activation-patching

The paper investigates neurons in frozen BERT that drive AI-text detection using sparse probing and activation patching on the RAID benchmark, identifying a small set of causally relevant neurons that generalize across generator families.

This paper uses activation patching to causally measure if reasoning steps in chain-of-thought are load-bearing, finding that behavioral tests overestimate faithfulness and larger models like Qwen3-4B maintain better faithfulness across reasoning depths.

This paper investigates how vision-language models read exact values from vertical bar charts using controlled counterfactual activation patching, revealing insights into internal computations and differences between models like Qwen and InternVL.

This paper uses mechanistic interpretability to analyze how LLaMA 3.1 8B models numerical sequences, revealing that it internally computes and stores first differences to extrapolate patterns.

This research investigates how multilingual large language models internally handle subject-verb agreement across languages, finding that models reuse shared computational structure for languages with overt inflection, indicating cross-lingual overlap in morphosyntactic processing.

This paper argues that LLM failures on hidden-constraint reasoning are routing problems, not knowledge problems, and introduces a quartet diagnostic to dissociate knowledge, symmetry, routing, and repair across 14 models, with activation probing and patching experiments.

This paper investigates how language models execute in-context conditional rules by probing whether testing and routing are separable mechanisms. Using activation patching across three open models and six languages, the authors find that predicate testing is modular while route representations are token-bound and non-transferable.

This paper presents a graph-to-answer mechanism-tracing case study for Graph-PRefLexOR-8B, a materials-science hypothesis generation model, using visualization and activation-based diagnostics to localize where mechanism support is lost or recovered during generation.

This paper tests whether increasing interventional data in pretraining improves LLMs' causal direction reasoning, using controlled Simpson's-paradox worlds. It finds that the training mixture does not govern interventional evidence use; instead the evidence type in the inference-time context is the decisive factor.

This paper presents a theoretical cost accounting for exhaustive sweeps and sequential mutations on reactive computational graphs, deriving speedup ratios and validating them on a Julia-based graph engine.

This paper investigates whether arithmetic heuristic neurons in LLMs are form-invariant across symbolic arithmetic, natural language word problems, and Python code. Using activation patching, they find a shared circuit of neurons that is necessary and sufficient for arithmetic computation, and that cross-format failures arise from activation states rather than distinct circuits.

This paper investigates how vision-language models resolve conflicts between visual evidence and world knowledge, revealing that visual grounding is the default while prior knowledge depends on a small set of late-layer attention heads. The authors perform causal analysis across three VLM families, demonstrating an asymmetric structure where ablating these heads shifts predictions from knowledge-grounded to visually grounded answers.

This paper re-derives activation patching from causal mediation analysis, revealing that the natural indirect effect (NIE) captures not only a component's causal effect but also interaction effects with other components. It demonstrates these hidden interactions in the GPT-2 IOI circuit and argues that they are a diagnostic tool rather than a nuisance.

This paper investigates why instruction-tuned language models give different answers to causal reasoning questions when variable names are replaced with placeholders, finding that the issue stems from representational misalignment rather than information loss. The authors introduce Vernier, a method using paired-view weight updates and mechanism inspection to reveal that answer-relevant content is still present in the placeholder view but misaligned.

The paper proposes the Unlearning Depth Score (UDS), a metric that uses activation patching to quantify how thoroughly target knowledge is erased from LLMs, achieving state-of-the-art faithfulness and robustness across multiple unlearning methods.

Proposes interaction locality, a task-geometry-aware framework for measuring whether information flow in spatial reasoning models stays within local cells or crosses into global structure, and applies it to HRM, TRM, and MTU3D models on grid benchmarks and embodied 3D grounding.

This paper proposes a five-stage methodology for causal feature analysis in transformer language models, demonstrated on GPT-2 small for the IOI task. It finds that features are specifically causal but not necessary, and exposes a gap between detection and causal robustness.

This paper investigates how large language models process emotional valence through mechanistic interpretability. Using activation patching and steering on three open-source LLMs, the authors find that negative valence is localized to early layers while positive valence peaks in mid-to-late layers, and they validate this through topic-controlled flip tests.

This paper presents causal evidence that hallucination in autoregressive language models results from early trajectory commitment governed by asymmetric attractor dynamics, using same-prompt bifurcation and activation patching experiments on Qwen2.5-1.5B to show that hallucinated trajectories diverge at the first token and exhibit strong causal asymmetry across model layers.