Credo: Declarative Control of LLM Pipelines via Beliefs and Policies

Duo Lu, Andrew Crotty, Uğur Çetintemel · 2026

Read Credo: Declarative Control of LLM Pipelines via Beliefs and Policies by Duo Lu, Andrew Crotty, Uğur Çetintemel (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-14401.txt.

Description

Lu, Crotty, and Çetintemel introduce Credo, a control layer for language-model pipelines that stores semantic state as beliefs and steers behavior with declarative policies over those beliefs. Imperative loops, fleeting memory, and prompt-buried logic make agent conduct opaque and brittle, so Credo instead routes execution via a database-backed semantic control plane. Beliefs and policies then guide choices like model selection, retrieval, and corrective re-execution without touching pipeline code. A decision-control demo shows the design adapting as evidence arrives and earlier conclusions change.

Catalog record

CREATORDuo Lu, Andrew Crotty, Uğur Çetintemel
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE19206
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.14401

Browse the full library · About the Tower of Babel