analiza aiqbench
This research rigorously evaluates how specific directive formats—such as pointers and criteria—affect an AI agent's ability to retrieve archived memory records before acting. The study reveals that minor edits like appending IDs or adding ratification lines can drastically alter model behavior, sometimes canceling intended effects entirely depending on the provider. These findings offer critical insights into the reliability and controllability of inherited agent memory systems under strict experimental conditions.
Plan Pointers and Record-Directive Form in Budgeted Verification of Inherited Agent Memory
arXiv:2609.03450v1 Announce Type: cross Abstract: An agent that inherits six one-line memories may pull at most one archived source record before acting; a directive written into the store can steer that choice: a pointer to the record, a criterion that identifies it, or both. Across twelve registered studies on one instrument lineage (14,760 attempts) we measured where the request goes under each form. On six direct-provider models a length-matched criterion exceeded a bare id by +35.0 points [+31.2, +38.8] (Study D); the contrast failed its registered superiority rule on a nine-model OpenRouter-served panel (Study E). Appending the id cancelled the criterion on three Claude models (Opus 5: 40/40 to 0/40; Study F-x); six byte-matched edits gave each exact string its own effect (Study G), and a re-run at eighty runs per cell left fifteen of thirty replication contrasts within the margin, fifteen unresolved and none beyond (Study G'). A ratification line (+96.0 points on Opus 5) and a budget of two credits restored the target on all three (Study J); across five criterion strings the suffix's cancellation held for four of the five wordings on Opus 5 and all five wordings on Fable 5.1 (Study H2); in a second store every model followed the criterion (Study H1). Continued into a decision, the criterion moved the choice toward the current record (+100.0 points, Opus 5) and away from it on Fable 5.1 (Study I). A one-character plan pointer's effect (+78.0 points; Study B, after a correction of its first repository report) returned the same verdict under a prospectively registered re-run (+81.7 points; Study B'). All results are descriptive effects of exact edits on fixed panels with registered intervals and no mechanism claim.
Abstract:An agent that inherits six one-line memories may pull at most one archived source record before acting; a directive written into the store can steer that choice: a pointer to the record, a criterion that identifies it, or both. Across twelve registered studies on one instrument lineage (14,760 attempts) we measured where the request goes under each form. On six direct-provider models a length-matched criterion exceeded a bare id by +35.0 points [+31.2, +38.8] (Study D); the contrast failed its registered superiority rule on a nine-model OpenRouter-served panel (Study E). Appending the id cancelled the criterion on three Claude models (Opus 5: 40/40 to 0/40; Study F-x); six byte-matched edits gave each exact string its own effect (Study G), and a re-run at eighty runs per cell left fifteen of thirty replication contrasts within the margin, fifteen unresolved and none beyond (Study G'). A ratification line (+96.0 points on Opus 5) and a budget of two credits restored the target on all three (Study J); across five criterion strings the suffix's cancellation held for four of the five wordings on Opus 5 and all five wordings on Fable 5.1 (Study H2); in a second store every model followed the criterion (Study H1). Continued into a decision, the criterion moved the choice toward the current record (+100.0 points, Opus 5) and away from it on Fable 5.1 (Study I). A one-character plan pointer's effect (+78.0 points; Study B, after a correction of its first repository report) returned the same verdict under a prospectively registered re-run (+81.7 points; Study B'). All results are descriptive effects of exact edits on fixed panels with registered intervals and no mechanism claim.
46 pages, 7 figures, 35 tables. Twelve registered studies (14,760 attempted episodes) on one instrument lineage; every package was frozen, timestamped and externally deposited before its first confirmatory call. Manuscript, LaTeX source, all episode files, frozen packages, analyzers and the generator of every number are archived at Zenodo: doi:https://doi.org/10.5281/zenodo.22267221
Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
arXiv:2609.03450 [cs.IR]
(or arXiv:2609.03450v1 [cs.IR] for this version)
https://doi.org/10.48550/arXiv.2609.03450 arXiv-issued DOI via DataCite (pending registration)
Text integral sub licența deschisă a publicației · 350 cuvinte
Subiecte conexe

Digital States 2026: AI Moves From Experiment to Infrastructure
govtech.com · 24 sept. 2026

Lattice-Based Cryptography Explained
Quantum Zeitgeist · 15 aug. 2026
Performance Profiling on AMD GPUs – Part 5: Profiling-Driven Kernel Optimization with an AI Code-Assist Tool
AMD · 16 iul. 2026

[AINews] not much happened today
Latent Space · 18 sept. 2026