How can we … help researchers build better language models?
18 August 2025
Large Language Models (LLMs), a core AI technology, are sometimes compared to black boxes in airplanes because their inner workings are hard to interpret. They can reflect unknown biases due to uncurated training datasets, lack transparency in their outputs, and are often developed and controlled by a small number of key players. Small Language Models (SLMs) offer a promising alternative. They tend to be cheaper to train and easier to adapt to specific use cases, and allow researchers to retain full control over the model’s data and weights. While sometimes seen as less powerful, well-trained SLMs can match or even outperform larger models in certain contexts.
