Principled context representation improves reasoning over large documents
Realize What Matters: Principled Context Representation for Large-Scale Reasoning
Computation and Language
Summary
AI systems often struggle to solve complex problems because important information is spread out across many documents beyond what they can easily process all at once. The authors suggest clear principles for how AI should organize and represent this scattered information so it can reason better. They built a system called R3Con that follows these principles and tested it against top existing methods. Their system performs much better, even enabling smaller AI models to outperform much larger ones, reducing the need for very large and expensive models.
What this means in practice
- •For legal technology developers: Enable AI systems to efficiently gather and reason over large volumes of legal documents to aid case analysis and decision support.
- •For financial analytics teams: Improve AI's ability to synthesize and reason about information from extensive, diverse financial reports and data sources.
Authors
Michael Theologitis, Dean Light, Shuyue Stella Li, Benjamin Newman, Yulia Tsvetkov, Dan Suciu
Abstract
Solving complex tasks in domains such as science, medicine, law, and finance often requires assembling interdependent information scattered across vast, heterogeneous sources far beyond model context limits. Existing approaches tackle this challenge by organizing information into more manageable representations over which models can reason, such as graphs, textual memories, and retrieval collections. These representations dictate what downstream reasoning is possible and, ultimately, whether it succeeds; yet their design and construction remain largely ad hoc. In this work, drawing on the cognitive theory of relevance realization, we propose concrete principles for designing AI systems that construct effective representations of very large contexts. We analyze existing approaches and show how their successes and failures map onto their alignment with these principles, and introduce R3Con, a harness designed to operationalize the principles more systematically. We evaluate R3Con against nine state-of-the-art baselines on two recent benchmarks of reasoning over large document corpora. On these benchmarks, R3Con substantially outperforms the strongest baseline, by $20$ and $8.4$ percentage points. It also enables smaller models to outperform much larger ones: R3Con with 4B and 9B models outperforms all evaluated 35B baselines, while R3Con with a 35B-A3B model outperforms Claude Code with Claude-Sonnet-5 at $3.7\times$ lower cost. Our results show that context representations following our principled approach can reduce reliance on model scale, pointing toward a future of AI systems with frontier-level performance powered by smaller models. Our code is available at https://github.com/michaeltheologitis/r3con