# Forewarned is Forearmed: When Non-Sequential Embedding Turns Into an Anomaly Detector

arXiv: https://arxiv.org/abs/2606.30196 · HTML: https://mlsanity.com/paper/2606.30196/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 14 of its 20 references that are in the map.
- First posted on arXiv 29 Jun 2026.

## Builds on (most cited first, 14 of 14)

- [BERTScore: Evaluating Text Generation with BERT](https://mlsanity.com/paper/1904.09675/) (2019)
- [No Language Left Behind: Scaling Human-Centered Machine Translation](https://mlsanity.com/paper/2207.04672/) (2022)
- [VoxCeleb: a large-scale speaker identification dataset](https://mlsanity.com/paper/1706.08612/) (2017)
- [Language-agnostic BERT Sentence Embedding](https://mlsanity.com/paper/2007.01852/) (2020)
- [Listen, Attend and Spell](https://mlsanity.com/paper/1508.01211/) (2015)
- [SONAR: Sentence-Level Multimodal and Language-Agnostic Representations](https://mlsanity.com/paper/2308.11466/) (2023)
- [Large Concept Models: Language Modeling in a Sentence Representation Space](https://mlsanity.com/paper/2412.08821/) (2024)
- [LLMs Will Always Hallucinate, and We Need to Live With This](https://mlsanity.com/paper/2409.05746/) (2024)
- [SAMU-XLSR: Semantically-Aligned Multimodal Utterance-level Cross-Lingual Speech Representation](https://mlsanity.com/paper/2205.08180/) (2022)
- [BESTOW: Efficient and Streamable Speech Language Model with the Best of Two Worlds in GPT and T5](https://mlsanity.com/paper/2406.19954/) (2024)
- [Semantic Energy: Detecting LLM Hallucination Beyond Entropy](https://mlsanity.com/paper/2508.14496/) (2025)
- [Audio-Linguistic Embeddings for Spoken Sentences](https://mlsanity.com/paper/1902.07817/) (2019)
- [Efficient Streaming LLM for Speech Recognition](https://mlsanity.com/paper/2410.03752/) (2024)
- [SONAR-LLM: Autoregressive Transformer that Thinks in Sentence Embeddings and Speaks in Tokens](https://mlsanity.com/paper/2508.05305/) (2025)

## Abstract (arXiv)

This paper offers an in-depth analysis of non-sequential multimodal sentence-level embeddings, with a particular focus on the SONAR model. We demonstrate that certain embedding dimensions are sensitive to perturbations and can serve as indicators of decoding anomalies. By leveraging the consistency between successive encoding and decoding, we successfully build an accurate detector. Additionally, we explore modifying specific dimensions of interest to attempt to correct them. This work underscores the importance of understanding and analyzing the embeddings themselves to enhance the reliability of multimodal representations.