Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2603.11679 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911702912598016 |
|---|---|
| author | Demirel, Ilker Shi, Lawrence Hussain, Zeshan Sontag, David |
| author_facet | Demirel, Ilker Shi, Lawrence Hussain, Zeshan Sontag, David |
| contents | As real-world datasets become more complex and heterogeneous, supervised learning is often bottlenecked by input representation design. Modeling multimodal data, such as time-series, free text, and structured records, often requires non-trivial domain expertise. We propose an agentic pipeline to streamline this process. First, an LLM analyzes a small but diverse subset of text-serialized input examples in-context to synthesize a global rubric, which acts as a programmatic specification for extracting and organizing evidence. This rubric is then used to transform naive text-serializations of inputs into a more standardized format for downstream models. We also describe local rubrics, which are task-conditioned interpretive summaries generated by an LLM. Across 15 clinical tasks from the EHRSHOT benchmark, our rubric approaches significantly outperform count-feature models, naive LLM baselines, and a clinical foundation model pretrained on orders of magnitude more data. Beyond performance, rubrics offer operational advantages such as being easy to audit, cost-effectiveness at scale, and facilitating tabular representations. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_11679 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | LLMs can construct powerful representations and streamline sample-efficient supervised learning Demirel, Ilker Shi, Lawrence Hussain, Zeshan Sontag, David Artificial Intelligence As real-world datasets become more complex and heterogeneous, supervised learning is often bottlenecked by input representation design. Modeling multimodal data, such as time-series, free text, and structured records, often requires non-trivial domain expertise. We propose an agentic pipeline to streamline this process. First, an LLM analyzes a small but diverse subset of text-serialized input examples in-context to synthesize a global rubric, which acts as a programmatic specification for extracting and organizing evidence. This rubric is then used to transform naive text-serializations of inputs into a more standardized format for downstream models. We also describe local rubrics, which are task-conditioned interpretive summaries generated by an LLM. Across 15 clinical tasks from the EHRSHOT benchmark, our rubric approaches significantly outperform count-feature models, naive LLM baselines, and a clinical foundation model pretrained on orders of magnitude more data. Beyond performance, rubrics offer operational advantages such as being easy to audit, cost-effectiveness at scale, and facilitating tabular representations. |
| title | LLMs can construct powerful representations and streamline sample-efficient supervised learning |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2603.11679 |