InsightBuild uses Granger causality and structural pruning to rank sensor causes, then a fine-tuned LLaMA 2 model turns the ranked causes into text explanations that outperform baseline systems on annotated building energy anomalies.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
InsightBuild: LLM-Powered Causal Reasoning in Smart Building Systems
InsightBuild uses Granger causality and structural pruning to rank sensor causes, then a fine-tuned LLaMA 2 model turns the ranked causes into text explanations that outperform baseline systems on annotated building energy anomalies.