Research

Paper

AI LLM March 12, 2026

LLMs can construct powerful representations and streamline sample-efficient supervised learning

Authors

Ilker Demirel, Larry Shi, Zeshan Hussain, David Sontag

Abstract

As real-world datasets become increasingly complex and heterogeneous, supervised learning is often bottlenecked by input representation design. Modeling multimodal data for downstream tasks, such as time-series, free text, and structured records, often requires non-trivial domain-specific engineering. 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 summaries generated by an LLM. Across 15 clinical tasks from the EHRSHOT benchmark, our rubric-based approaches significantly outperform traditional count-feature models, naive text-serialization-based LLM baselines, and a clinical foundation model, which is pretrained on orders of magnitude more data. Beyond performance, rubrics offer several advantages for operational healthcare settings such as being easy to audit, cost-effectiveness to deploy at scale, and they can be converted to tabular representations that unlock a swath of machine learning techniques.

Metadata

arXiv ID: 2603.11679
Provider: ARXIV
Primary Category: cs.AI
Published: 2026-03-12
Fetched: 2026-03-14 05:03

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.11679v1</id>\n    <title>LLMs can construct powerful representations and streamline sample-efficient supervised learning</title>\n    <updated>2026-03-12T08:44:06Z</updated>\n    <link href='https://arxiv.org/abs/2603.11679v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.11679v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>As real-world datasets become increasingly complex and heterogeneous, supervised learning is often bottlenecked by input representation design. Modeling multimodal data for downstream tasks, such as time-series, free text, and structured records, often requires non-trivial domain-specific engineering. 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 summaries generated by an LLM. Across 15 clinical tasks from the EHRSHOT benchmark, our rubric-based approaches significantly outperform traditional count-feature models, naive text-serialization-based LLM baselines, and a clinical foundation model, which is pretrained on orders of magnitude more data. Beyond performance, rubrics offer several advantages for operational healthcare settings such as being easy to audit, cost-effectiveness to deploy at scale, and they can be converted to tabular representations that unlock a swath of machine learning techniques.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <published>2026-03-12T08:44:06Z</published>\n    <arxiv:primary_category term='cs.AI'/>\n    <author>\n      <name>Ilker Demirel</name>\n    </author>\n    <author>\n      <name>Larry Shi</name>\n    </author>\n    <author>\n      <name>Zeshan Hussain</name>\n    </author>\n    <author>\n      <name>David Sontag</name>\n    </author>\n  </entry>"
}