Demonstrating CatDB: LLM-based Generation of Data-centric ML Pipelines
Bibliographic record
Abstract
AutoML systems automate finding Machine Learning (ML) pipelines but struggle to scale with large datasets due to time-consuming data analysis and complex hyper-parameter search spaces. LLMs (Large Language Models) offer flexibility and scalability for code generation with strong generalization across coding tasks. However, generating data-centric ML pipeline scripts is more challenging, as it requires complex reasoning to align the needs of a dataset with coding tasks, such as data cleaning or feature transformation. Thus, LLMs struggle to generate effective and efficient ML pipelines. This demo paper presents CatDB, which overcomes these challenges by dynamically generating dataset-specific instructions to guide LLMs in generating effective pipelines. CatDB profiles datasets to extract metadata, including refined data catalog information and statistics, and then uses this metadata to break down pipeline generation into instructions of tasks such as data cleaning, transformation, and model training, tailored to specifics of the dataset at hand. This process enables CatDB to leverage LLM coding capabilities more effectively. Our evaluation shows CatDB outperforms existing LLM-based and AutoML systems with up to orders of magnitude faster runtime on large datasets. The audience will experience CatDB's capabilities with commercial and open-source LLMs, using a variety of real datasets, as shown in our demo video and Colab notebook.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.016 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.003 | 0.005 |
| Open science | 0.004 | 0.005 |
| Research integrity | 0.001 | 0.003 |
| Insufficient payload (model declined to judge) | 0.021 | 0.013 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".