🤖 AI Summary
This study investigates the representational capacity of quantum-inspired high-dimensional document embeddings in information retrieval and their potential to complement conventional dense embeddings. The authors propose a 1024-dimensional embedding framework incorporating EigAngle semantic projection, circuit-inspired feature mapping, and optional teacher–student distillation. To evaluate its efficacy, they develop a diagnostic toolkit comprising dynamic and static BM25–embedding interpolation, candidate fusion strategies, and alpha-oracle upper-bound analysis. Experimental results indicate that, when used in isolation, the proposed embedding yields weak and unstable ranking signals and suffers from distance compression issues. However, when integrated with BM25 in a hybrid retrieval setting, it significantly enhances retrieval performance, thereby demonstrating its value as a complementary signal rather than a standalone representation.
📝 Abstract
Text embeddings are central to modern information retrieval and Retrieval-Augmented Generation (RAG). While dense models derived from Large Language Models (LLMs) dominate current practice, recent work has explored quantum-inspired alternatives motivated by the geometric properties of Hilbert-like spaces and their potential to encode richer semantic structure. This paper presents an experimental framework for constructing quantum-inspired 1024-dimensional document embeddings based on overlapping windows and multi-scale aggregation. The pipeline combines semantic projections (e.g., EigAngle), circuit-inspired feature mappings, and optional teacher-student distillation, together with a fingerprinting mechanism for reproducibility and controlled evaluation. We introduce a set of diagnostic tools for hybrid retrieval, including static and dynamic interpolation between BM25 and embedding-based scores, candidate union strategies, and a conceptual alpha-oracle that provides an upper bound for score-level fusion. Experiments on controlled corpora of Italian and English documents across technical, narrative, and legal domains, using synthetic queries, show that BM25 remains a strong baseline, teacher embeddings provide stable semantic structure, and standalone quantum-inspired embeddings exhibit weak and unstable ranking signals. Distillation yields mixed effects, improving alignment in some cases but not consistently enhancing retrieval performance, while hybrid retrieval can recover competitive results when lexical and embedding-based signals are combined. Overall, the results highlight structural limitations in the geometry of quantum-inspired embeddings, including distance compression and ranking instability, and clarify their role as auxiliary components rather than standalone retrieval representations.