LG AI Research Launches EXAONE 3.0: A Bilingual Language Model for Open Research
08/08/2024
This model is the pioneering open-access member of the EXAONE family, designed to democratize expert-level AI capabilities.
This model is the pioneering open-access member of the EXAONE family, designed to democratize expert-level AI capabilities.
A 6 DoF ABB 1100 robotic arm on linear gantries won 45% of matches against human players of various skill
LLaVA-OneVision excels in chart interpretation, visual reasoning, and real-world image comprehension, rivaling advanced commercial models like GPT-4V.
AMT halves token sequence length, reducing computational load and doubling MeshAnything V2’s mesh capacity to 1600 faces.
Midjourney 6.1 boosts image coherence with new 2x upscalers, improved texture quality, and 25% faster processing speeds.
The MMAU benchmark features 20 tasks and over 3,000 prompts for a detailed assessment of LLM capabilities, aiming to pinpoint
Tora outperforms existing methods in motion control, achieving 3-5 times better trajectory accuracy, especially in generating longer videos up to
MoNE’s key advantage is its ability to adapt to varying inference-time compute budgets with a single trained model.
Priced at $99 with no subscription fees, Friend allows users to speak their minds, pause interactions, and review conversations.
“To become more useful, wearables need embedded AI agents that are context-aware and proactive,” said Cayden Pierce.
The Llama 3.1 405B model matches top closed models, supports 128k context length, eight languages, and excels in code generation,
Outages are often caused due to network issues, including design/configuration, hardware, capacity, software, and environmental threats.
With new AI models from Google and Anthropic emerging, studies highlight GPT-4o’s leading performance, though other models also show strong
The Llama3-ChatQA-2-70B model can process contexts up to 128,000 tokens, matching GPT-4-Turbo’s capacity.
The CLP framework enhances language models for summarization, conversational agents, and social norms encoding by balancing multiple objectives for real-world
NeedleBench tests bilingual long-context capabilities with tasks from 4,000 to over 1 million tokens
These tools offer accounting solutions, automate payable processes, categorise transactions, and provide a more efficient way of accounting.
“This is my first visit to India, and I was very impressed with the level of innovation and new ideas,
“People will soon realise that today 50% of the images available in the top microstock agencies can be generated in
The new model compresses spreadsheets by up to 96%, enabling LLMs to manage larger datasets within token limits.
These tools offer deep consumer insights, personalised marketing, and optimised lead identification, boosting sales conversion rates.
Claude 3.5 Sonnet’s strengths in data science, analysis, and vision capabilities were highlighted by Swami Sivasubramanian, head of AI at
PaliGemma excels in image captioning and video understanding, surpassing larger models with its versatile architecture and achieving top results on
MiraData offers videos averaging 72.1 seconds with detailed captions of 318 words, surpassing other datasets in length and descriptive detail.
FunLLM’s SenseVoice offers multilingual ASR, emotion recognition, and audio event detection, while CosyVoice excels in multilingual voice generation and cross-lingual
LivePortrait model introduces innovative stitching and retargeting modules, enabling more realistic and expressive animated portraits.
Companies like Microsoft, Google, and IBM provide cloud services for the public sector; however, the AWS initiative is unique and
Claude Sonnet 3.5’s features spark debate and interest, showcasing its strengths as a compelling model.
The model is available via Anthropic API, Amazon Bedrock, and Google Cloud’s Vertex AI, priced at $3 per million input
“AI can remember everything about you and everything you’ve said, enhancing your memory and personalisation,” says Microsoft chief Mustafa Suleyman.