Our philosophy was to keep it simple. We gave the model just two tools: a Bash Tool for running commands and an Edit Tool for viewing/editing files. Then we let the model decide how to tackle each problem.
PROMPT ENGINEERING
-
Benchmark Tests Complete Agent System Beyond Just Model
By
–
The benchmark tests the entire "agent" system – not just the model, but also the software scaffolding around it that handles prompts, parses outputs, and manages the interaction loop.
-
Chain-of-Thought Reasoning: Limited Benefits Beyond Math Tasks
By
–
Here’s an interesting paper (link in image/alt) that shows chain-of-thought (CoT) benefits in math and symbolic reasoning but only marginal benefits on other tasks. On MMLU, generating an answer without CoT leads to almost identical accuracy as CoT. The extra cost and time
-
Llama-3 Function Calling Workshop at ODSC West
By
–
Tomorrow, join our engineers at @_odsc West, where they will be hosting a hands-on workshop on Function Calling with Llama-3 Running at 1,000 tokens/s! Get ready to unlock the potential of Llama-3 with us. See you there! #Devs #GenAI
-
Superhuman’s AI Agent Evolution and Prompt Engineering Techniques
By
–
Email is a place where everyone thinks AI should be applied, and Superhuman is leading the charge there Really fun to hear the details of how their agent evolved, as well as cool prompt engineering tricks they developed ("double dipping") https://
langchain.com/breakoutagents
/superhuman
… -
GPT Action Simulation: Frame Testing and Prompt Optimization
By
–
not actually play play — i had a bunch of screen-caps and forward-simulated what GPT would take as an action.
I tried sending one frame, 10 frames, 24 frames, etc. and asked it to take one of several actions — and played with the prompting to adjust the state space and action -
LLM Sampler: Visualizing Language Model Sampling Techniques
By
–
I also can't resist recommending LLM sampler again. It's a great website to visualize the effect of different techniques. https://
artefact2.github.io/llm-sampling/i
ndex.xhtml
… -
Top-k and Top-p Sampling Strategies for LLMs Explained
By
–
There's been a lot of interest around sampling recently with entropix.
— Maxime Labonne @ ICLR (@maximelabonne) 29 octobre 2024
I just uploaded my first LLM blog post to HF, which talked about top-k and top-p sampling with easy-to-understand visualizations and code.
📝 Article: https://t.co/cgoP59fED4 pic.twitter.com/DJ45ToDtPnThere's been a lot of interest around sampling recently with entropix. I just uploaded my first LLM blog post to HF, which talked about top-k and top-p sampling with easy-to-understand visualizations and code. Article: https://
huggingface.co/blog/mlabonne/
decoding-strategies
… -
GitHub Copilot Adds Claude 3.5 and Gemini 1.5 Pro
By
–
Github Copilot got support for Claude 3.5 (new) and Gemini 1.5 Pro models besides previously available models from OpenAI.
— 🚨 AI News | TestingCatalog (@testingcatalog) 29 octobre 2024
A collab you would not expect ❤️ https://t.co/OEsaiefUNX pic.twitter.com/E1SqCST6C2Github Copilot got support for Claude 3.5 (new) and Gemini 1.5 Pro models besides previously available models from OpenAI. A collab you would not expect
-
Building Agentic Apps: Insights from Industry Leaders
By
–
How do you build an agentic app? We sat down with companies leading the charge on agents (Replit, Perplexity, Ramp and SuperHuman) to better understand how they actually shipped agents. Talked about:
UX
Cognitive Architecture
Prompting
Evals https://
langchain.com/breakoutagents