ERRAND: Budgeted Maintenance of Agent Memory

Abstract:Deployed agents run on handed-over knowledge: a frozen policy consults a briefing of consolidated items written before the stream begins. The world then moves while the store stands still: paths close, flags change, price bands move; every item was true at handover, and the failure is staleness, not ignorance. We introduce ERRAND, which treats revalidation as a priced errand: a recheck competes with the task it protects for the same scarce actions, funded only when the value per action of resolving a doubt clears a running wage. The errand index is single-peaked, vanishing at both ends of belief, so certainty in either direction costs nothing; free en-route receipts maintain on-path knowledge, and repair writes a version, never a deletion. Under equal action budgets in two drifting tool-use worlds, ERRAND clears every non-oracle policy on the preregistered calibers, primary in every setting and conditional at every binding budget, leading eager revalidation by 10.0pp at the base cap. Restraint wins: given no cap, ERRAND stops on its own, spending 11.0% of steps, while uncapped eager revalidation spends 70.7% and still finishes 4.5pp behind capped ERRAND. The margin sits where the briefing's coverage is thinnest, the shadow price of long-tail knowledge: a small budget, well priced, beats a bigger store that never rechecks.

Artificial Intelligence (cs.AI)

arXiv:2609.29545 [cs.AI]

(or arXiv:2609.29545v1 [cs.AI] for this version)

https://doi.org/10.48550/arXiv.2609.29545 Focus to learn more arXiv-issued DOI via DataCite (pending registration)

release/README.md

release/arms/__init__.py

release/arms/base.py

release/arms/bigmem.py

release/arms/blind_belief.py

release/arms/eager.py

release/arms/fixed_probe.py

release/arms/frozen.py

release/arms/gated_comparator.py

release/arms/index_ablations.py

release/arms/literature.py

release/arms/myopic.py

release/arms/no_regime_return.py

release/arms/oracle_filter.py

release/arms/priced.py

release/arms/random_errand.py

release/arms/registry.py

release/arms/whittle.py

release/audit/__init__.py

release/audit/dual_domain.py

release/audit/flip_intercept.py

release/audit/stratified.py

release/audit/whitelist.py

release/errand/__init__.py

release/errand/beliefs.py

release/errand/comparator.py

release/errand/config.py

release/errand/gate.py

release/errand/grades.py

release/errand/index.py

release/errand/lifecycle.py

release/errand/scheduler.py

release/errand/store.py

release/errand/voi.py

release/errand/wage.py

release/harness/__init__.py

release/harness/fingerprint.py

release/harness/ledger.py

release/harness/pairing.py

release/harness/runner.py

release/harness/seeds.py

release/harness/smoke.py

release/metrics/__init__.py

release/metrics/bootstrap.py

release/metrics/calibers.py

release/metrics/domains.py

release/metrics/paired_t.py

release/metrics/spend.py

release/metrics/stale_share.py

release/pyproject.toml

release/scripts/__init__.py

release/scripts/export_atoms.py

release/scripts/run_lifeboat.py

release/scripts/run_numerics.py

release/scripts/run_wave.py

release/world/__init__.py

release/world/agentodyssey_rules.py

release/world/atoms.py

release/world/dispatch.py

release/world/drift.py

release/world/orders.py

release/world/receipts.py

release/world/topology.py

release/world/truth_log.py

(59 additional files not shown) You must enabled JavaScript to view entire file list.

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.