> For the complete documentation index, see [llms.txt](https://docs.quantaria.xyz/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.quantaria.xyz/deep-dives/evaluation-methodology.md). # Evaluating against matched dense baselines A claim that a tensor-network model is smaller or better only means something against a fair comparison. This page explains the evaluation method Quantaria commits to, and what each metric measures. ## Matched baselines Every Quantaria model is compared with a dense transformer of the same size, trained on the same data. Matching the size means the comparison isolates the effect of the tensor-network layers, rather than differences in scale. ## Metrics * **Loss curves.** Training and validation loss over time. Shows whether the model learns and whether it overfits. * **Perplexity.** How well the model predicts held-out text. Lower is better. * **FID.** Image quality against real images. Lower is better. * **Memory use.** Peak memory for training and inference. This is where tensor-network layers should show their advantage. ## Why all four A smaller model with worse loss is not an improvement. A lower perplexity with much higher memory is also not an improvement. Reporting the four together shows the trade-off. ## Reporting rules * State the model size, dataset, and training budget for every number. * Publish results whether they favour the tensor-network model or not. * Separate paper results from Quantaria results. See [Research Foundations](https://github.com/bhavy2908/dataai_docs/tree/main/architecture/research-foundations.md#how-to-read-these-numbers). ## Common traps * **Scale mismatch.** A comparison at different sizes says little about the method. * **Cherry-picked settings.** Reporting the best bond dimension without the others hides the trade-off. * **Toy results presented as general.** A result on a small model and small dataset applies to that setting. Say so. ## Quantaria's use Hilbert-1 Nano is trained on TinyStories text and CIFAR-scale images and compared against a dense transformer of the same size. Benchmarks are planned for each model as it is evaluated. See [Roadmap](https://github.com/bhavy2908/dataai_docs/tree/main/overview/roadmap.md). ## Related * [Features](https://github.com/bhavy2908/dataai_docs/tree/main/product/features.md#evaluation-commitments) * [FAQ](https://github.com/bhavy2908/dataai_docs/tree/main/faq.md) --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://docs.quantaria.xyz/deep-dives/evaluation-methodology.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.