Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deterministic Legal Agents: A Canonical Primitive API for Auditable Reasoning over Temporal Knowledge Graphs

Hudson de Martim

TL;DR

The paper addresses the lack of determinism and auditability in RAG-based legal AI by introducing a Canonical Primitive API built atop a temporal SAT-Graph. It defines a library of atomic, composable primitives for discovery, deterministic fetch, navigation, causality, and introspection, enabling planner-guided agents to produce auditable execution plans. By decoupling knowledge representation from reasoning, the approach delivers a verifiable audit log of the agent’s steps and results, improving trust in high-stakes domains. The work demonstrates use cases for deterministic point-in-time retrieval, causal pinpointing with version diffs, and thematic-hierarchical impact analysis, and discusses implications and future directions for extending auditable reasoning beyond law.

Abstract

For autonomous legal agents to operate safely in high-stakes domains, they require a foundation of absolute determinism and auditability-guarantees that standard Retrieval-Augmented Generation (RAG) frameworks cannot provide. When interacting with temporal knowledge graphs that model the complex evolution of legal norms, agents must navigate versioning, causality, and hierarchical structures with precision, a task for which black-box vector search is ill-suited. This paper introduces a new architectural pattern to solve this: a formal Primitive API designed as a secure execution layer for reasoning over such graphs. Instead of a monolithic query engine, our framework provides a library of canonical primitives-atomic, composable, and auditable primitives. This design empowers planner-guided agents to decompose complex legal questions into transparent execution plans, enabling critical tasks with full verifiability, including: (i) precise point-in-time version retrieval, (ii) robust causal lineage tracing, and (iii) context-aware hybrid search. Ultimately, this architecture transforms opaque retrieval into auditable reasoning, turning the agent's internal process from a black box into a verifiable log of deterministic primitives and providing a blueprint for building the next generation of trustworthy legal AI.

Deterministic Legal Agents: A Canonical Primitive API for Auditable Reasoning over Temporal Knowledge Graphs

TL;DR

The paper addresses the lack of determinism and auditability in RAG-based legal AI by introducing a Canonical Primitive API built atop a temporal SAT-Graph. It defines a library of atomic, composable primitives for discovery, deterministic fetch, navigation, causality, and introspection, enabling planner-guided agents to produce auditable execution plans. By decoupling knowledge representation from reasoning, the approach delivers a verifiable audit log of the agent’s steps and results, improving trust in high-stakes domains. The work demonstrates use cases for deterministic point-in-time retrieval, causal pinpointing with version diffs, and thematic-hierarchical impact analysis, and discusses implications and future directions for extending auditable reasoning beyond law.

Abstract

For autonomous legal agents to operate safely in high-stakes domains, they require a foundation of absolute determinism and auditability-guarantees that standard Retrieval-Augmented Generation (RAG) frameworks cannot provide. When interacting with temporal knowledge graphs that model the complex evolution of legal norms, agents must navigate versioning, causality, and hierarchical structures with precision, a task for which black-box vector search is ill-suited. This paper introduces a new architectural pattern to solve this: a formal Primitive API designed as a secure execution layer for reasoning over such graphs. Instead of a monolithic query engine, our framework provides a library of canonical primitives-atomic, composable, and auditable primitives. This design empowers planner-guided agents to decompose complex legal questions into transparent execution plans, enabling critical tasks with full verifiability, including: (i) precise point-in-time version retrieval, (ii) robust causal lineage tracing, and (iii) context-aware hybrid search. Ultimately, this architecture transforms opaque retrieval into auditable reasoning, turning the agent's internal process from a black box into a verifiable log of deterministic primitives and providing a blueprint for building the next generation of trustworthy legal AI.

Paper Structure

This paper contains 50 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. AI in Legal Information Retrieval
  4. Retrieval-Augmented Generation and its Limitations
  5. Knowledge Graphs for Retrieval (Graph RAG)
  6. Agentic Frameworks for Reasoning and Retrieval
  7. Positioning Our Contribution
  8. Specification of the Canonical Primitive API
  9. Design Principles
  10. Core Data Models
  11. Architectural Note: Separation of Normative and Conceptual Entities
  12. Canonical Primitive API Specification
  13. Bridging Baseline RAG with a Temporal Knowledge Graph
  14. Discovery and Search Primitives
  15. Deterministic Fetch Primitives
  16. ...and 35 more sections

Figures (3)

  • Figure 1: From "flat-text" RAG to SAT Graph RAG
  • Figure 2: Diagram illustrating a reasoning agent decomposing a user prompt into tasks that are executed through primitives provided by the SAT-Graph API.
  • Figure 3: SAT-Graph Ontology