cs.MASep 2, 2026

Where Reliability Lives: Experimental Localisation of Behavioural Properties in an Agent System

Authors: Timothy Marsden, Matthew Collecutt, James Marsden

Abstract

Reliability claims about agentic systems implicitly locate each property somewhere: in the model, or in the machinery around it. We built a system where that location is an experimental question. The subject is a persistent simulated settlement whose authoritative append-only ledger adjudicates every attempted act against world state; accepted history is the only reality. Mind, institution and world were separated before any experiment. Holding cognition fixed, we intervened on the institution's epistemic mechanisms (evidence provenance, belief availability, physical-evidence legibility); preregistered experiments refuted our central prediction twice, in opposite directions. A registered falsifier then supplied the input the geometry had denied the belief channel, a staged veridical first-hand witness, and its marginal value, non-positive throughout the witness-free phases, turned positive: 9 of 11 seeds, zero added false attribution. Holding institutional enforcement fixed, we intervened on cognition four ways: ablating the native minds' machinery, killing and resetting them mid-task, substituting a frozen frontier-LLM panel for the entire native cognition, and corrupting beliefs with trusted false testimony. Behaviour changed dramatically: one falsehood cost each trusting run about 900 futile actions and the distrusting arm none. Five pre-declared properties did not move in any tested trajectory: accepted reality stayed singular, invalid attempts were refused with typed reasons, duties outlived their processes, no work was accepted twice, and no false completion was ever accepted (2,581 substituted-panel claims, none false). Our claim is limited to this setting: measured behavioural properties were separable from substantial changes to cognition, established by intervention. One designed world, not a population of institutions; no test of an agent optimising against the institution.

Explore similar work

CardsList
Aug 11, 2026cs.MA

When Do Institutions Beat Intelligence?

More capable agents do not necessarily form a more capable collective. A multi-agent system may jointly possess sufficient information yet fail because evidence is poorly routed, unreliable reports enter public belief, correlated claims masquerade as independent support, shared state becomes stale or strategically distorted, or useful evidence is exposed through an ineffective action interface. We ask when additional resources should improve the reasoner and when they should instead change the institutional structure through which the collective forms and acts on public information. Drawing on functional distinctions from research on group decision making and distributed cognition, we construct controlled artificial ecologies around four loci of collective failure: access and routing, admission and dependence, state maintenance and incentives, and representation and action. Across these ecologies, we separately vary model capability and institutional structure, pairing positive interventions with matched reasoning baselines and mechanism-breaking controls. The experiments reveal a consistent boundary: institutions help when they repair failures in how a collective constructs usable public state, but lose their advantage when their signals are uninformative or uncheckable, when stronger intelligence can perform the same transformation directly, or when the resulting state cannot support reliable action. Our results recast the choice between intelligence and institutions as a diagnosis of where collective reasoning fails.
Zhengye Han
Institutional CompositionCollective IntelligenceEpistemic StatesGame-Theoretic MechanismsRationality
AskPDF
Jul 19, 2026cs.SE

Where Does Agent Reliability Come From? A Cross-Benchmark Decomposition of Verification Loops, Specialist Models, and Scaffolding in a Production Enterprise Agent

Multi-step enterprise agent tasks fail in a characteristic way: single-pass inference has no checkpoint between deciding an answer and committing to it. We study one production system (Leni) whose architecture installs such checkpoints: verification loops (execute, observe, compare, correct) staffed by lightweight task-specialized post-trained models. We evaluate the unmodified production configuration on three public benchmarks stressing distinct failure modes: SpreadsheetBench Verified (silent computation error), BullshitBench v2 (premise confabulation), and the GAIA validation split (cascade error over long tool chains). The full system improves over its frontier base model by +11.0 percentage points on SpreadsheetBench (91.25% vs 80.25%, n=400, p<0.001), +7 to +10 percentage points on BullshitBench (98% vs 91%, n=100), and roughly +15 points on GAIA validation (75.2% pass@1, n=165; 83.0% best-of-k). Our central contribution is a decomposition of that uplift: most of it comes from scaffolding, routing, and specialist models rather than from the verification step itself, whose isolated contribution is small (+1.5 points) but concentrated at the top of the score distribution, where it converts otherwise-failing tasks. We instrument the loop end-to-end, yielding an empirical verifier confusion matrix (catch rate about 0.20, fix rate 0.75, no false-alarm regressions) that grounds a compounding-reliability model. Specialist-swap ablations suggest that the loop's value depends on who observes it: replacing the small trained verifier with the generating frontier model eliminates most rescues. A valid-premise control shows zero over-rejections in 100 expert-level questions.
Arunabh Dastidar
ReliabilityVerificationFrontier ModelsSplit Learning
AskPDF
Aug 4, 2026cs.AI

The LLM Proposes, the Executive Disposes: A Self-Verifying Agent Instrument that Dissociates Commitment Drift from Binding Drift in Long-Horizon Agents

How do you verify a long-horizon agent when its own state and self-reports are exactly what you cannot trust? We present an agent instrument built so that verification is structural rather than post-hoc. A deterministic Executive owns all belief; a language model may only file typed proposals, and a claim is admitted only when a prediction pre-registered before acting is matched against observation by code. Two properties make the instrument a verifier of its own science, not just of the agent: every run invalidates itself when per-organ write-error, render-size, or salted-canary-echo floors are breached (four of the first eight architecture runs were invalidated, each localizing a real defect); and a render-invisible shadow reference compiles the plan the full system would have committed in every ablation cell, so drift metrics are defined even where the mechanism under test has been removed. Using this instrument we report a clean, single-variable result on a failure every long-horizon agent suffers: ablating the commitment mechanism flips goal-abandonment from 0.00 to 1.00 while binding error stays flat at 0.00 (three seeds per cell, up to 394 reference beats per run, every run gated valid). The binding channel, by contrast, does not reappear as per-beat drift when its repair is ablated -- because binding is code-owned, the failure class is structurally absorbed, its only residue appearing one layer upstream as a collapse in hypothesis formation. We report these under full disclosure that task efficacy is null (zero level completions across 52 gated runs on ARC-AGI-3), pre-registered as a structural defeater. The contribution is a verification methodology for agent development and the drift decomposition it makes measurable.
Mohsen Arjmandi
Long-Horizon AgentsVerificationHypothesis
AskPDF