Matheus Marques

Matheus Marques, Carlos Alberto Malcher, Cledson de Sousa

Modern AIOps environments operating within multi-campus institutional infrastructures suffer acutely from topological drift and black-box unmanaged physical network segments. Classical layer-2 discovery pipelines rely on uniform administrative cooperation and ubiquitous SNMP polling, which routinely fail in heterogeneous, multi-vendor, and multi-tenant overlay infrastructures. This paper introduces B.O.D.Y. (Beyond-Overlay Deterministic topologY), a deterministic structural grounding layer for AIOps ecosystems operating under fragmented administrative boundaries. B.O.D.Y. bypasses the NP-hard incomplete Address Forwarding Table (AFT) resolution dilemma by formalizing a multi-modal data fusion pipeline that orchestrates ephemeral MAC address forwarding tables collected via non-privileged terminal sessions, passive OUI fingerprinting, PoE telemetry, and declarative state storage. The resulting topology graph reconstructs unmanaged physical segments and maps logical asset semantics without administrative network privileges, providing downstream reasoning systems with an immutable, auditable source of physical ground truth. Evaluation across five campuses of the Universidade Federal Fluminense, resolving 530 of 541 registered edge devices, demonstrates that deterministic topological grounding eliminates a critical failure mode in prob bilistic AIOps reasoning: confident causal inference decoupled from physical reality.

Chris Su, Harrison Li, Matheus Marques et al.

Recent work reports that Large Reasoning Models (LRMs) undergo a collapse in performance on solving puzzles beyond certain perplexity thresholds. In subsequent discourse, questions have arisen as to whether the nature of the task muddles an evaluation of true reasoning. One potential confound is the requirement that the model keep track of the state space on its own. We provide a large language model (LLM) with an environment interface for Tower of Hanoi problems, allowing it to make a move with a tool call, provide written justification, observe the resulting state space, and reprompt itself for the next move. We observe that access to an environment interface does not delay or eradicate performance collapse. Furthermore, LLM-parameterized policy analysis reveals increasing divergence from both optimal policies and uniformly random policies, suggesting that the model exhibits mode-like collapse at each level of complexity, and that performance is dependent upon whether the mode reflects the correct solution for the problem. We suggest that a similar phenomena might take place in LRMs.