Mechanistic Interpretability

Recent momentum

-8%

11 papers in the last 28 days · 0.2% of indexed attention

Twelve weeks of publication activity for this topic as it is defined today.

Weekly history

Recent digests

What was published in this topic, kept on the site without email delivery.

Period ending 2026-09-21

4 new papers

A weekly snapshot of new work published in Mechanistic Interpretability.

Period ending 2026-09-14

2 new papers

A weekly snapshot of new work published in Mechanistic Interpretability.

Period ending 2026-09-07

6 new papers

A weekly snapshot of new work published in Mechanistic Interpretability.

158 papers

Latest in Mechanistic Interpretability

May 7, 2026cs.LG

PLOT: Progressive Localization via Optimal Transport in Neural Causal Abstraction

Causal abstraction offers a principled framework for mechanistic interpretability, aligning a high-level causal model with the low-level computation realized by a neural network through counterfactual intervention analysis. Existing methods such as distributed alignment search (DAS) learn expressive subspace interventions, but the relevant neural site is unknown a priori, so finding a handle requires a computationally burdensome search over candidate sites. We introduce PLOT (Progressive Localization via Optimal Transport), a transport-based framework that localizes causal variables from the output effect geometry of abstract and neural interventions. PLOT fits an optimal transport coupling between abstract variables and candidate neural sites, yielding a global soft correspondence that can be calibrated into intervention handles. In simple settings, a single coupling over individual neurons suffices. In larger models, PLOT is applied progressively, moving from coarse sites such as tokens, timesteps, or layers to finer supports such as coordinate groups or PCA spans, and optionally guiding DAS based on the localized signal. Across experiments of increasing complexity, transport-only PLOT handles are exceedingly fast and competitive on accuracy, while PLOT-guided DAS reaches DAS-level accuracy at a fraction of full DAS runtime, providing an efficient localization engine for causal abstraction research at scale.
Jonathn Chang, Arya Datla, Ziv Goldfeld
May 7, 2026cs.AI

Beyond the Black Box: Interpretability of Agentic AI Tool Use

AI agents are promising for high-stakes enterprise workflows, but dependable deployment remains limited because these tool-use decisions are difficult to diagnose and control. Agents may skip required tool calls, invoke tools unnecessarily, or take actions whose consequences become visible only after execution. Existing observability methods are external: prompts reveal correlations, evaluations score outputs, and logs arrive only after the model has already acted. In long-horizon settings, these failures are costly because an early tool mistake can alter the rest of the execution trajectory, increase token consumption, and create downstream safety and security risk. We introduce a mechanistic-interpretability toolkit built on Sparse Autoencoders (SAEs), which decompose activations into sparse internal features, and linear probes, lightweight classifiers that read signals from those features. The framework reads model states before each action and infers whether a tool is needed and how risky the next tool action is. It identifies the model layers and features most associated with tool decisions and tests their functional importance through feature ablation. We train the probes on multi-step agent execution traces from the NVIDIA Nemotron function-calling dataset and apply the same workflow to GPT-OSS 20B and Gemma 3 27B models. The goal is not to replace external evaluation, but to add a missing layer: visibility into what the model signaled internally before action. This helps surface deeper causes of agent failure, especially in long-horizon runs where an early mistake can impact subsequent agent behavior. More broadly, the paper shows how mechanistic interpretability can support internal observability for monitoring tool calls and risk in agent systems.
Hariom Tatsat, Ariye Shater
May 7, 2026cs.LG

SoftSAE: Dynamic Top-K Selection for Adaptive Sparse Autoencoders

Sparse Autoencoders (SAEs) have become an important tool in mechanistic interpretability, helping to analyze internal representations in both Large Language Models (LLMs) and Vision Transformers (ViTs). By decomposing polysemantic activations into sparse sets of monosemantic features, SAEs aim to translate neural network computations into human-understandable concepts. However, common architectures such as TopK SAEs rely on a fixed sparsity level. They enforce the same number of active features (K) across all inputs, ignoring the varying complexity of real-world data. Natural data often lies on manifolds with varying local intrinsic dimensionality, meaning the number of relevant factors can change significantly across samples. This suggests that a fixed sparsity level is not optimal. Simple inputs may require only a few features, while more complex ones need more expressive representations. Using a constant K can therefore introduce noise in simple cases or miss important structure in more complex ones. To address this issue, we propose SoftSAE, a sparse autoencoder with a Dynamic Top-K selection mechanism. Our method uses a differentiable Soft Top-K operator to learn an input-dependent sparsity level k. This allows the model to adjust the number of active features based on the complexity of each input. As a result, the representation better matches the structure of the data, and the explanation length reflects the amount of information in the input. Experimental results confirm that SoftSAE not only finds meaningful features, but also selects the right number of features for each concept. The source code is available at: https://github.com/St0pien/SoftSAE.
Jakub Stępień, Marcin Mazur, Jacek Tabor +1
May 7, 2026cs.AI

From Token Lists to Graph Motifs: Weisfeiler-Lehman Analysis of Sparse Autoencoder Features

Sparse autoencoders (SAEs) have become central to mechanistic interpretability, decomposing transformer activations into monosemantic features. Yet existing analyses characterise features almost exclusively through top-activating token lists or decoder weight vectors, leaving the higher-order co-occurrence structure shared across features largely unexamined. We introduce a graph-structured representation in which each SAE feature is modelled as a token co-occurrence graph: nodes are the tokens most frequent near strong activations, and edges connect pairs that co-occur within local context windows. A custom WL-style, frequency-binned graph kernel then provides a similarity measure over this structural space. Applied as a proof of concept to features from a large SAE trained on GPT-2 Small and probed with a synthetic mixed-domain corpus, our clustering recovers heuristic motif families (punctuation-heavy patterns, language and script clusters, and code-like templates) that are not recovered by clustering on decoder cosine similarity. A token-histogram baseline achieves higher overall purity, so the contribution of the graph view is complementary rather than dominant: it surfaces structural relationships that token-frequency and decoder-weight views alone do not capture. Cluster assignments are stable across graph-construction hyperparameters and random seeds.
Ruben Fernandez-Boullon, Pablo Magariños-Docampo, Javier Perez-Robles
May 7, 2026cs.AI

Patch-Effect Graph Kernels for LLM Interpretability

Mechanistic interpretability aims to reverse-engineer transformer computations by identifying causal circuits through activation patching. However, scaling these interventions across diverse prompts and task families produces high-dimensional, unstructured datasets that are difficult to compare systematically. We propose a framework that reframes mechanistic analysis as a graph machine-learning problem by representing activation-patching profiles as patch-effect graphs over model components. We introduce three graph-construction methods: direct-influence via causal mediation, partial-correlation, and co-influence and apply graph kernels to analyze the resulting structures. Evaluating this approach on GPT-2 Small using Indirect Object Identification (IOI) and related tasks, we find that patch-effect graphs preserve discriminative structural signals. Specifically, localized edge-slot features provide higher classification accuracy than global graph-shape descriptors. A screened paired-patching validation suggests that CI and PC selected candidate edges correspond to stronger activation-influence effects than random or low-rank candidates. Crucially, by evaluating these representations against rigorous prompt-only and raw patch-effect controls, we make the evidential scope of the benchmark explicit: graph features compress structured patching signal, while raw tensors and surface cues define strong baselines that any circuit-level claim should address. Ultimately, our framework provides a compression and evaluation pipeline for comparing patching-derived structures under controlled baselines, separating robust slice-discriminative evidence from stronger task-general causal-circuit claims.
Ruben Fernandez-Boullon, David N. Olivieri
May 7, 2026cs.LG

Playing the network backward: A Game Theoretic Attribution Framework

Attribution methods explain which input features drive a model's prediction, making them central to model debugging and mechanistic interpretability. Yet backward attribution methods, including gradients, LRP, and transformer-specific rules, lack a shared framework in which to compare the underlying backward calculations. We introduce such a framework by recasting backward attribution as a two-player game on an extended network graph, building on Gaubert and Vlassopoulos' ReLU Net Game. Gradients and the full alpha-beta-LRP family arise as integrals over game trajectories under specific equilibria, so attribution maps become projections of trajectory distributions rather than the primary object. Desired explanation properties, such as localisation focus, robustness to input noise, or stable attention routing, can be specified as game-theoretic concepts, including policy regularization, risk aversion, and extended action sets, and translate directly into novel adaptations of the well-known backward rules. On ViT-B/16, one such selected adaptation of alpha-beta-LRP outperforms prior transformer-specific backward methods across all considered localisation metrics.
Jakob Paul Zimmermann, Jim Berend, Georg Loho +2
May 7, 2026cs.CL

Negative Before Positive: Asymmetric Valence Processing in Large Language Models

Mechanistic interpretability has revealed how concepts are encoded in large language models (LLMs), but emotional content remains poorly understood at the mechanistic level. We study whether LLMs process emotional valence through dedicated internal structure or through surface token matching. Using activation patching and steering on open-source LLMs, we find that negative and positive valence are processed at different network depths. Negative outcomes localize to early layers while positive outcomes peak at mid-to-late layers. Holding topic fixed while flipping valence produces sign-opposite responses, ruling out topic detection. Steering with the good-news direction at the identified layers shifts neutral prompts toward positive valence, showing these layers encode valence as a manipulable direction. Emotional valence in LLMs is localized, causal and steerable, making it a concrete target for interpretability-based oversight.
Sohan Venkatesh
May 6, 2026cs.LG

Superposition Is Not Necessary: A Mechanistic Interpretability Analysis of Transformer Representations for Time Series Forecasting

Transformer architectures have been widely adopted for time series forecasting, yet whether the representational mechanisms that make them powerful in NLP actually engage on time series data remains unexplored. The persistent competitiveness of simple linear models such as DLinear has fueled ongoing debate, but no mechanistic explanation for this phenomenon has been offered. We address this gap by applying sparse autoencoders (SAEs), a tool from mechanistic interpretability, to probe the internal representations of PatchTST. We first establish that a single-layer, narrow-dimensional transformer matches the forecasting performance of deeper configurations across commonly used benchmarks. We then train SAEs on the post-GELU intermediate FFN activations with dictionary sizes ranging from 0.5x to 4.0x the native dimensionality. Expanding the dictionary yields negligible downstream performance change (average 0.214%), with large portions of overcomplete dictionaries remaining inactive. Targeted causal interventions on dominant latent features produce minimal forecast perturbation. Across all evaluated settings, we observe no empirical evidence that the analyzed FFN representations rely on strong superposition. Instead, the representations remain sparse, stable under aggressive dictionary expansion, and largely insensitive to latent interventions. These results demonstrate that superposition is not necessary for competitive performance on standard forecasting benchmarks, suggesting they may not demand the rich compositional representations that drive transformer success in language modeling, and helping explain the persistent competitiveness of simple linear models
Alper Yıldırım
May 4, 2026cs.LG

Moral Sensitivity in LLMs: A Tiered Evaluation of Contextual Bias via Behavioral Profiling and Mechanistic Interpretability

Large language models (LLMs) are increasingly deployed in settings that require nuanced ethical reasoning, yet existing bias evaluations treat model outputs as simply "biased" or "unbiased." This binary framing misses the gradual, context-sensitive way bias actually emerges. We address this gap in two stages: behavioral profiling and mechanistic validation. In the behavioral stage, we introduce the Moral Sensitivity Index (MSI), a metric that quantifies the probability of biased output across a graduated, seven-tier stress test ranging from abstract numerical problems to scenarios rooted in historical and socioeconomic injustice. Evaluating four leading models (Claude 3.5, Qwen 3.5, Llama 3, and Gemini 1.5), we identify distinct behavioral signatures shaped by alignment design: for instance, Gemini 1.5 reaches 72.7% MSI by Tier 5 under socioeconomic framing, while Claude exhibits sharp suppression consistent with identity-based safety training. We then verify these behavioral patterns mechanistically. We select criminal-bias scenarios, which produced the highest MSI scores across models, as probes and apply logit lens, attention analysis, activation patching, and semantic probing to a controlled set of six models spanning three capability tiers: small language models (SLMs), instruction-tuned base models, and reasoning-distilled variants. Circuit-level analysis reveals a U-curve of bias: SLMs exhibit strong criminal bias; scaling to instruction-tuned models eliminates it; reasoning distillation reintroduces bias to SLM-like levels despite identical parameter counts, suggesting distillation compresses reasoning traces in ways that reactivate shallow statistical associations. Critically, the socially loaded cues that drive high MSI scores activate the same bias-driving circuits identified mechanistically, providing cross-stage validation.
Yash Aggarwal, Atmika Gorti, Vinija Jain +3
May 4, 2026cs.LG

Neuron-Anchored Rule Extraction for Large Language Models via Contrastive Hierarchical Ablation

A central goal of explainable AI is to express large language model (LLM) decision logic symbolically and ground it in internal mechanisms. Existing rule-extraction methods usually learn ungrounded symbolic surrogates, while mechanistic interpretability links behavior to neurons but often requires hand-crafted hypotheses and costly interventions. We introduce MechaRule, a pipeline that grounds rule extraction in LLM circuits by localizing sparse agonist activations whose ablation disrupts rule-related behavior. MechaRule rests on two findings. First, in a fixed baseline/flip regime, sparse agonist effects can exhibit overtopping: a few high-effect activations remain detectable within larger groups, dominate weaker ones, and flip many of the same examples. In such regimes, adaptive group testing with confidence-guided conservative pruning requires O(k log(N/k) + k) interventions over N candidates when k << N are agonists. Second, agonists are localized more reliably on data splits aligned with close-to-faithful rule behavior; spectral splits provide a rule-free fallback, whereas unfaithful splits degrade localization. Empirically, on arithmetic and jailbreaking, MechaRule recalls 97.0% of highest-effect agonists in matched brute-force validations at only 2.14% of exhaustive-ablation cost on average. Ablating the localized agonists eliminates 97.6--100.0% of eligible correct arithmetic answers and jailbreaks, and can correct arithmetic errors or induce jailbreaks by up to 72.8% and 32.5%.
Francesco Sovrano, Gabriele Dominici, Marc Langheinrich
May 2, 2026cs.CL

Automated Interpretability and Feature Discovery in Language Models with Agents

We introduce an autonomous multiagent framework for mechanistic interpretability that automates both explaining and finding internal features in large language models. The system runs two coupled loops: (1) explanation refinement, where an agent proposes competing hypotheses and iteratively tests them with targeted prompt controls and a multi-metric evaluation; and (2) feature discovery, where an agent generates prompt sets, constructs a k-nearest-neighbor graph in activation space, and retrieves candidate features using statistical separability and semantic coherence criteria. On Gemma-2 family models and MLP neurons in weight-sparse transformers, our agent improves over one-shot auto-interpretations, discovers language-specific and safety-relevant features, and produces auditable explanation traces, showing that agent-driven empirical loops yield sharper and more falsifiable explanations than one-shot labels.
Arnau Marin-Llobet, Javier Ferrando
May 1, 2026cs.LG

Knowing when to trust machine-learned interatomic potentials

Prevailing machine-learned interatomic potential (MLIP) uncertainty-quantification methods rely on ensembles of independently trained backbones. These methods scale unfavorably with foundation-scale MLIPs, and their member-disagreement signals correlate weakly with per-molecule prediction error. Here we probe the frozen per-atom representations of a pretrained MLIP with a compact discriminative classifier, recasting MLIP uncertainty quantification as selective classification rather than error regression. The resulting method, PROBE (Post-hoc Reliability frOm Backbone Embeddings), produces a per-prediction reliability probability that monotonically tracks actual error without modification to the underlying model. Across large held-out evaluation sets and two structurally distinct MLIP architectures, PROBE outperforms ensemble disagreement as a binary reliability signal, which strengthens with the expressiveness of the backbone representation, implying a favorable scaling trajectory toward foundation-scale MLIPs. Multi-head self-attention additionally yields per-atom importance maps, providing chemically interpretable diagnostics at no additional computational cost. PROBE is post-hoc and architecture-agnostic, and is directly deployable on any MLIP that exposes per-atom representations.
Shams Mehdi, Ilkwon Cho, Olexandr Isayev
Apr 28, 2026cs.LG

reward-lens: A Mechanistic Interpretability Library for Reward Models

Every RLHF-trained language model is shaped by a reward model, yet the mechanistic interpretability toolkit -- logit lens, direct logit attribution, activation patching, sparse autoencoders -- was built for generative LLMs whose primitives all project onto a vocabulary unembedding. Reward models replace that with a scalar regression head, breaking each tool. We present reward-lens, an open-source library that ports this toolkit to reward models, organised around one observation: the reward head's weight vector wrw_r is the natural axis for every interpretability question. The library provides a Reward Lens, component attribution, three-mode activation patching, a reward-hacking probe suite, TopK SAE feature attribution, cross-model comparison, and five theory-grounded extensions (distortion index, divergence-aware patching, misalignment cascade detection, reward-term conflict analysis, concept-vector analysis). A ten-method adapter protocol covers Llama, Mistral, Gemma-2, and ArmoRM multi-objective heads, with a generic adapter for any HuggingFace sequence classification model. We validate on two production reward models across ~695 RewardBench pairs. The central empirical finding is negative: linear attribution does not predict causal patching effects (mean Spearman ρ=0.256ρ= -0.256 on Skywork, 0.027-0.027 on ArmoRM). The framework treats this disagreement as a property to expose, not a bug -- motivating a design that keeps observational and causal views first-class and directly comparable.
Mohammed Suhail B Nadaf
Apr 28, 2026cs.CL

Quantifying Aleatoric Uncertainty of In-Context Learning for Robust Measure of LLM Prediction Confidence

In-Context Learning (ICL) allows LLMs to adapt to new tasks from a few demonstrations, but its reliability remains a concern: predictions are highly sensitive to both prompt design and the model's ability to understand the context, obscuring whether failures arise from data properties or model limitations. Uncertainty decomposition-separating aleatoric from epistemic sources-is particularly crucial in this setting, yet existing methods, designed for standard generation tasks, fail to capture the unique dynamics of ICL. To address this, we introduce a concept of self-function vectors, built upon Bayesian views and the mechanistic interpretability of ICL. These vectors leverage internal model representations to model the latent concept learned during in-context prompting, thereby enabling a direct estimation of aleatoric uncertainty within a Bayesian framework and circumventing the reliance on brittle input or decoding manipulations. Given the lack of established benchmarks and suitable evaluation protocols, we also propose the first and rigorous evaluation protocol, in which data is manipulated in controlled ways so as to quantify aleatoric uncertainty precisely and separately from epistemic uncertainty. With this new evaluation framework, initially grounded in synthetic tasks for conceptual development and subsequently extended to real-world datasets, we show that our proposed methodology can measure uncertainty of LLM predictions made under ICL more reliably than existing alternative methods. Moreover, we show it can be used as a practical tool for trustworthy-related applications, such as hallucination detection. Our findings pave a new direction for connecting the quantitative view of uncertainty with the mechanistic understanding of model behavior.
Jinseok Chung, Minkyoung Song, Hyunji Jung +1
Apr 28, 2026cs.AI

From Insight to Action: A Novel Framework for Interpretability-Guided Data Selection in Large Language Models

While mechanistic interpretability tools like Sparse Autoencoders (SAEs) can uncover meaningful features within Large Language Models (LLMs), a critical gap remains in transforming these insights into practical actions for model optimization. We bridge this gap with the hypothesis that data selection guided by a model's internal task features is a effective training strategy. Inspired by this, we propose Interpretability-Guided Data Selection (IGDS), a framework that first identifies these causal task features through frequency recall and interventional filtering, then selects ``Feature-Resonant Data'' that maximally activates task features for fine-tuning. We validate IGDS on mathematical reasoning, summarization, and translation tasks within Gemma-2, LLaMA-3.1, and Qwen3 models. Our experiments demonstrate exceptional data efficiency: on the Math task, IGDS surpasses full-dataset fine-tuning by a remarkable 17.4% on Gemma-2-2B while using only 50% of the data, and outperforms established baselines focused on data quality and diversity. Analysis confirms a strong positive correlation between feature amplification and task performance improvement. IGDS thus provides a direct and effective framework to enhance LLMs by leveraging their internal mechanisms, validating our core hypothesis.
Ling Shi, Xinwei Wu, Xiaohu Zhao +7
Apr 27, 2026q-bio.MN

Learning biophysical models of gene regulation with probability flow matching

Cellular differentiation is governed by gene regulatory networks, the high-dimensional stochastic biochemical systems that determine the transcriptional landscape and mediate cellular responses to signals and perturbations. Although single-cell RNA sequencing provides quantitative snapshots of the transcriptome, current methods for inferring gene-regulatory dynamics often lack mechanistic interpretability and fail to generalize to unseen conditions. Here we introduce Probability Flow Matching (PFM), a scalable framework for learning biophysically consistent stochastic processes directly from time-resolved single-cell measurements. Applying PFM to three hematopoiesis datasets, we show that models with similar interpolation accuracy can encode fundamentally different dynamics, with only biophysically consistent formulations accurately capturing mechanisms of lineage transitions, fate specification, and gene perturbation responses. We further demonstrate that PFM accommodates unbalanced populations, enabling simultaneous inference of cellular proliferation and death dynamics. Together, these results establish PFM as a flexible, scalable framework for integrating mechanistic modeling with single-cell omics.
Suryanarayana Maddu, Victor Chardès, Michael J. Shelley
Apr 27, 2026cs.CL

Differentiable Faithfulness Alignment for Cross-Model Circuit Transfer

Mechanistic interpretability has made it possible to localize circuits underlying specific behaviors in language models, but existing methods are expensive, model-specific, and difficult to scale to larger architectures. We introduce \textbf{Differentiable Faithfulness Alignment (DFA)}, a framework that transfers circuit information from a smaller source model to a larger target model through a learned differentiable alignment. DFA projects source-model node importance scores into the target model and trains this mapping with a soft faithfulness objective, avoiding full circuit discovery on the target model. We evaluate DFA on Llama-3 and Qwen-2.5 across six tasks spanning factual retrieval, multiple-choice reasoning, and arithmetic. The strongest results occur on Llama-3 11B3\rightarrow3B, where aligned circuits are often competitive with direct node attribution and zero-shot transfer remains effective. Recovery weakens for larger source--target gaps and is substantially lower on Qwen-2.5, suggesting that transfer becomes harder as architectural and scaling differences increase. Overall, DFA consistently outperforms simple baselines and, in some settings, recovers target-model circuits with faithfulness comparable to or stronger than direct attribution. These results suggest that smaller models can provide useful mechanistic priors for larger ones, while highlighting both the promise and the limits of node-level cross-model circuit alignment.\footnote{Code is available at https://github.com/jasonshaoshun/dfa-circuits.
Shun Shao, Binxu Wang, Shay B. Cohen +2
Apr 26, 2026cs.CL

AIPsy-Affect: A Keyword-Free Clinical Stimulus Battery for Mechanistic Interpretability of Emotion in Language Models

Mechanistic interpretability research on emotion in large language models -- linear probing, activation patching, sparse autoencoder (SAE) feature analysis, causal ablation, steering vector extraction -- depends on stimuli that contain the words for the emotions they test. When a probe fires on "I am furious", it is unclear whether the model has detected anger or detected the word "furious". The two readings have very different consequences for every downstream claim about emotion circuits, features, and interventions. We release AIPsy-Affect, a 480-item clinical stimulus battery that removes the confound at the stimulus level: 192 keyword-free vignettes evoking each of Plutchik's eight primary emotions through narrative situation alone, 192 matched neutral controls that share characters, setting, length, and surface structure with the affect surgically removed, plus moderate-intensity and discriminant-validity splits. The matched-pair structure supports linear probing, activation patching, SAE feature analysis, causal ablation, and steering vector extraction under a strong methodological guarantee: any internal representation that distinguishes a clinical item from its matched neutral cannot be doing so on the basis of emotion-keyword presence. A three-method NLP defense battery -- bag-of-words sentiment, an emotion-category lexicon, and a contextual transformer classifier -- confirms the property: bag-of-words methods see only situational vocabulary, and a contextual classifier detects affect (p < 10^-15) but cannot identify the category (5.2% top-1 vs. 82.5% on a keyword-rich control). AIPsy-Affect extends our earlier 96-item battery (arXiv:2603.22295) by a factor of four and is released openly under MIT license.
Michael Keeman
Apr 24, 2026cs.CY

Make Mechanistic Interpretability Auditable: A Call to Develop Guidelines via Continuous Collaborative Reviewing

While mechanistic interpretability (MI) has produced important insights into neural network internals, the field has yet to establish a standardized system to audit experiments. As such, many of its findings remain underutilized in safety-critical applications such as medical AI and autonomous systems, as stakeholders cannot certify their validity. Recent work demonstrates this concretely: two papers found conflicting conclusions for the same behavior, and a third study revealed that both were partially correct but incomparable due to methodological inconsistencies. Without standardized auditing, such ambiguities hinder adoption in high-stakes contexts requiring strong correctness guarantees. We call for the MI community to work towards developing a novel reviewing system that complements peer review via: (1) Continuous reviewing supported by a \emph{Collaborative Reviewing Platform} where meta-science results and discussions (such as critiques, negative results, post-hoc extensions, reproductions, replications, and partial results) that fit outside of papers are organized and discussed, allowing for comments and revisions to be made at any time (2) Generalizing good practices found on this platform into expert-verified guidelines and protocols to improve auditing efficiency, and (3) Source-based auditing systems that track arguments which claims depend on. This position paper encourages constructive debate over the necessity, design and implementation of such a framework, providing early concrete examples to help catalyze these dialogues. Overall, we propose that auditing MI itself is essential for its application in AI safety, industry, and governance.
Michael Lan, Narmeen Fatimah Oozeer, Chaithanya Bandi +4
Apr 24, 2026cs.CL

Fine-Grained Analysis of Shared Syntactic Mechanisms in Language Models

While language models demonstrate sophisticated syntactic capabilities, the extent to which their internal mechanisms align with cross-constructional principles studied in linguistics remains poorly understood. This study investigates whether models employ shared neural mechanisms across different syntactic constructions by applying causal interpretability methods at a granular level. Focusing on filler-gap dependencies and negative polarity item (NPI) licensing, we utilize activation patching to identify the functional roles of specific attention heads and MLP blocks. Our results reveal a highly localized and shared mechanism for filler-gap dependencies located in the early to middle layers, whereas NPI processing exhibits no such unified mechanism. Furthermore, we find that these mechanisms identified by activation patching generalize to out-of-distribution, while distributed alignment search, a supervised interpretability method, is susceptible to overfitting on narrow linguistic distributions. Finally, we validate our findings by demonstrating that the manipulation of the identified components improves model performance on acceptability judgment benchmarks.
Ryoma Kumon, Hitomi Yanaka
Apr 23, 2026physics.ed-ph

Locating acts of mechanistic reasoning in student team conversations with mechanistic machine learning

STEM education researchers are often interested in identifying moments of students' mechanistic reasoning for deeper analysis, but have limited capacity to search through many team conversation transcripts to find segments with a high concentration of such reasoning. We offer a solution in the form of an interpretable machine learning model that outputs time-varying probabilities that individual students are engaging in acts of mechanistic reasoning, leveraging evidence from their own utterances as well as contributions from the rest of the group. Using the toolkit of intentionally-designed probabilistic models, we introduce a specific inductive bias that steers the probabilistic dynamics toward desired, domain-aligned behavior. Experiments compare trained models with and without the inductive bias components, investigating whether their presence improves the desired model behavior on transcripts involving never-before-seen students and a novel discussion context. Our results show that the inductive bias improves generalization -- supporting the claim that interpretability is built into the model for this task rather than imposed post hoc. We conclude with practical recommendations for STEM education researchers seeking to adopt the tool and for ML researchers aiming to extend the model's design. Overall, we hope this work encourages the development of mechanistically interpretable models that are understandable and controllable for both end users and model designers in STEM education research.
Kaitlin Gili, Mainak Nistala, Kristen Wendell +1
Apr 23, 2026stat.ML

There Will Be a Scientific Theory of Deep Learning

In this paper, we make the case that a scientific theory of deep learning is emerging. By this we mean a theory which characterizes important properties and statistics of the training process, hidden representations, final weights, and performance of neural networks. We pull together major strands of ongoing research in deep learning theory and identify five growing bodies of work that point toward such a theory: (a) solvable idealized settings that provide intuition for learning dynamics in realistic systems; (b) tractable limits that reveal insights into fundamental learning phenomena; (c) simple mathematical laws that capture important macroscopic observables; (d) theories of hyperparameters that disentangle them from the rest of the training process, leaving simpler systems behind; and (e) universal behaviors shared across systems and settings which clarify which phenomena call for explanation. Taken together, these bodies of work share certain broad traits: they are concerned with the dynamics of the training process; they primarily seek to describe coarse aggregate statistics; and they emphasize falsifiable quantitative predictions. We argue that the emerging theory is best thought of as a mechanics of the learning process, and suggest the name learning mechanics. We discuss the relationship between this mechanics perspective and other approaches for building a theory of deep learning, including the statistical and information-theoretic perspectives. In particular, we anticipate a symbiotic relationship between learning mechanics and mechanistic interpretability. We also review and address common arguments that fundamental theory will not be possible or is not important. We conclude with a portrait of important open directions in learning mechanics and advice for beginners. We host further introductory materials, perspectives, and open questions at learningmechanics.pub.
Jamie Simon, Daniel Kunin, Alexander Atanasov +11
Apr 22, 2026physics.ao-ph

Mechanistic Interpretability Tool for AI Weather Models

Artificial Intelligence (AI) weather models are improving rapidly, and their forecasts are already competitive with long-established traditional Numerical Weather Prediction (NWP). To build confidence in this new methodology, it is critical that we understand how these predictions are generated. This is a huge challenge as these AI weather models remain largely black boxes. In other areas of Machine Learning (ML), mechanistic interpretability has emerged as a framework for understanding ML predictions by analysing the building blocks responsible for them. Here we present an open-source, highly adaptable tool which incorporates concepts from mechanistic interpretability. The tool organises internal latent representations from the model processor and allows for initial analyses, including cosine similarity and Principal Component Analysis (PCA), enabling the user to identify directions in latent space potentially associated with meteorological features. Applying our tool to the graph neural network GraphCast, we present preliminary case studies for mid-latitude synoptic-scale waves and specific humidity. These demonstrate the tool's ability to identify linear combinations of latent channels that appear to correspond to interpretable features.
Kirsten I. Tempest, Matthias Beylich, George C. Craig
Feb 26, 2026cs.LG

MetaOthello: A Controlled Study of Multiple World Models in Transformers

Foundation models must handle multiple generative processes, yet mechanistic interpretability largely studies capabilities in isolation; it remains unclear how a single transformer organizes multiple, potentially conflicting "world models". Previous experiments on Othello playing neural-networks test world-model learning but focus on a single game with a single set of rules. We introduce MetaOthello, a controlled suite of Othello variants with shared syntax but different rules or tokenizations, and train small GPTs on mixed-variant data to study how multiple world models are organized in a shared representation space. We find that transformers trained on mixed-game data do not partition their capacity into isolated sub-models; instead, they converge on a mostly shared board-state representation that transfers causally across variants. Linear probes trained on one variant can intervene on another's internal state with effectiveness approaching that of matched probes. For isomorphic games with token remapping, representations are equivalent up to a single orthogonal rotation that generalizes across layers. When rules partially overlap, early layers maintain game-agnostic representations while a middle layer identifies game identity, and later layers specialize. MetaOthello offers a path toward understanding not just whether transformers learn world models, but how they organize many at once.
Aviral Chawla, Galen Hall, Juniper Lovato
Feb 20, 2026cs.LG

A Probabilistic Framework for LLM-Based Model Discovery

Automated methods for discovering mechanistic simulator models from observational data offer a promising path toward accelerating scientific progress. Such methods often take the form of agentic-style iterative workflows that repeatedly propose and revise candidate models by imitating human discovery processes. However, existing LLM-based approaches typically implement such workflows via hand-crafted heuristic procedures, without an explicit probabilistic formulation. We recast model discovery as probabilistic inference, i.e., as sampling from an unknown distribution over mechanistic models capable of explaining the data. This perspective provides a unified way to reason about model proposal, refinement, and selection within a single inference framework. As a concrete instantiation of this view, we introduce ModelSMC, an algorithm based on Sequential Monte Carlo sampling. ModelSMC represents candidate models as particles which are iteratively proposed and refined by an LLM, and weighted using likelihood-based criteria. Experiments on real-world scientific systems illustrate that this formulation discovers models with interpretable mechanisms and improves posterior predictive checks. More broadly, this perspective provides a probabilistic lens for understanding and developing LLM-based approaches to model discovery.
Stefan Wahl, Raphaela Schenk, Ali Farnoud +2
Feb 8, 2026cs.CL

Patches of Nonlinearity: Instruction Vectors in Large Language Models

Despite the recent success of instruction-tuned language models and their ubiquitous usage, very little is known of how models process instructions internally. In this work, we address this gap from a mechanistic point of view by investigating how instruction-specific representations are constructed and utilized in different stages of post-training: Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO). Via causal mediation, we identify that instruction representation is fairly localized in models. These representations, which we call Instruction Vectors (IVs), demonstrate a curious juxtaposition of linear separability along with non-linear causal interaction, broadly questioning the scope of the linear representation hypothesis commonplace in mechanistic interpretability. To disentangle the non-linear causal interaction, we propose a novel method to localize information processing in language models that is free from the implicit linear assumptions of patching-based techniques. We find that, conditioned on the task representations formed in the early layers, different information pathways are selected in the later layers to solve that task, i.e., IVs act as circuit selectors.
Irina Bigoulaeva, Jonas Rohweder, Subhabrata Dutta +1
Feb 4, 2026cs.LG

Towards Isolated Interventions via Almost Orthogonal Features in Language Models

A central premise in mechanistic interpretability is that meaningful concepts in language models are represented by linear features in activation space. For such features to support reliable interventions, manipulating one feature should not substantially alter the effects of others. In practice, however, feature entanglement leads to interference such that localized interventions can have unintended downstream effects. Motivated by the \textit{Independent Causal Mechanisms} principle, we propose to constrain internal features to be almost orthogonal. We argue that this promotes modular representations amenable to causal intervention. We formalize this problem by characterizing the gap between an idealized isolated intervention and its realized effect on model outputs in terms of feature interference. We upper-bound the propagation of feature interference in terms of the self-coherence of the feature dictionary, and relate this discrepancy to an explicit orthogonality regularization on the dictionary itself. Empirically, we show that this regularization enables more isolated interventions on mathematical reasoning concepts while preserving model performance. Our code is available under \texttt{https://github.com/mrtzmllr/sae-icm}.
Moritz Miller, Florent Draye, Bernhard Schölkopf
Feb 4, 2026cs.CL

Hierarchical Latent Structures in Data Generation Process Unify Mechanistic Phenomena across Scale

Contemporary studies in mechanistic interpretability have uncovered many puzzling phenomena in the neural information processing of Transformer-based language models, such as induction heads, function vectors, and the Hydra effect. Some of these individual phenomena have been independently tied to different data distributional properties, while some have been loosely associated with model architecture and how Transformers process information. However, a unified understanding of the relationship between data, model architecture, and optimization remains lacking, failing to answer the fundamental question: why do these three phenomena appear universally across different model families and scales, despite their seeming disconnect? In this work, we answer this question by unifying these three phenomena as consequences of hierarchical latent structures in the data generation process, coupled with decorrelated gradients across additive model components and directional concavity in the representation geometry. We validate our theoretical results in a toy model regime and in a large-scale synthetic data regime, comparing them with language models trained on natural language data.
Jonas Rohweder, Subhabrata Dutta, Iryna Gurevych
Feb 2, 2026cs.LG

Universal Redundancies in Time Series Foundation Models

Time Series Foundation Models (TSFMs) leverage extensive pretraining to accurately predict unseen time series during inference, without the need for task-specific fine-tuning. Through large-scale evaluations on standard benchmarks, we find that leading transformer-based TSFMs exhibit redundant components in their intermediate layers. We introduce a set of tools for mechanistic interpretability of TSFMs, including ablations of specific components and direct logit attribution on the residual stream. Our findings are consistent across several leading TSFMs with diverse architectures, and across a diverse set of real-world and synthetic time-series datasets. We discover that all models in our study are robust to ablations of entire layers. Furthermore, we develop a theoretical framework framing transformers as kernel regressors, motivating a purely intrinsic strategy for ablating heads based on the stable rank of the per-head projection matrices. Using this approach, we uncover the specific heads responsible for degenerate phenomena widely observed in TSFMs, such as parroting of motifs from the context and seasonality bias. Our study sheds light on the universal properties of this emerging class of architectures for continuous-time sequence modeling.
Anthony Bao, Venkata Hasith Vattikuti, Jeffrey Lai +1
Jan 6, 2026cs.LG

Bridging Mechanistic Interpretability and Prompt Engineering with Gradient Ascent for Interpretable Persona Control

Controlling emergent behavioral personas (e.g., sycophancy, hallucination) in Large Language Models (LLMs) is critical for AI safety, yet remains a persistent challenge. Existing solutions face a dilemma: manual prompt engineering is intuitive but unscalable and imprecise, while automatic optimization methods are effective but operate as "black boxes" with no interpretable connection to model internals. We propose a novel framework that adapts gradient ascent to LLMs, enabling targeted prompt discovery. In specific, we propose two methods, RESGA and SAEGA, that both optimize randomly initialized prompts to achieve better aligned representation with an identified persona direction. We introduce fluent gradient ascent to control the fluency of discovered persona steering prompts. We demonstrate RESGA and SAEGA's effectiveness across Llama 3.1, Qwen 2.5, and Gemma 3 for steering three different personas, sycophancy, hallucination, and myopic reward. Crucially, on sycophancy, our automatically discovered prompts achieve significant improvement (49.90% compared with 79.24%). By grounding prompt discovery in mechanistically meaningful features, our method offers a new paradigm for controllable and interpretable behavior modification. We release our scripts for RESGA and SAEGA in this github repo: https://github.com/HarshSaini10/RESGA_SAEGA.
Harshvardhan Saini, Yiming Tang, Dianbo Liu
Oct 28, 2025cs.CL

Emergence of Minimal Circuits for Indirect Object Identification in Attention-Only Transformers

Mechanistic interpretability aims to reverse-engineer large language models (LLMs) into human-understandable computational circuits. However, the complexity of pretrained models often obscures the minimal mechanisms required for specific reasoning tasks. In this work, we train small, attention-only transformers from scratch on a symbolic version of the Indirect Object Identification (IOI) task, a benchmark for studying coreference-like reasoning in transformers. Surprisingly, a single-layer model with only two attention heads achieves perfect IOI accuracy, despite lacking MLPs and normalization layers. Through residual stream decomposition, spectral analysis, and embedding interventions, we find that the two heads specialize into additive and contrastive subcircuits that jointly implement IOI resolution. Furthermore, we show that a two-layer, one-head model composes information from the previous layer primarily through query-key interactions. These results demonstrate that task-specific training induces highly interpretable, minimal circuits, offering a controlled testbed for probing the computational foundations of transformer reasoning.
Rabin Adhikari
Oct 14, 2025cs.LG

Attribution Graphs and Causal Probing for Mechanistic Discovery and Bias Repair in Multimodal Generative Learning

We treat the internals of generative models as mechanistic objects rather than black boxes. We introduce \textbf{Attribution Graphs} (AGs), which extend GradCAM++ to circuit-level representations, and \textbf{Causal Probing}, a do-calculus intervention method for identifying causal latent structures, enabling detection and correction of spurious correlations, demographic biases, and misaligned decision circuits during training. We further propose the \textbf{Cognitive Alignment Score (CAS)}, quantifying agreement between model-internal representations and human concepts, a \textbf{saliency-first privacy mechanism} sharing only thresholded attribution nodes, a bias-aware regularizer aligning subgroup statistics, and a Reveal-to-Revise loop integrating attribution signals into parameter updates without separate fine-tuning. Evaluated on CelebA, FairFace, Jigsaw, and HateXplain, our method achieves \textbf{94.1%} accuracy, \textbf{92.3%} macro F1, \textbf{79.4%} IoU-XAI, and \textbf{12.7} FID at 72--76% adversarial robustness, while reducing subgroup disparity ΔbiasΔ_{\mathrm{bias}} by \textbf{41%}, demonstrating that mechanistic interpretability, fairness, and generative performance can be jointly optimized.
Noor Islam S. Mohammad, Uluğ Bayazıt
Jun 18, 2025cs.LG

Interpretability and Generalization Bounds for Learning Spatial Physics

While there are many applications of ML to scientific problems that look promising, visuals can be deceiving. Using numerical analysis techniques, we rigorously quantify the accuracy, convergence rates, and generalization bounds of certain ML models applied to linear differential equations for parameter discovery or solution finding. Beyond the quantity and discretization of data, we identify that the function space of the data is critical to the generalization of the model. A similar lack of generalization is empirically demonstrated for commonly used models, including physics-specific techniques. Counterintuitively, we find that different classes of models can exhibit opposing generalization behaviors. Based on our theoretical analysis, we also introduce a new mechanistic interpretability lens on scientific models whereby Green's function representations can be extracted from the weights of black-box models. Our results inform a new cross-validation technique for measuring generalization in physical systems, which can serve as a benchmark.
Alejandro Francisco Queiruga, Theo Gutman-Solo, Shuai Jiang
May 26, 2025cs.LG

Tensorization is a powerful but underexplored tool for compression and interpretability of neural networks

Tensorizing a neural network involves reshaping some or all of its dense weight matrices into higher-order tensors and approximating them using low-rank tensor network decompositions. This technique has shown promise as a model compression strategy for large-scale neural networks. However, despite encouraging empirical results, tensorized neural networks (TNNs) remain underutilized in mainstream deep learning. In this position paper, we offer a perspective on both the potential and current limitations of TNNs. We argue that TNNs represent a powerful yet underexplored framework for deep learning--one that deserves greater attention from both engineering and theoretical communities. Beyond compression, we highlight the value of TNNs as a flexible class of architectures with distinctive scaling properties and increased interpretability. A central feature of TNNs is the presence of bond indices, which introduce new latent spaces not found in conventional networks. These internal representations may provide deeper insight into the evolution of features across layers, potentially advancing the goals of mechanistic interpretability. We conclude by outlining several key research directions aimed at overcoming the practical barriers to scaling and adopting TNNs in modern deep learning workflows.
Safa Hamreras, Sukhbinder Singh, Román Orús
Mar 8, 2025cs.LG

Using Mechanistic Interpretability to Craft Adversarial Attacks against Large Language Models

Traditional white-box methods for creating adversarial perturbations against LLMs typically rely only on gradient computation from the targeted model, ignoring the internal mechanisms responsible for attack success or failure. Conversely, interpretability studies that analyze these internal mechanisms lack practical applications beyond runtime interventions. We bridge this gap by introducing a novel white-box approach that leverages mechanistic interpretability techniques to craft practical adversarial inputs. Specifically, we first identify acceptance subspaces - sets of feature vectors that do not trigger the model's refusal mechanisms - then use gradient-based optimization to reroute embeddings from refusal subspaces to acceptance subspaces, effectively achieving jailbreaks. This targeted approach significantly reduces computation cost, achieving attack success rates of 80-95% on state-of-the-art models including Gemma2, Llama3.2, and Qwen2.5 within minutes or even seconds, compared to existing techniques that often fail or require hours of computation. We believe this approach opens a new direction for both attack research and defense development. Furthermore, it showcases a practical application of mechanistic interpretability where other methods are less efficient, which highlights its utility. The code and generated datasets are available at https://github.com/Sckathach/subspace-rerouting.
Thomas Winninger, Boussad Addad, Katarzyna Kapusta
Aug 2, 2024cs.AI

Interpreting Global Perturbation Robustness of Image Models using Axiomatic Spectral Importance Decomposition

Perturbation robustness evaluates the vulnerabilities of models, arising from a variety of perturbations, such as data corruptions and adversarial attacks. Understanding the mechanisms of perturbation robustness is critical for global interpretability. We present a model-agnostic, global mechanistic interpretability method to interpret the perturbation robustness of image models. This research is motivated by two key aspects. First, previous global interpretability works, in tandem with robustness benchmarks, e.g. mean corruption error (mCE), are not designed to directly interpret the mechanisms of perturbation robustness within image models. Second, we notice that the spectral signal-to-noise ratios (SNR) of perturbed natural images exponentially decay over the frequency. This power-law-like decay implies that: Low-frequency signals are generally more robust than high-frequency signals -- yet high classification accuracy can not be achieved by low-frequency signals alone. By applying Shapley value theory, our method axiomatically quantifies the predictive powers of robust features and non-robust features within an information theory framework. Our method, dubbed as \textbf{I-ASIDE} (\textbf{I}mage \textbf{A}xiomatic \textbf{S}pectral \textbf{I}mportance \textbf{D}ecomposition \textbf{E}xplanation), provides a unique insight into model robustness mechanisms. We conduct extensive experiments over a variety of vision models pre-trained on ImageNet to show that \textbf{I-ASIDE} can not only \textbf{measure} the perturbation robustness but also \textbf{provide interpretations} of its mechanisms.
Róisín Luo, James McDermott, Colm O'Riordan
Date pendingcs.CL

Where is the Mind? Persona Vectors and LLM Individuation

The individuation problem for large language models asks which entities associated with them, if any, should be identified as minds. We approach this problem through mechanistic interpretability, engaging in particular with recent empirical work on persona vectors, persona space, and emergent misalignment. We argue that three views are the strongest candidates: the virtual instance view and two new views we introduce, the (virtual) instance-persona view and the model-persona view. First, we argue for the virtual instance view on the grounds that attention streams sustain quasi-psychological connections across token-time. Then we present the persona literature, organised around three hypotheses about the internal structure underlying personas in LLMs, and show that the two persona-based views are promising alternatives.
Pierre Beckmann, Patrick Butlin
Date pendingcs.AI

SAEScientist-Bench: Can AI Agents Conduct Autonomous SAE Interpretability Research?

While research on recursive self-improvement (RSI) has predominantly automated model training pipelines, reliable autonomous development demands a missing pillar: post-hoc monitoring and auditing to understand what models learn and ensure safe alignment. Mechanistic interpretability tools are essential to bridge this gap, among which Sparse Autoencoders (SAEs) serve as a cornerstone by isolating interpretable features for model inspection and steering. In this paper, we introduce SAEScientist-Bench to evaluate whether AI agents can act as scientists utilizing SAE tools for autonomous mechanistic discovery. Given a target concept, an agent designs contrastive probes and navigates a Gemma Scope dictionary of 131K+ features in Gemma-2-9B-IT to discover the optimal feature, evaluated against curated expert reference features anchored on Neuronpedia across activation rank, concept selectivity on contrastive texts, and causal steering. Across 10 agent configurations and 20 tasks, frontier agents demonstrate genuine discovery capabilities and lead different evaluation dimensions, but remain well behind the expert baseline, approaching expert levels on separating target concepts from contrastive controls while lagging substantially in causal generation steering. Further analysis reveals that although agents can design contrasts to rule out spurious candidates, they frequently misinterpret experimental measurements. These results establish experimental model understanding as a measurable capability for closed-loop autonomous AI R&D. Our code is available at https://github.com/Trae1ounG/SAEScientist.
Yuqiao Tan, Shizhu He, Jun Zhao +1