Driving down inference cost is the #1 requirement to enabling great AI-first apps. @YI01_Yi's new model Yi-Lightning is ranked #6 (
https://
lnkd.in/gvCSxTxz) in the world, and costs only ~5% cost of its peers. Interview with @EleanorOlcott of @FinancialTimes
LLMS
-
Yi-Lightning Model Reduces Inference Costs by 95 Percent
By
–
-
High Quality 5T Token Dataset Released for LLM Training
By
–
They also released a 5T token high quality, processed dataset for training LLMs https://
huggingface.co/datasets/Zyphr
a/Zyda-2
… -
LLM Benchmark Reveals Sequential Reasoning Limitations
By
–
Fascinating new LLM benchmark alert! My colleagues @DCasBol and @thefillm have developed a simple yet powerful test exposing LLMs' struggles with sequential reasoning. Key findings:
• Most LLMs start erring after just 2 operations • OpenAI o1-mini unsurprisingly -
Meta Researchers Develop AI Models to Think Before Answering
By
–
Meta researchers develop method to make AI models "think" before answering Scientists from Meta, UC Berkeley, and NYU have created a new technique to improve how large language models (LLMs) approach general tasks. Called "Thought Preference Optimization" (TPO), the method aims
-
Self-Harmonized Chain of Thought Advances LLM Reasoning
By
–
Self-Harmonized Chain of Thought https://
bit.ly/4gQ65um
#AI #MachineLearning #DeepLearning #LLMs #DataScience -
Fine-tune Llama-3-8b for SQL Code Generation with Small Models
By
–
#CodeGen is a popular use case for LLMs, but you don't need a huge or proprietary model to get good performance. Check out this #tutorial — sample data + notebook included — to learn how to fine-tune a small open-source model (#SLM) LLama-3-8b for SQL generation, achieving a
-
Meta Releases SpiritLM Research Model with Code and Weights
By
–
(5/6) Learn more & access Meta's research paper, model weights, and code: Research Paper: https://
arxiv.org/abs/2402.05755 Blog & Overview: https://
ai.meta.com/blog/fair-news
-segment-anything-2-1-meta-spirit-lm-layer-skip-salsa-lingua/?utm_source=twitter&utm_medium=organic_social&utm_content=thread&utm_campaign=fair
… GitHub Repository & Demo: https://
github.com/facebookresear
ch/spiritlm
… -
Meta Releases Breakthrough Text-Speech AI Model for Innovation
By
–
(4/6) By releasing this breakthrough model, Meta hopes to spark innovation within the research community, inspiring new approaches to text and speech integration. The future of AI voice experiences just got a major upgrade!
-
Anthropic Developing New Artifacts Attachments and RAG Features
By
–
Anthropic keeps preparing Artifacts Attachments feature for the release It is still not referenced anywhere directly and it is not fully clear what this feature will be doing exactly. Potentially it may be related to the "Reference Files" RAG feature which is also in
-
Meta Spirit LM: Open Source Multimodal Language Model for Text and Speech
By
–
Today we released Meta Spirit LM — our first open source multimodal language model that freely mixes text and speech.
— AI at Meta (@AIatMeta) 18 octobre 2024
Many existing AI voice experiences today use ASR to techniques to process speech before synthesizing with an LLM to generate text — but these approaches… pic.twitter.com/gMpTQVq0nEToday we released Meta Spirit LM — our first open source multimodal language model that freely mixes text and speech. Many existing AI voice experiences today use ASR to techniques to process speech before synthesizing with an LLM to generate text — but these approaches