An exciting step toward responsible AI: @MLCommons just announced the creation of the AI Safety working group which will focus on developing safety benchmarks for LLMs using the groundbreaking HELM framework from @StanfordCRFM
. Congrats @percyliang & team! https://
mlcommons.org/en/news/format
ion-ai-safety-working-group/
…
LLMS
-
MLCommons launches AI Safety working group with HELM framework
By
–
-
Peak On-Device Model Capacity Reaches 13B Parameters
By
–
Currently peak on device is 13B parameters
-
Llama2-7B Reaches 30 Tokens Per Second On-Device
By
–
Here we go. Llama2-7B at 30 tokens per second ON CHIP. Give it 2 more cycles and we’ll see the larger models on device at these speeds. LLMs will be as omnipresent as calculators, simple maths.
-
Stop Sequences Control AI Text Generation Length and Chain Prompts
By
–
Stop sequences dictate text generation termination if you want shorter answers or more detailed ones. It's useful when you chain prompts together and wish to insert other knowledge in between the model's responses.
-
Optimize LLMs: Temperature, Stop Sequences, and Penalty Parameters
By
–
Optimize your Large Language Models (LLMs) by adjusting parameters like temperature, stop sequences, and penalties for frequency and presence. Higher temperature increases diversity, and lower enhances determinism.
-
Training State-of-the-Art LLM with AutoTrain Advanced
By
–
Training state of the art LLM model with experiment tracking using weights & biases without writing a single line of code with AutoTrain Advanced
-
LLM-Based OS or Advanced PRO Products in 2026?
By
–
Could it be that we will see an LLM based OS already… Or will we see some scary fast PRO products? Whats your take?
-
Overview of LLM retrieval and RAG tuning
By
–
Good intro/overview to LLM retrieval (RAG) from @c_viernickel and Bryan Zhu on Scale's blog — it's not RAG vs. fine-tuning, it's tuning for RAG; for interesting tasks you want both:
-
Generative agents with memory presented at UIST 2023
By
–
A large language model enables generative agents that can remember interactions and build relationships. @joon_s_pk and colleagues will present their work at the User Interface Software and Technology Symposium (UIST) 2023. @ACMUIST
-
AI’s $200 Billion Question: Investment and Future Impact
By
–
AI’s $200B Question https://
bit.ly/3Q2q7oU #AI #MachineLearning #DeepLearning #LLMs #DataScience