128,000 token length, weights available under a non-commercial license or a commercial API, fine-tuned for both tool use and citation-providing RAG $3/M input, $15/M output – same price as Claude 3 Sonnet
LLMS
-
Prompt Engineering Techniques for Haiku Model Optimization
By
–
The two tricks I'm trying with Haiku that seem to work incredibly well are feeding it a small targeted sequence of fake "user"/"assistant" examples in the messages array and finishing with a prefilled "assistant" message to point it in the right direction
-
RAG and LLM: Advancing Dynamic Language Modeling Frontier
By
–
RAG and LLM: A New Frontier in Dynamic Language Modeling
by @_odsc Read more: https://
buff.ly/4129LBL #AI #BigData #DataScience #MachineLearning #DeepLearning cc: @pascal_bornet @pbalakrishnarao @rtehrani -
AI-powered coding with Codeium
By
–
2. AI coding with Codieum: Boost your coding with Codium's AI-powered autocomplete and chat features. Try it here: https://
codeium.com -
Emergence in AI Models: Beyond Questionable Titles and Rigorous Evaluation
By
–
the title of the post overreaches with the question mark as cop out . even the mirage paper acknowledges that some abilities do emerge even after a more rigorous rewrite of evals to be more linear. i dont think anybody seriously thinks emergence completely doesnt exist
-
Assessing LLM Brain Damage: Rapid Evaluation Methods
By
–
how do you assess brain damage of an llm this quickly?
-
Statistical Literacy Requires Priors and Baselines for JSON Mode
By
–
you need priors and baselines to be statistically literate. i expected json mode to be <half of fn calling, esp since fn calling is more than twice as old.
-
JSON Mode Outperforms Function Calling in Large Sample Study
By
–
sample size n>2000, json mode beat out function calling! completely unexpected and in fact changed how i think about the tradeoffs. recirculating bc poll concluded
-
Groq Launches Tool Calling with LangChain Structured Output Support
By
–
Groq tool calling + structured output @GroqInc just dropped tool calling! We've added LangChain support (including the popular `withStructuredOutput` method!) so you can try it in your favorite chains and apps. It supports @MistralAI Mixtral, Llama 70B, and Google
-
GPT-4 Debate Effectively Reduces Conspiracy Theory Beliefs
By
–
Unexpected & big: it is famously hard to get people to stop believing in conspiracy theories, but… A controlled trial finds a 3 round debate with GPT-4 arguing the other side robustly lowers conspiracy theory beliefs and the effects persist over time, even for true believers.