The paper introduces a mechanistic interpretability framework inspired by aphasia to probe large language models (LLMs). Using classifier‑guided lesioning and continuous activation steering, the authors induce aphasia‑like breakdowns and find that language functions can be isolated to discrete structural circuits and continuous vectors, yet these functions are densely entangled with significant overlap across different aphasic profiles. While steering generalizes across downstream tasks, structural lesioning shows asymmetric constraints, highlighting the densely entangled architecture of LLMs that can be mapped through clinically‑informed interventions.
arXiv:2608. 12717v1 Announce Type: new Abstract: Mechanistic interpretability of large language models lacks spatially resolved, falsifiable tools for testing whether internal components are specialized for distinct cognitive operations.
The study investigates how progressive neuronal loss affects language models by systematically ablating computational units in LSTM and Transformer architectures. LSTMs show a gradual decline in syntactic and semantic abilities, while Transformers exhibit a threshold-like collapse under MLP ablation and a smoother sigmoid-like decline under attention-head ablation. These distinct failure modes mirror different aphasic syndromes, suggesting that architecture-dependent degradation patterns can provide computational insights into neurodegenerative diseases.
arXiv:2601. 17952v2 Announce Type: replace-cross Abstract: Interpretability remains a key challenge for deploying language models (LM) in clinical settings such as progression diagnosis of Alzheimer disease, where early and trustworthy predictions are essential.
arXiv:2608. 06429v1 Announce Type: cross Abstract: Interpretability methods for large language models (LLMs) describe internal state but do not directly test whether that state is causally sufficient to produce the observed behavior.
arXiv:2607. 00397v1 Announce Type: cross Abstract: Understanding how complex cognitive functions are organized within artificial systems is central to interpreting large language models (LLMs) and relating them to biological cognition.
The paper argues that neural networks can exhibit modular organization even when individual neurons are not specialized. It introduces a nested hierarchy of modularity—unit-level, orthogonal, and linear—and shows that many vision, language, and RNN models lack unit-level modularity but still display clear functional modularity at the subspace level. The authors demonstrate that these subspaces can be causally validated through ablation and align with brain data, revealing distinct content/style or syntax/semantics separations.
arXiv:2607. 11621v1 Announce Type: new Abstract: Aphasia following stroke commonly produces systematic naming errors with characteristic profiles, but whether general-purpose language models not designed for clinical simulation can reproduce these patterns remains untested.
arXiv:2606. 05843v1 Announce Type: cross Abstract: While Multimodal Large Language Models (MLLMs) demonstrate remarkable proficiency on complex vision-language tasks, the mechanisms by which they extract query-relevant visual features from complex, noisy contexts remain opaque.
arXiv:2606. 09287v1 Announce Type: new Abstract: Understanding how transformer representations evolve across layers, not merely what they encode, remains an open problem in mechanistic interpretability.
The study investigates how a large language model (LLaMA2‑7B) aligns with human brain activity during naturalistic language comprehension by using SCRIBE, an interpretable semantic representation of 65 neurocognitively grounded properties. Variance partitioning on fMRI data from seven participants reveals that shared variance between LLaMA2‑7B and SCRIBE dominates across the brain, yet both representations also contribute unique variance. Further analysis shows that concrete semantic properties correlate more strongly with shared variance, while abstract properties do not show increased unique variance, and ablation experiments demonstrate that multiple semantic dimensions cumulatively influence both shared and unique brain-predictive variance.
arXiv:2605. 28854v2 Announce Type: replace-cross Abstract: Large language models (LLMs) exhibit remarkable flexibility in adapting to novel tasks from in-context examples without parameter updates, a capability known as in-context learning (ICL).
arXiv:2606. 28719v1 Announce Type: new Abstract: Test-time adaptation (TTA) of vision-language models (VLMs) is essential for their robust deployment in dynamic, real-world environments.