Skip to content
EUCLIDE

Keeping project context across conversations

IFCMemoryBench: Evaluating Long-Term Memory of LLM-Based Agents in BIM Information Retrieval

Workshop paper ·

Du, Vosseler, Mazza & Borrmann · KDD Agentic AI Evaluation Workshop

Document intelligenceBIM information managementEvidence & evaluation

IFCMemoryBench: Evaluating Long-Term Memory of LLM-Based Agents in BIM Information Retrieval
Original paper · first page

What it proposes

IFCMemoryBench tests whether agents can combine earlier project decisions with live queries of BIM models.

Limits of the evidence

General-purpose memory still struggles in this specialist benchmark.

Not tested by EUCLIDE

Original paper

Source checked: · duhan