LG released new version of generative AI, EXAONE 3.5

LG

Four months after unveiling EXAONE 3.0 in August, LG AI Research has open-sourced its latest AI model, EXAONE 3.5, further enhancing its performance.

  • Top performance among global open source AI models in 20 benchmarks, including real-world usability, long text processing, coding, and math. 

Unlike previous EXAONE 3.0, LG AI Research will open source all three models, an ultra-lightweight model for on-device use (2.4 billion parameters), a lightweight model for general purpose (7.8 billion parameters), and a high-performance model for specialized application (32 billion parameters).

To reduce the hallucination and to increase the accuracy and reliability of answers, LG AI Research has advanced Retrieval-Augmented Generation (RAG) technology, which generates answers based on real-time web search results or uploaded documents, and Multi-step Reasoning, which allows AI to deconstruct user’s inquiry step-by-step to develop logical reasoning.

Depending on the length of sentence inputs, EXAONE 3.5 can process long texts, roughly equivalent to around 100 pages at a time.

LG AI Research also published a technical report on the performance evaluation of the model, demonstrating confidence in the technology’s ability to compete on a global scale.

According to a technical report released by LG AI Research, EXAONE 3.5 is the world’s best class in real-world usability, long text processing, coding, and math.

LG AI Research disclosed both individual scores for the 20 benchmarks it used to evaluate performance and the average scores.

All three models, from on-device to high-performance, can be used for research purposes, and anyone can check and validate the models.

LG AI Research expects the open-source release to promote an open AI research ecosystem and increase the pace of innovation.

Also Read: SutiDAnalytics Boosts Data Analysis with Intelligent Features and Seamless Integration

‘As AI technology becomes a key strategic asset for each country, developing AI models with our own technology is meaningful in contributing to enhancing national AI competitiveness,’ said an official from LG AI Research.

  • LG employees can use AI agent, real-time web, and document-based question-and-answer capabilities without any data leakage or security issues

LG AI Research has officially launched its enterprise AI agent, ChatEXAONE for its employees, starting the ‘era of everyday AI’.

From the 9th of this month, LG employees can sign up and start using ChatEXAONE in their work immediately.

Based on the new EXAONE 3.5 model, ChatEXAONE applies information encryption and privacy protection technology so that employees can use it in their work without worrying about leaking internal data within the company’s security environment.

LG AI Research expects ChatEXAONE to help employees increase their work productivity and efficiency, from real-time web information search to document summarization, translation, report writing, data analysis, and coding.

With the application of EXAONE 3.5 to ChatEXAONE, LG AI Research has enhanced its performance and added ‘Deep’ and ‘Dive’ functions.

‘Deep’ is a feature that enables ChatEXAONE to analyze and infer multiple questions in stages and provide a comprehensive answer when a complex question is asked, and can be used when you want accurate and in-depth report-level results.

‘Dive’ is a feature that allows you to select a search scope such as general, global, academic, and YouTube to get answers based on the exact source according to your purpose.

ChatEXAONE recommends 133 job-specific prompts based 14 different job functions and provides personalized answers, and employees can set their interests according to their use.

LG AI Research plans to continue to expand the number of jobs and job categories based on employee feedback.

ChatEXAONE currently supports 32,000 tokens that can process 20,000 words in Korean (23,000 English words) simultaneously, enabling long-form questions and answers, and plans to expand to 128,000 tokens in the first half of next year.

  • Four years on, LG AI Research accelerates innovation, expands AI ecosystem and delivers tangible results

LG AI Research, the AI think tank for LG Group, which celebrated its fourth anniversary on 7 December 2020, is preparing for the beyond of EXAONE 3.5 already.

LG AI Research has researched on Large Action Model (LAM) that plans and acts on its own and plans to develop an AI agent based on the technology in 2025.

‘The recent advancement of generative AI models has accelerated, and it is important to upgrade them quickly,’ said Bae Kyunghoon, the president of LG AI Research. ‘We will speed up the pace of innovation and develop them into frontier models that represent Korea, with the goal of artificial super intelligence that can be applied to real-world industries.’

LG AI Research has strengthened its collaborations with AWS, Dell Technologies, Elsevier, Google Cloud, Intel, Jackson Laboratory, NVIDIA, Parsons School of Design, UiPath, University of MichiganUniversity of Pittsburgh Medical Centre(UPMC), University of Toronto and others to expand the ecosystem to create real-world outcomes with EXAONE and other AI technologies.

Meanwhile, LG has been accelerating its AI transformation for the past four years, centered on the LG AI Research Institute, which Koo Kwang-mo, Chairman and CEO of LG Group has been working on as a future business, and all affiliates are actively working to preempt future technologies and recruit talent.

SOURCE: PRNewswire