aiqbench analysis
This is an academic NLP study showing that path-signature features built from Word2Vec trajectories track character-level edit distances far better than average pooling, which mostly misses them. Its claims are limited to lexical-similarity correlations on five string metrics, not improvements on semantic tasks, and the authors position signatures as interpretable tools rather than replacements for Transformer embeddings. Readers should treat the results as evidence about what embedding geometry encodes, not as a practical gain in retrieval or generation.
Iterated Integrals as Lexical Fingerprints: Path Signature Embeddings for Lexical Similarity
arXiv:2610.41855v1 Announce Type: new Abstract: While sentence embeddings are primarily designed for semantic generalization, their latent geometry implicitly encodes surface-level lexical properties, a relationship that remains poorly understood. Shallow pooling methods discard word order, while deep Transformer models obscure these lexical alignments behind computational complexity and uninterpretable spaces. We propose constructing sentence embeddings via the path signature of Word2Vec trajectories, leveraging the iterated integrals of rough path theory to create representations that explicitly capture sequential and lexical structure in a geometrically tractable manner. Through extensive ablations spanning Word2Vec dimensions, signature depths, and PCA compression, we correlate embedding distances against five lexical string metrics: Levenshtein, Jaro-Winkler, word Jaccard, phoneme Levenshtein, and character TF-IDF. We reveal that while average pooling almost entirely fails to capture character-level edit distances, path signatures strongly align with them, with depth 2 signatures optimally balancing character and word-level sensitivities. Furthermore, we demonstrate that the critical lexical variance is concentrated in a highly compact PCA subspace, and unlike pooling, signatures structurally preserve and quantify word order. These findings establish path signatures not as semantic generalizers, but as rigorous mathematical lenses that explicitly encode discrete lexical structures and word order, offering an interpretable alternative to opaque deep learning for lexical tracking.
Abstract:While sentence embeddings are primarily designed for semantic generalization, their latent geometry implicitly encodes surface-level lexical properties, a relationship that remains poorly understood. Shallow pooling methods discard word order, while deep Transformer models obscure these lexical alignments behind computational complexity and uninterpretable spaces. We propose constructing sentence embeddings via the path signature of Word2Vec trajectories, leveraging the iterated integrals of rough path theory to create representations that explicitly capture sequential and lexical structure in a geometrically tractable manner. Through extensive ablations spanning Word2Vec dimensions, signature depths, and PCA compression, we correlate embedding distances against five lexical string metrics: Levenshtein, Jaro-Winkler, word Jaccard, phoneme Levenshtein, and character TF-IDF. We reveal that while average pooling almost entirely fails to capture character-level edit distances, path signatures strongly align with them, with depth 2 signatures optimally balancing character and word-level sensitivities. Furthermore, we demonstrate that the critical lexical variance is concentrated in a highly compact PCA subspace, and unlike pooling, signatures structurally preserve and quantify word order. These findings establish path signatures not as semantic generalizers, but as rigorous mathematical lenses that explicitly encode discrete lexical structures and word order, offering an interpretable alternative to opaque deep learning for lexical tracking.
Computation and Language (cs.CL)
arXiv:2610.41855 [cs.CL]
(or arXiv:2610.41855v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2610.41855 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 292 words
Related coverage
Efficient, Robust, and Anti-Collusion Fingerprinting of Image Diffusion Models
arXiv · Jun 12, 2026
Multi-View Molecular Representation Learning with Hierarchical Graphs and Contextualized Fingerprints
arXiv · Sep 16, 2026
Self-Supervised Pretraining of Molecular Graph Encoders with LeJEPA
arXiv · Sep 7, 2026
From Construction to Injection: Edit-Based Fingerprints for Large Language Models
arXiv · Oct 1, 2026