AI Insight
This paper introduces the Hyperdimensional Probe, a new method for interpreting how Large Language Models internally represent information. The approach combines Vector Symbolic Architectures with neural probing to analyze both input features and output distributions simultaneously, addressing limitations of existing interpretability methods like supervised probes, Sparse Autoencoders, and logit-based approaches. Experiments show the method successfully extracts meaningful semantic information across different language models and tasks, providing deeper insights into how LLMs process and generate text.
Why it matters
Understanding LLM internal representations is crucial for improving AI safety, debugging model failures, and building more reliable AI systems. This unified interpretability framework could help researchers and developers better understand why models make certain predictions and identify potential biases or reasoning errors before deployment.
Understand the Science
arXiv:2509.25045v3 Announce Type: replace-cross
Abstract: Despite their capabilities, Large Language Models (LLMs) remain opaque with limited understanding of their internal representations. Current interpretability methods either focus on input-oriented feature extraction, such as supervised probes and Sparse Autoencoders (SAEs), or on output distribution inspection, such as logit-oriented approaches. A full understanding of LLM vector spaces, however, requires integrating both perspectives, something existing approaches struggle with due to constraints on latent feature definitions. We introduce the Hyperdimensional Probe, a hybrid supervised probe that combines symbolic representations with neural probing. Leveraging Vector Symbolic Architectures (VSAs) and hypervector algebra, it unifies prior methods: the top-down interpretability of supervised probes, SAE’s sparsity-driven proxy space, and output-oriented logit investigation. By combining the supervised learning paradigm of traditional probes with the dictionary-based representation principle of SAEs, our approach enables deeper input-focused feature extraction while supporting output-oriented analysis. Our experiments demonstrate that our approach consistently extracts meaningful semantic information across different LLMs, embedding sizes, and configurations, uncovering concept-oriented insights into LLM inference across two distinct scenarios: input-completion tasks and QA-focused text generation. VSA-based probing overcomes the limitations of logit-based analyses, which are constrained by the model’s token vocabulary, while also mitigating the noisier interpretability outcomes often produced by SAEs in settings with a bounded conceptual feature space. By supporting a joint investigation of input-output features, this work advances the semantic understanding of neural representations while unifying the complementary perspectives of prior methods.
Source: Hyperdimensional Probe: Decoding LLM Representations via Vector Symbolic Architectures