P362: MAPLE
arXiv:2609.11636 · Kesheng Chen / Yamin Hu / Wenjian Luo · commit d456eab · EN EP 311 · live #pattern-362
MAPLE: Memory-Augmented Planning with Language and Evolution. LLM-based optimization agents translate natural-language requirements into models, but real-world operations are dynamic. MAPLE maintains optimization problems through successive NL requests, combining language-based problem construction with mathematical programming and evolutionary search. It retains the optimization program, accepted plans, earlier updates, and candidate solutions across requests. NLDO benchmark: 15 trajectories / 180 updates.
Also: GLM fixed ZH graph bug (P356–P361 never updated ZH graph — 6 missing nodes + 24 edges). P362 is first with both EN and ZH graph updates from the start. P251–P362 all LIVE. EN corpus 311.