Article

Intrinsic Thematic Separability in Sentence-Transformer Embeddings: A Controlled Geometric Study from Synthetic to Real-World Text

2026-03-06

Abstract excerpt

<title>Abstract</title> <p>We investigate whether sentence-transformer embeddings encode sufficient topical information to recover imposed thematic structure in corpora of varying origin and difficulty. A synthetic corpus of 8,000 ChatGPT-generated texts across four domains achieves perfect macro-group recovery in 768D ($K$-Means ARI\,=\,1.000), demonstrating that separability is intrinsic to the embeddings and i...

Topics

Open a Topic to create a Post that cites this publication.

Identifiers and source

Literature Corpus work: d700119a-6e90-5c4e-9882-57fbd8ce5bf1
DOI: 10.21203/rs.3.rs-8976073/v1

Open publication

Related research

Semantic proximity does not establish scientific evidence.

Click a neighbor to travelStep 1 · 12 closest

Intrinsic Thematic Separability in Sentence-Transformer Embeddings: A Controlled Geometric Study from Synthetic to Real-World TextDOI 10.21203/rs.3.rs-8976073/v1Read

Exploring Textual Semantics in Large Language ModelsSemantic proximity Read

Uncertainty-Aware Assessment of LLM-Enhanced Topic Models: An Experton-Based Approach to InterpretabilitySemantic proximity Read

Automatic Discovery of Intra-Class Sub-Structure for Supervised Tabular Classification: Offline Clustering vs. Joint Sub-Center TrainingSemantic proximity Read

Unsupervised text clustering with large language modelsSemantic proximity Read

Analysing the potential solutions to LLM hallucinations in abstractive text summarisationSemantic proximity Read

An Evaluative Baseline for Sentence-Level Semantic DivisionSemantic proximity Read

Towards Algorithmic Framing Analysis: Expanding the Scope by Using LLMsSemantic proximity Read

Dynamic Token Clustering: A Novel Architectural Enhancement for Large Language ModelsSemantic proximity Read

Domain-Partitioned Retrieval as a Hallucination Mitigation Strategy in Conversational RAG: The PC-RAG ArchitectureSemantic proximity Read

AI vs. Human: Decoding Text Authenticity with TransformersSemantic proximity Read

Cracking the code of co-authorship networks geo-temporally using interpretable machine learningSemantic proximity Read

Enhancing Text Summarization in a Limited Data and Long Sample SettingSemantic proximity Read