🗺️ Connected Papers — Citation Map for PUMA
Tool: Connected Papers (https://www.connectedpapers.com) Phase: Phase 1 — Research Step: 02 — Scientific Mapping Methodology: Citation Network Analysis
Prompt A — Benchmark Foundation Map
Map the citation neighborhood around the seed paper on LLM agents and project management automation. Focus on: prior works that established the foundational methods, derivative works that applied them to PM, and conceptual neighbors using different terminology for the same problems. Identify the most connected hub papers.
Seed: Assalaarachchi (2026) arXiv:2601.16392 OR Cinkusz et al. (2025) arXiv:2508.16678
Prompt B — Methodological Contrast Analysis
Show papers that pursue similar goals (issue triage, effort estimation, PM automation) with different methods: classical ML approaches (SVM, Random Forest) versus LLM-based agents versus multi-agent orchestration. Highlight papers that provide direct empirical comparisons between these approaches.
Prompt C — Evaluation Lineage Tracing
Trace the evaluation lineage around this seed paper. Highlight works that introduced benchmarks, datasets, or reproducible experimental protocols usable in PUMA. Focus on: what datasets were used, what metrics reported, and whether artifacts were made publicly available.
Prompt D — Governance Branch Discovery
Identify papers in this citation map most relevant to: governance of agentic systems, human-in-the-loop mechanisms, traceability requirements, hallucination detection in PM metrics, and safe deployment in enterprise project environments.
Three-Step Usage Protocol
- Input: Paste the DOI or exact title of a seed paper
- Explore: Identify hub papers (highest edge count), prior/derivative split
- Export: Download the cluster visualization → save as PNG for PUMA thesis figures
- Import: Export BibTeX → Zotero collection
PUMA-Literature-2026
PUMA Relevance
Connected Papers revealed the cluster structure of the PM+AI field for PUMA: a foundational cluster (ReAct, MetaGPT, AutoGen), an applied PM cluster (Assalaarachchi, Cinkusz), and an evaluation methodology cluster (SWE-bench, MASAI). This three-cluster structure organizes PUMA’s Section 2 state-of-the-art.