{"paper":{"title":"APWA: A Distributed Architecture for Parallelizable Agentic Workflows","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"APWA breaks complex agentic tasks into independent subproblems that run in parallel without communication.","cross_cats":["cs.DC","cs.MA"],"primary_cat":"cs.AI","authors_text":"Alina Oprea, Cristina Nita-Rotaru, Evan Rose, Matthew D. Laws, Tushin Mallick","submitted_at":"2026-05-14T17:40:20Z","abstract_excerpt":"Autonomous multi-agent systems based on large language models (LLMs) have demonstrated remarkable abilities in independently solving complex tasks in a wide breadth of application domains. However, these systems hit critical reasoning, coordination, and computational scaling bottlenecks as the size and complexity of their tasks grow. These limitations hinder multi-agent systems from achieving high-throughput processing for highly parallelizable tasks, despite the availability of parallel computing and reasoning primitives in the underlying LLMs. We introduce the Agent-Parallel Workload Archite"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"In our evaluation, we demonstrate that APWA can dynamically decompose complex queries into parallelizable workflows and scales on larger tasks in settings where prior systems fail completely.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That complex agentic workflows can be reliably decomposed into non-interfering subproblems that require no cross-communication and can be executed independently on heterogeneous resources.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"APWA is a distributed multi-agent architecture that decomposes parallelizable agentic workflows into non-interfering subproblems for scalable execution on heterogeneous resources.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"APWA breaks complex agentic tasks into independent subproblems that run in parallel without communication.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"2a715faf45f71c4b27f46b68234afa92ccbff58c4ffc9922b56a364cea5020fd"},"source":{"id":"2605.15132","kind":"arxiv","version":1},"verdict":{"id":"cbd5c22d-b836-4b48-a441-c784b0f8fa2e","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T03:07:47.800391Z","strongest_claim":"In our evaluation, we demonstrate that APWA can dynamically decompose complex queries into parallelizable workflows and scales on larger tasks in settings where prior systems fail completely.","one_line_summary":"APWA is a distributed multi-agent architecture that decomposes parallelizable agentic workflows into non-interfering subproblems for scalable execution on heterogeneous resources.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That complex agentic workflows can be reliably decomposed into non-interfering subproblems that require no cross-communication and can be executed independently on heterogeneous resources.","pith_extraction_headline":"APWA breaks complex agentic tasks into independent subproblems that run in parallel without communication."},"references":{"count":68,"sample":[{"doi":"","year":2024,"title":"PII masking 300k dataset","work_id":"d7f51784-15a0-4eec-9c05-f5602d7e9177","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Hadoop.https://hadoop.apache.org/","work_id":"5d5ded8f-4cf6-43fd-85e0-2205545839cd","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Schema-driven information extraction from heterogeneous tables","work_id":"4e98478a-e868-4bc3-bacc-ba1c6bceca41","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Boiko, Robert MacKnight, Ben Kline, and Gabe Gomes","work_id":"06c818e9-9eb4-4be4-9f23-dba80fb8bf3a","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1901,"title":"Language models are few-shot learners","work_id":"e5d2dfec-d7bd-4be4-9e02-3244247b810a","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":68,"snapshot_sha256":"4f1c1ba604e27d6d3d6a0c34b5aa0ac2d8e01d34f45fe34ce41b86ff8a4c4951","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}