CybAI news

[Submitted on 12 Jul 2024] View a PDF of the paper titled Investigating the Interplay of Prioritized Replay and Generalization, by Parham Mohammad Panahi and 3 other authors View PDF HTML (experimental) Abstract:Experience replay is ubiquitous in reinforcement learning, to reuse past data and improve sample efficiency. Though a variety of smart sampling schemes have been introduced to improve performance, uniform sampling by far remains the most common approach. One exception is Prioritized Experience Replay (PER), where sampling is done proportionally to TD errors, inspired by the success of prioritized sweeping in dynamic programming. The original work on PER showed…

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs. Source link lol

arXiv:2407.09722v1 Announce Type: new Abstract: Transformer-based Large language models (LLMs) have demonstrated their power in various tasks, but their inference incurs significant time and energy costs. To accelerate LLM inference, speculative decoding uses a smaller model to propose one sequence of tokens, which are subsequently validated in batch by the target large model. Compared with autoregressive decoding, speculative decoding generates the same number of tokens with fewer runs of the large model, hence accelerating the overall inference by $1$-$2times$. However, greedy decoding is not the optimal decoding algorithm in terms of output perplexity, which is a direct measurement of the…

(treety/Shutterstock) Some of the biggest AI companies in the world are using material taken from thousands of content creators on YouTube to their AI models without compensating the creators of those videos, ProofNews reported today. According to the article by ProofNews authors Annie Gilbertson and Alex Reisner, AI companies like Anthropic, Apple, and Nvidia used a dataset called “YouTube Subtitles” that contained transcribed text from more than 173,000 YouTube videos to train their models. YouTube Subtitles is part of a larger, open-source data set created by EleutherAI called the Pile. According to a 2020 paper by EleutherAI researchers, the Pile…