The study investigates the consistency and specificity of language model circuits across six tasks and five models, focusing on component-level (attention heads and MLP blocks) and neuron-level circuits. Component-level circuits are highly consistent and causally important but lack task specificity, as ablating a circuit for one task similarly harms performance on other tasks. Neuron-level circuits show higher task specificity but lower consistency, with overlap mainly between closely related tasks. The analysis of Llama‑3.2‑3B reveals that shared components are predominantly MLP blocks, while attention heads act as generic attention‑sink heads.
arXiv:2606. 05378v1 Announce Type: new Abstract: We test whether a single screen-and-ablate recipe -- identify attention-head circuits by task-pattern selectivity, then verify by causal ablation against a matched-random null -- produces consistent mechanistic claims across model families.
arXiv:2605.09129v2 Announce Type: replace Abstract: Circuit study aims to explain how language models (LMs) implement a task by localizing and interpreting a circuit: a computational subgraph respons...
arXiv:2606. 24026v1 Announce Type: new Abstract: Mechanistic interpretability has made substantial progress in automatically localizing circuits, but explaining what localized components do remains labor-intensive and difficult to standardize.
arXiv:2609.25602v1 Announce Type: new Abstract: In language models, the choice between believing the prompt and believing the weights is made by a handful of identifiable attention heads. Instruction...
The study investigates how task context reshapes the roles of shared attention heads in large language models. By applying identical graph facts and measuring interventions on attention heads and downstream activations, the authors show that task changes redistribute influence across five open‑weight models, causing the same heads to respond to different facts and affect later components differently. Linear response maps before and after activation edits reveal significant reductions in prediction error, indicating that task context alters both the facts processed and the downstream impact of shared components.
The paper investigates whether linguistic and arithmetic tasks in large language models (LLMs) rely on shared neural circuits. Using mechanistic interpretability techniques on five LLMs, the authors find that circuit similarity aligns primarily with task domain rather than input/output vocabularies or syntactic structure. This suggests that LLMs may employ largely separate neural circuits for language and arithmetic, indicating a modular organization of abilities.
arXiv:2609.32964v2 Announce Type: replace Abstract: Language models (LMs) often hallucinate by committing to confident answers rather than abstaining, even when they do not have enough information to...
Fine‑tuning reshapes internal representations of large language models, affecting attention patterns and layer‑wise activations. The study shows that components identified by EAP as important for task performance cluster in specific layers, yet these layers do not align with those undergoing the largest representational changes. Additionally, overlapping EAP components across different tasks do not guarantee cross‑task transfer and can even degrade performance when tasks differ in nature.
arXiv:2606. 16920v1 Announce Type: cross Abstract: Circuit discovery is a key technique in mechanistic interpretability to pinpoint the model components that are crucial for performing a given task.
The paper introduces Circuit Condensation, a post‑training method that prunes low‑attribution edges from large causal graphs and trains a low‑rank adapter to preserve behavior. Across four behaviors and eight models, the condensed circuits are on average 8.1× smaller than the strongest frozen baseline, with reductions up to 316×. Experiments show that weight updates drive the size reduction, and detailed ablations reveal dependencies among remaining edges and a more focused set of heads for indirect object identification.
The paper proposes that two architectural assumptions—(1) attention and MLPs share a key‑value form <phi(S)>U, and (2) components read from an additive residual stream—are sufficient to answer three interpretability questions: component interaction, information routing, and token attribution. By treating these selections as a computational graph, the authors develop Unpack, a backward attribution method that validates interaction scores, recovered routes, and token attribution against established tests across models ranging from 160M to 6.9B parameters. The study also shows that contribution and causal effect can differ, with a recognizable signature in how components change when a task is removed.
The paper investigates whether structural differences in circuits discovered by circuit discovery methods reflect distinct mechanisms. By varying input-token frequency while keeping the task constant, the authors find that although circuits appear specialized by frequency structurally, functional and representational analyses reveal no reliable differences, a phenomenon they call phantom specialization. Across multiple models and tasks, structurally distinct circuits implement the same computation, with core shared subgraphs recovering most of the performance and interchangeable internal representations confirmed by causal interventions.