aiqbench analysis
This paper proposes Probabilistic Agentic Programming, which makes an LLM agent's mandate, authority, side effects, and audit evidence explicit in code rather than scattered across prompts and configuration. Its safety claims are conditional: the inductive proofs hold only if the guards are correct and mediation is complete, and the seven-case exploratory study measures task completion, not safety under adversarial conditions. Readers building agent systems should treat the mediation-kernel approach as a design pattern to evaluate, and the proofs and empirical results as scoped to their stated assumptions.
Responsibility Boundaries for Agentic Systems: A Comparative Evaluation of Probabilistic Agentic Programming
arXiv:2610.33312v1 Announce Type: cross Abstract: Language-model agents combine reasoning, tools, memory, delegation, and external effects. Authority is often distributed across prompts, host code, configuration, and logs. Probabilistic Agentic Programming (PAP) makes responsibility boundaries explicit programming abstractions connecting mandate, authority, effects, and evidence. We compare PAP with programming paradigms, agent-oriented and BDI systems, workflow and policy languages, and LLM tool-use frameworks along seven dimensions. A mediation kernel formalizes checks before tool execution, operation-bound approval, resource limits, and evidence. Inductive proofs establish conditional safety for arbitrary model responses, assuming correct guards and complete mediation. A backward simulation matches graph executions to source executions in a defined sequential Touba subset. Its human checkpoints release program continuations; operation-bound approval remains a separate kernel contract. Finite-state exploration checks the model. Scripted enforcement tests exercise selected refusal paths and permitted controls. An exploratory study combines seven business cases, including two three-agent workflows, seven models from three providers, five methods, and five repetitions per case and method. Shared English instructions and exact checks of stored outcomes yield 1,225 assessed observations. Native Touba completes all 35 observations on six models and 30 on the remaining model; structured Touba ranges from 21 to 35. Native matches the highest observed count for each model, including ties with the frameworks. All paired case intervals include zero. These observed counts describe the evaluated tasks; the runtime authority boundary remains fixed. The results connect responsibility contracts, execution rules, and evidence.
Abstract:Language-model agents combine reasoning, tools, memory, delegation, and external effects. Authority is often distributed across prompts, host code, configuration, and logs. Probabilistic Agentic Programming (PAP) makes responsibility boundaries explicit programming abstractions connecting mandate, authority, effects, and evidence. We compare PAP with programming paradigms, agent-oriented and BDI systems, workflow and policy languages, and LLM tool-use frameworks along seven dimensions. A mediation kernel formalizes checks before tool execution, operation-bound approval, resource limits, and evidence. Inductive proofs establish conditional safety for arbitrary model responses, assuming correct guards and complete mediation. A backward simulation matches graph executions to source executions in a defined sequential Touba subset. Its human checkpoints release program continuations; operation-bound approval remains a separate kernel contract. Finite-state exploration checks the model. Scripted enforcement tests exercise selected refusal paths and permitted controls. An exploratory study combines seven business cases, including two three-agent workflows, seven models from three providers, five methods, and five repetitions per case and method. Shared English instructions and exact checks of stored outcomes yield 1,225 assessed observations. Native Touba completes all 35 observations on six models and 30 on the remaining model; structured Touba ranges from 21 to 35. Native matches the highest observed count for each model, including ties with the frameworks. All paired case intervals include zero. These observed counts describe the evaluated tasks; the runtime authority boundary remains fixed. The results connect responsibility contracts, execution rules, and evidence.
Software Engineering (cs.SE); Artificial Intelligence (cs.AI); Programming Languages (cs.PL)
arXiv:2610.33312 [cs.SE]
(or arXiv:2610.33312v1 [cs.SE] for this version)
https://doi.org/10.48550/arXiv.2610.33312 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
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.
Full text under the publisher's open licence · 332 words
Related coverage
Quantum Dice and North Wales Police Test Probabilistic Computing for Emergency Response
The Quantum Insider · Sep 15, 2026
TRACER: Adaptive Multi-Robot Social Navigation via Joint Human-Response Prediction and Interaction-Aware Replanning
arXiv · Sep 17, 2026
Using Probabilistic Programs to Train Inductive Reasoning in Large Language Models
arXiv · Jun 10, 2026
Clarify, Abstain or Answer? Strategising in Conversation with Belief-Augmented Generation
arXiv · Sep 9, 2026