Research relevant to Merlion (2025–2026)
Snapshot as of 2026-09-22. Each row states the paper’s finding and where the specs apply it.
| \ | Paper | Venue | Finding | Applied in |
|---|---|---|---|---|
| 1 | L. Nachmanson, X. Chen. Browsing Large Graphs with Tile Pyramids and Sleeve Routing in the Browser. arXiv:2605.17498 | arXiv, 2026-05 | Semantic zoom that keeps the labels of the highest-ranked nodes readable at every zoom level. Edges routed through free space with a constrained Delaunay triangulation and the funnel algorithm. 32,768 nodes and 236,978 edges handled client-side in MSAGL.js. | layout.md spline routing; viewer.md semantic zoom |
| 2 | Z. Alsuwaykit et al. ARCOL: Aspect Ratio Constrained Orthogonal Layout. arXiv:2603.29618 | arXiv, 2026-03 (rev. 2026-09) | Aspect ratio as a soft constraint in stress minimisation and tree re-attachment (built on HOLA); balanced layouts at any target ratio; validated by a user study and expert interviews. | layout.md container fit |
| 3 | F. V. Fomin, P. A. Golovach, T. Inamdar, S. Saurabh, M. Zehavi. Tight Parameterized (In)tractability of Layered Crossing Minimization. arXiv:2510.13335 | SODA 2026 | Exact crossing minimisation: 2^O(√k log k) + n·k^O(1) for 2 layers, 2^O(k^(2/3) log k) + n·k^O(1) for 3 layers (k = number of crossings); no 2^o(k/log k)·n^O(1) algorithm for 5+ layers unless the Exponential Time Hypothesis fails. | layout.md exact refinement: why refinement stays one pair of layers at a time |
| 4 | P. Giannopoulos et al. One-Sided Local Crossing Minimization. arXiv:2510.00331 | arXiv, 2025-09 (rev. 2026-04) | Minimising the maximum crossings per edge is NP-hard even for forests of stars; the median heuristic with specific tie-breaking is a 3-approximation. | layout.md local objective |
| 5 | S. van Wageningen, T. Mchedlidze, A. C. Telea. Same Quality Metrics, Different Graph Drawings. arXiv:2508.15557 | GD 2025 | Drawings can be morphed into arbitrary shapes with quality metrics nearly unchanged; metrics can rate poor drawings as good. | benchmark.md; ADR-0007 |
| 6 | G. J. Mooney, J. Miller, M. Wybrow, S. Kobourov, H. C. Purchase. Stress in Graph Drawings: Perception, Preference, and Performance. GD 2025, LIPIcs 38 | GD 2025 | People perceive stress, prefer lower stress, and find shortest paths faster as stress falls. | benchmark.md stress metric |
| 7 | B. C. B. Charytitsch, M. C. V. Nascimento. An Efficient Hybridization of Graph Representation Learning and Metaheuristics for the Constrained Incremental Graph Drawing Problem. arXiv:2508.15949 | EJOR 330(2), 2026 | Formalises incremental layered drawing with bounded movement of existing nodes (C-IGDP); an embedding-seeded GRASP beats earlier GRASP heuristics on the primal integral. | layout.md stable layout; ADR-0006 |
| 8 | P. Schaad, T. Ben-Nun, T. Hoefler. VEIL: Reading Control Flow Graphs Like Code. arXiv:2511.05066 | arXiv, 2025-11 | Dominator-based layering preserves execution order; adds happens-before ordering and edge-direction grouping as criteria; beats general-purpose layouts on real control-flow graphs. | layout.md flowchart layering |
| 9 | C. Liang, J. You. DiagramEval: Evaluating LLM-Generated Diagrams via Graphs. arXiv:2510.25761 | EMNLP 2025 | Scores SVG diagrams by extracting a graph (text → nodes, connections → edges) and aligning nodes and paths with the intended graph. | benchmark.md round-trip correctness |
| 10 | B. Shbita, F. Ahmed, C. DeLuca. MermaidSeqBench. arXiv:2511.14967 | NeurIPS 2025 workshop | 132 test cases for turning natural language into Mermaid sequence diagrams; LLMs differ widely on syntax correctness. Dataset Apache-2.0. | benchmark.md llm corpus; parser.md |
Background: X. Qian et al., GraphTide (arXiv:2604.12624, 2026) found that revealing a nested graph step by step improves comprehension, which is relevant to a future viewer mode. The ELK design is described in arXiv:2311.00533 (2023).