Filipe Gracio, PhD

Comparing models.. Finding that a small 1B #llm hallucinates less than a 4B for semantic to structural matching. Small llms for targeted tasks for the win. The new Meta llama3.2 1B is better than expected and cheap on token cost.

3

FunctionGemma, a new 270M parameter language model from Google, designed specifically for function calling and tool execution — not general chat.   Why is FunctionGemma special? 🔹 Extremely lightweight FunctionGemma can run in ~550MB RAM, making it possible to deploy on laptops, edge devices, and even phones — all without large GPUs. 🔹 Purpose-built for function calling Unlike general LLMs, this model is trained for text-only tool calling, making it ideal for automation, agents, and API-driven workflows. 🔹 Easy local deployment The model is released in GGUF format, so it works seamlessly with llama.cpp and Ollama using quantized versions (Q4/Q5/Q8). 🔹 Fine-tuning ready from day one Thanks to Unsloth’s day-zero support, you can: Run FunctionGemma on CPU or GPU Fine-tune it locally using Colab notebooks Choose full fine-tuning or LoRA   What can you fine-tune it for? FunctionGemma is meant to be customized for specific tasks, such as: Multi-turn function-calling agents Mobile actions (calendar events, reminders, flashlight, timers) Task-specific automation and workflows Lightweight, offline AI assistants This model clearly shows a shift toward small, focused, and efficient AI — where models are trained to do one thing well and can be adapted quickly to real-world use cases.   Model (GGUF): https://lnkd.in/gxVuBhZA Docs & fine-tuning guides: https://lnkd.in/gchpAjUF Run and Deploy LLMs on your iOS or AndroidPhone: https://lnkd.in/grk72Bey

5

Anthropic's recent blogpost discusses the important shift from evaluating simple LLMs ---> testing complex AI agents.  It highlights why standard benchmarks fall short for agentic workflows and offers a practical roadmap for building robust, multi-turn evaluations." Packed with insights, but couple of takeaways that stood out to me:  🎯 1. 𝐂𝐨𝐦𝐩𝐨𝐮𝐧𝐝𝐢𝐧𝐠 𝐯𝐚𝐥𝐮𝐞 𝐨𝐟 𝐭𝐡𝐞𝐬𝐞 𝐬𝐲𝐬𝐭𝐞𝐦𝐬: where costs are visible upfront but their benefits accumulate over time. (quicker iterations, faster deployment) 🎯 𝟐. 𝐓𝐡𝐞 𝐂𝐨𝐧𝐬𝐢𝐬𝐭𝐞𝐧𝐜𝐲 𝐅𝐢𝐱 / 𝐃𝐞𝐭𝐞𝐫𝐦𝐢𝐧𝐢𝐬𝐦 𝐢𝐧 𝐄𝐯𝐚𝐥𝐮𝐚𝐭𝐢𝐨𝐧𝐬: Regardless of agent type, agent behavior varies between runs. Sometimes, we want to measure how often (what proportion of the trials) an agent succeeds for a task, while for others, we want to measure if Agent succeeds atleast once. 🚀 Single success is enough across runs ? Use pass@k. 🚨 Need reliability across runs? Use pass^k. The post contains many more valuable insights worth exploring. Read more here: https://lnkd.in/gwCxqhDj

59

2 Comments

New paper from Yann LeCun. It's a great read as always and it comes out at the right time when I think thousands [of software engineers] are really getting into their coding agents and discovering glimpses of their static nature (forget to save a session?). I think part of the art of writing a high-impact paper is generating a reaction. I can't help but feel like the threat of a biologically inspired engineered optimization algorithm is more for the dramatic story telling. https://lnkd.in/eZAuqWkN

4

Is "Bigger" Always Better? Challenging Scaling Laws with a 19M Parameter Model Most progress in LLMs today is driven by one factor - scale. More parameters, more data, massive compute clusters. But what if we looked in the opposite direction? I am starting a new series on Vicky, a project exploring a core hypothesis: If we encode strong inductive biases directly into the architecture, a small model can outperform much larger LLMs in a narrow domain. Specifically, in ML/AI code generation. Instead of training a massive vanilla Transformer, I designed a specialized, parameter-efficient hybrid architecture. It doesn't try to know everything, it tries to be an expert in Python syntax and PyTorch logic. The Architecture: This isn't just a shrunk-down GPT-2. It utilizes: Factorized Token Embeddings: drastically reducing parameter waste in the vocabulary layer. Rotary Positional Encoding (RoPE): for better handling of relative positions in code structure. Low-Rank Attention: optimizing the "brain" of the model. SSM-like Convolutional Memory: adding a structural bias to track long-term dependencies efficiently. The goal is to force the model to "think like code" rather than brute-forcing statistical correlations. The Specs: Parameters: ~19M Hardware: Runs on a consumer edge GPU (4GB VRAM) Check out the diagram. #AIArchitecture #LLMResearch #EfficientAI #CodeGeneration #MachineLearning #DeepLearning #VickySLM

3

Streamline your machine learning workflows with CI/CD in Julia. From model validation to deployment, discover how to build scalable, test-driven, and reproducible ML systems using Julia’s powerful ecosystem. https://lnkd.in/esWGw8MT #JuliaLang #MLOps #CI_CD #MachineLearning #programming

17