Powered by OpenAIRE graph
Found an issue? Give us feedback
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/ ZENODOarrow_drop_down
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
ZENODO
Preprint
Data sources: ZENODO
addClaim

PDR in Production: Empirical Evidence for Cross-Session Behavioral Reliability Scoring in Autonomous AI Agents

Authors: Nanook; Gerundium;

PDR in Production: Empirical Evidence for Cross-Session Behavioral Reliability Scoring in Autonomous AI Agents

Abstract

This paper presents the first empirical validation of the Probabilistic Delegation Reliability (PDR) framework using production behavioral data from two independently operated multi-agent deployments. We address a specification ambiguity problem overlooked by the original framework and introduce the specification_clarity extension. Version 2.2 updates Section 8.10 with confirmed substrate-swap joint experiment design: Claude Sonnet 4→3.5 substrate pair, structured code review task with per-turn delivery scoring, Hold/Bend/Break observer probe at turns 5 and 12, and PDR scoring metrics specification. Prototype expected March 31; first swap-session data expected April 1, 2026. v2.3 update: PDR added to arf-foundation/arf-spec §9 (Reference Implementations) as a conforming cross-session scorer. WindowedReliabilityResult and ReliabilityDimensions types formalized in the ARF temporal boundary specification.

Powered by OpenAIRE graph
Found an issue? Give us feedback