give some hints as to how to address a third (How to build LLMs with attribution?).

The field of Question Answering (QA) has made remarkable progress in recent years, thanks to the advent of large pre-trained language models, newer realistic benchmark datasets with leaderboards, and novel algorithms for key components such as retrievers and readers. In this paper, we introduce PRIMEQA: a one-stop and open-source QA repository with an aim to democratize QA re-search and facilitate easy replication of state-of-the-art (SOTA) QA methods. PRIMEQA supports core QA functionalities like retrieval and reading comprehension as well as auxiliary capabilities such as question generation.It has been designed as an end-to-end toolkit for various use cases: building front-end applications, replicating SOTA methods on pub-lic benchmarks, and expanding pre-existing methods. PRIMEQA is available at : https://github.com/primeqa.

NEW YORK - In 1964, Stanley Kubrick, on the recommendation of the science-fiction author Arthur C. Clarke, bought a telescope. "He got this Questar and he attached one of his cameras to it," said Katharina Kubrick, the filmmaker's stepdaughter. "On a night where there was a lunar eclipse, he dragged us all out onto the balcony and we were able to see the moon like a big rubber ball. I don't think I've seen it as clearly since. He looked at it all the time."

Although much of the world focused on the the first Earth-Moon conversation at 4:18 p.m. EDT July 20, 1969, when Neil Armstrong and Buzz Aldrin landed the lunar module Eagle with just 30 seconds of fuel remaining, the entire mission was actually recorded. After painstakingly digitising the original tapes, the audio has now been put online. 'The Eagle has landed': NASA releases Apollo 11 mission audio Melania Trump visits a children's hospital in Nashville Matt Lauer says'I think they see me as an easy mark' Hundreds of audio conversations between the flight controllers and other teams supporting the mission were going on every minute of the mission over an intricate intercom system. 'The untold story of that momentous 8-day, 3-hour, mission by Armstrong, Aldrin and Michael Collins was the round-the-clock support provided by Mission Control and many communications and support locations around the globe,' said NASA. The trio of space explorers lifted off 9:32 a.m.

To mark the 47th anniversary of the Apollo 11 moon-landing mission, the Smithsonian has revealed a stunning high resolution 3D model of the command module'Columbia,' the spacecraft that carried astronauts Neil Armstrong, Edwin'Buzz' Aldrin and Michael Collins to the moon. The stunning virtual 3-D model that will allow the public a look inside the car-sized Apollo 11 command module. As one of the most sophisticated scans ever made of a historic artifact, it employed seven different scanning technologies to capture nearly 1 trillion high-resolution measurements producing more than a terabyte of compressed data resulting in a highly detailed master model. To mark the 47th anniversary of the Apollo 11 moon-landing mission, the Smithsonian has revealed a stunning high resolution 3D model of the command module'Columbia,' the spacecraft that carried astronauts Neil Armstrong, Edwin'Buzz' Aldrin and Michael Collins to the moon. As one of the most sophisticated scans ever made of a historic artifact, researchers at the Smithsonian employed seven different scanning technologies to capture nearly 1 trillion high-resolution measurements, producing more than a terabyte of compressed data.