{"paper":{"title":"A Two-Dimensional Framework for AI Agent Design Patterns: Cognitive Function and Execution Topology","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A 7-by-6 matrix of cognitive functions and execution topologies classifies 27 distinct AI agent design patterns.","cross_cats":["cs.MA","cs.SE"],"primary_cat":"cs.AI","authors_text":"Jia Huang, Joey Tianyi Zhou","submitted_at":"2026-03-16T04:01:01Z","abstract_excerpt":"Existing frameworks for LLM-based agent architectures describe systems from a single perspective: industry guides (Anthropic, Google, LangChain) focus on execution topology -- how data flows -- while cognitive science surveys focus on cognitive function -- what the agent does. Neither axis alone disambiguates architecturally distinct systems: the same Orchestrator-Workers topology can implement Plan-and-Execute, Hierarchical Delegation, or Adversarial Verification -- three patterns with fundamentally different failure modes and design trade-offs.\n  We propose a two-dimensional classification t"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"The resulting 7x6 matrix identifies 27 named patterns, 13 with original names. We demonstrate orthogonality through systematic cross-axis analysis, define eight representative patterns in detail, and validate descriptive coverage across four real-world domains. Cross-domain analysis yields five empirical laws of pattern selection governing the relationship between environmental constraints and architectural choices.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the seven cognitive-function categories and six execution-topology archetypes are exhaustive and mutually orthogonal enough to disambiguate all architecturally distinct systems, and that qualitative analysis across four domains is sufficient to derive general empirical laws.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"A 7x6 matrix classifies AI agent patterns into 27 types by combining cognitive functions and execution topologies, yielding five empirical laws linking task constraints to architectural choices.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A 7-by-6 matrix of cognitive functions and execution topologies classifies 27 distinct AI agent design patterns.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"45deacd4c50ad80e207cf23cae7ad499fa6f591051e4cd7755435993431ec415"},"source":{"id":"2605.13850","kind":"arxiv","version":1},"verdict":{"id":"a4db2eb5-f336-40cd-b2d4-997dd762e6e6","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T11:00:11.023341Z","strongest_claim":"The resulting 7x6 matrix identifies 27 named patterns, 13 with original names. We demonstrate orthogonality through systematic cross-axis analysis, define eight representative patterns in detail, and validate descriptive coverage across four real-world domains. Cross-domain analysis yields five empirical laws of pattern selection governing the relationship between environmental constraints and architectural choices.","one_line_summary":"A 7x6 matrix classifies AI agent patterns into 27 types by combining cognitive functions and execution topologies, yielding five empirical laws linking task constraints to architectural choices.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the seven cognitive-function categories and six execution-topology archetypes are exhaustive and mutually orthogonal enough to disambiguate all architecturally distinct systems, and that qualitative analysis across four domains is sufficient to derive general empirical laws.","pith_extraction_headline":"A 7-by-6 matrix of cognitive functions and execution topologies classifies 27 distinct AI agent design patterns."},"references":{"count":25,"sample":[{"doi":"","year":2024,"title":"E. Schluntz and B. Zhang, “Building effective agents,” Anthropic Research Blog, Dec. 2024","work_id":"a2bcfeaa-4a5a-4c39-a6ac-57f3a13e6a83","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Agent Development Kit: A flexible framework for building multi-agent systems,","work_id":"cf430de7-241f-47a6-a026-eb9a238c791f","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"H. Chase et al., “LangGraph: Multi-agent workflows,” LangChain Documentation, Feb. 2025","work_id":"a069cec2-2f8f-4952-ba17-3c6111992eec","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"What’s next for AI agentic workflows,","work_id":"98f1b9a1-2212-457f-b19c-1f22ba53ab33","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"A survey on large language model based autonomous agents,","work_id":"9889259e-cb74-406c-b7dc-2c7b17e0736c","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":25,"snapshot_sha256":"8c36ba481dd4706927d6ba149425348d0e0041d74e468ec0b3cae99c21a30071","internal_anchors":4},"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"}