How Small Models Learn Tool Use
Frontier AI Seminar
One of the clearest engineering trends in agentic AI today is the growing use of small language models (SLMs) inside agent loops.
This is not because SLMs outperform large models across the board. Rather, many agent workloads place real constraints on latency, cost, deployment footprint, and controllability—constraints under which smaller models are often a practical choice, especially when the core task is tool invocation rather than long-form generation.
Last December right before many of us slowed down to enjoy the holidays with families, Amazon Web Services released a study examining how SLMs perform on tool-calling tasks.
What makes this study a big deal is not the headline numbers, but the fact that a major technology company chose to publicly document, in detail, how a small language model can be trained, stabilized, and evaluated for tool use in a production-oriented setting.
Much of the public discussion so far has focused on results.
In the next Frontier seminar, I want to focus on the mechanics.
Specifically, I will address five important questions motivated by this study:
SLM: What is OPT-350M, the small language model used by AWS?
SFT: What is ToolBench-style Supervised Fine Tuning?
ToolBench: How is ToolBench data turned into training examples?
Loss: What loss is optimized during tool-use training?
Stability: Why does this setup remain stable in a single epoch?
I’m also excited to share that this session will feature a special guest: Dr. Xu Han from Amazon, who will share her first-hand experience training small language models for one of Amazon’s key products.
Curious which product? You’ll find out during the seminar which product she has been responsible for—and what practical lessons emerged from training SLMs in a production setting.
The goal of this seminar is not to argue that SLMs replace large models, but to understand why SLMs are increasingly used in agentic systems, and what that implies for training, evaluation, and deployment.
As always, I’ll teach this by hand ✍️.
👉 Sign up for the next Frontier seminar here: https://byhand.ai/seminar
