# Do LLM Embedding Spaces Recover Expert Structure?

arXiv: https://arxiv.org/abs/2606.23394 · HTML: https://mlsanity.com/paper/2606.23394/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 11 of its 29 references that are in the map.
- First posted on arXiv 22 Jun 2026.

## Builds on (most cited first, 11 of 11)

- [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://mlsanity.com/paper/1908.10084/) (2019)
- [Axiomatic Attribution for Deep Networks](https://mlsanity.com/paper/1703.01365/) (2017)
- [ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT](https://mlsanity.com/paper/2004.12832/) (2020)
- [Probing Classifiers: Promises, Shortcomings, and Advances](https://mlsanity.com/paper/2102.12452/) (2021)
- [What you can cram into a single vector: Probing sentence embeddings for linguistic properties](https://mlsanity.com/paper/1805.01070/) (2018)
- [Explaining Black Box Predictions and Unveiling Data Artifacts through Influence Functions](https://mlsanity.com/paper/2005.06676/) (2020)
- [Large language models predict human sensory judgments across six modalities](https://mlsanity.com/paper/2302.01308/) (2023)
- [SBERT studies Meaning Representations: Decomposing Sentence Embeddings into Explainable Semantic Features](https://mlsanity.com/paper/2206.07023/) (2022)
- [Investigating semantic subspaces of Transformer sentence embeddings through linear structural probing](https://mlsanity.com/paper/2310.11923/) (2023)
- [Interpretable Text Embeddings and Text Similarity Explanation: A Survey](https://mlsanity.com/paper/2502.14862/) (2025)
- [How Well Do Deep Learning Models Capture Human Concepts? The Case of the Typicality Effect](https://mlsanity.com/paper/2405.16128/) (2024)

## Abstract (arXiv)

Pretrained text embeddings are increasingly used as representational maps, yet high category separability does not imply that their geometry recovers expert-defined structure. We study this problem in mental-health-related language, where symptom relations provide an external reference and online communities introduce strong domain, affective, stylistic, and discourse confounds. Using 28 Reddit communities, we compare pretrained and supervised fine-tuned Qwen3 embedding spaces at two scales (0.6B and 4B). We construct category prototypes, evaluate their representational dissimilarity matrices against an expert symptom matrix with representational similarity analysis, and complement this global test with prototype-based typicality and multi-baseline confound controls. Pretrained embeddings show measurable alignment with expert structure within the mental-health subset; fine-tuning strengthens this alignment most at the finest category level; and larger scale improves both zero-shot alignment and supervision-induced gains. Residual alignment remains substantial after controlling for VAD, LIWC, lexical style, and topic-distribution structure. These results suggest that LLM embeddings can recover expert-relevant category geometry, but this recovery is level-dependent and should be tested against explicit confounds rather than inferred from classification alone.