论文检索

Social Aspects · Accountability, Transparency and Interpretability

Angie Boggust, Donghao Ren, Yannick Assogba, Dominik Moritz, Arvind Satyanarayan, Fred Hohman

Automated interpretability aims to translate large language model (LLM) features into human understandable descriptions. However, natural language feature descriptions can be vague, inconsistent, and require manual relabeling. In response, we introduce *semantic regexes*, structured language descriptions of LLM features. By combining primitives that capture linguistic and semantic patterns with modifiers for contextualization, composition, and quantification, semantic regexes produce precise and expressive feature descriptions. Across quantitative benchmarks and qualitative analyses, semantic regexes match the accuracy of natural language while yielding more concise and consistent feature descriptions. Their inherent structure affords new types of analyses, including quantifying feature complexity across layers, scaling automated interpretability from insights into individual features to model-wide patterns. Finally, in user studies, we find that semantic regexes help people build accurate mental models of LLM features.

Social Aspects · Accountability, Transparency and Interpretability

Zhengfu He, Junxuan Wang, Rui Lin, Xuyang Ge, Wentao Shu, Qiong Tang, Junping Zhang, Xipeng Qiu

We propose Low-Rank Sparse Attention (Lorsa), a sparse replacement model of Transformer attention layers to disentangle original Multi Head Self Attention (MHSA) into individually comprehensible components. Lorsa is designed to address the challenge of \textit{attention superposition} to understand attention-mediated interaction between features in different token positions. Lorsa helps find cleaner and finer-grained versions of previously discovered MHSA behaviors like induction heads, successor heads, attention sink, and a comprehensive family of arithmetic-specific Lorsa heads. Interestingly, we identify a novel head type called \emph{subtoken induction heads} that function at character level rather than token level. Automated interpretability analysis indicates that Lorsa achieves parity with SAE in interpretability while Lorsa exhibits superior circuit discovery properties. We also conduct extensive experiments on architectural design ablation, correlation to original MHSA heads and error analysis. Our early attempt to fully sparsify a toy Transformer succeeds to reveal clean global circuits. Eventually, we hope Lorsa would help us greatly understand attention computation and enable full sparsification of model computation along with its MLP counterparts. Lorsa is open-sourced at https://anonymous.4open.science/r/Lorsa-5686/.

Social Aspects · Accountability, Transparency and Interpretability

Wengang Li, Lingqi Zhang, Toshio Endo, Mohamed Wahib

Mixture-of-Experts (MoE) has been a prevalent method for scaling up large language models at a reduced computational cost. Despite its effectiveness, the routing mechanism of MoE still lacks a clear understanding from the perspective of cross-layer mechanistic interpretability. We propose a light-weight methodology at which we can break down the routing decision for MoE to the contribution of model components, in a recursive fashion. We use our methodology to dissect the routing mechanism by decomposing the input of routers into model components. We study how different model components contribute to the routing in different widely used open models. Our findings on four different LLMs reveal patterns such as: a) MoE layer outputs usually contribute more than attention layer outputs to the routing decisions of subsequent layers, b) *MoE entanglement* at which MoE firing up in layers consistently correlate with firing up of MoE in subsequent layers, and c) some components can persistently influence the routing in many following layers. Our study also includes findings on how different models have different patterns when it comes to long-range and short-range inhibiting/promoting effects that components can have over MoE in subsequent layers. Our results indicate the importance of quantifying the impact of components across different layers on MoE to understand the routing mechanism.

Social Aspects · Accountability, Transparency and Interpretability

Luke Guerdan, Justin Whitehouse, Kimberly Truong, Ken Holstein, Steven Wu

As Generative AI (GenAI) systems see growing adoption, a key concern involves the external validity of evaluations, or the extent to which they generalize from lab-based to real-world deployment conditions. Threats to the external validity of GenAI evaluations arise when the source sample of human raters and system outputs used to obtain a system quality estimate differs from the target distribution at deployment time. In this work, we propose a doubly-robust estimation framework designed to address this evaluation sampling bias. Key to our approach is the use of synthetic "persona" ratings -- produced by prompting an LLM evaluator (i.e., an LLM-as-a-judge) to behave as a human rater with specific sociodemographic characteristics. Our doubly-robust framework combines these informative yet imperfect persona ratings with human ratings obtained under evaluation sampling bias to produce statistically valid system quality estimates. In particular, we show that our approach yields valid system quality estimates when either: (i) a model trained to predict human ratings using persona ratings and source data observed under sampling bias, or (ii) a reweighting model that corrects for sampling bias is of sufficient quality. We validate our framework theoretically and via a novel Persona Simulation Framework (PSF) designed to systematically manipulate persona quality and the degree of evaluation sampling bias present in source data. Our work provides a principled foundation for combining imperfect persona ratings with human ratings observed under sampling bias to obtain valid system quality estimates.

Social Aspects · Accountability, Transparency and Interpretability

Michael Hanna, Emmanuel Ameisen

LLMs can perform seemingly planning-intensive tasks, like writing coherent stories or functioning code, without explicitly verbalizing a plan; however, the extent to which they implicitly plan is unknown. In this paper, we define *latent planning* as occurring when LLMs possess internal planning representations that (1) cause the generation of a specific future token or concept, and (2) shape preceding context to license said future token or concept. We study the Qwen-3 family (0.6B-14B) on simple planning tasks, finding that latent planning ability increases with scale. Models that plan possess features that represent a planned-for word like *accountant*, and cause them to output *an* rather than *a*; moreover, even the less-successful Qwen-3 4B-8B have nascent planning mechanisms. On the more complex task of completing rhyming couplets, we find that models often identify a rhyme ahead of time, but even large models seldom plan far ahead. However, we can elicit some planning that increases with scale when steering models towards planned words in prose. In sum, we offer a framework for measuring planning and mechanistic evidence of how models' planning abilities grow with scale.

Social Aspects · Accountability, Transparency and Interpretability

jacob drori, Luke Marks, Bryce Woodworth, Alex Cloud, Alexander Turner

Recently, OpenAI (2025) showed that training against a chain of thought (CoT) monitor can cause obfuscated CoTs, which contain bad behavior the monitor cannot detect. They proposed to keep CoTs monitorable by training only against output monitors that do not have access to CoT. We show that such training can still cause obfuscated CoTs via two mechanisms. First, when a model is trained to produce a safe-looking output, that model may generalize to making its CoTs look safe. Second, since later tokens are conditioned on earlier ones, safe‑looking CoTs may increase the likelihood of safe outputs, causing safe-looking CoTs to be reinforced. We introduce two mitigations to address these two issues, which achieve a Pareto improvement in terms of monitorability and task performance compared to regular training. To our knowledge, we are the first to identify and mitigate these problems. Our work implies that preserving CoT monitorability is more difficult than previously thought; we suggest practical guidelines for AI developers to maintain monitorable CoTs.

Social Aspects · Accountability, Transparency and Interpretability

Bharat Chandra Yalavarthi, Nalini Ratha, Venu Govindaraju

Vision-Language Models (VLMs) have demonstrated impressive zero-shot capabilities across a wide range of tasks and domains. However, their performance is often compromised by learned spurious correlations, which can adversely affect downstream applications. Existing mitigation strategies typically depend on additional data, model retraining, labeled features or classes, domain-specific expertise, or external language models posing scalability and generalization challenges. In contrast, we introduce a fully interpretable, zero-shot method that requires no auxiliary data or external supervision named DIAL (Disentangle, Identify, And Label-free removal). Our approach begins by filtering the representations that might be disproportionately influenced by spurious features, using distributional analysis. We then apply a sparse autoencoder to disentangle the representations and identify the feature directions associated with spurious features. To mitigate their impact, we remove the subspace spanned by these spurious directions from the affected representations. Additionally, for cases where prior knowledge of spurious features in a dataset is unknown, we introduce DIAL+ which can detect and mitigate the spurious features. We validate our method through extensive experiments on widely used spurious correlation benchmarks. Results show that our approach consistently outperforms or matches existing baselines in terms of overall accuracy and worst-group performance, offering a scalable and interpretable solution to a persistent challenge in VLMs.

Social Aspects · Accountability, Transparency and Interpretability

Xinting Huang, Michael Hahn

Understanding internal representations of neural models is a core interest of mechanistic interpretability. Due to its large dimensionality, the representation space can encode various aspects about inputs. To what extent are different aspects organized and encoded in separate subspaces? Is it possible to find these "natural" subspaces in a purely unsupervised way? Somewhat surprisingly, we can indeed achieve this and find interpretable subspaces by a seemingly unrelated training objective. Our method, neighbor distance minimization (NDM), learns non-basis-aligned subspaces in an unsupervised manner. Qualitative analysis shows subspaces are interpretable in many cases, and encoded information in obtained subspaces tends to share the same abstract concept across different inputs, making such subspaces similar to "variables" used by the model. We also conduct quantitative experiments using known circuits in GPT-2; results show a strong connection between subspaces and circuit variables. We also provide evidence showing scalability to 2B models by finding separate subspaces mediating context and parametric knowledge routing. Viewed more broadly, our findings offer a new perspective on understanding model internals and building circuits.

Social Aspects · Accountability, Transparency and Interpretability

Dasha Metropolitansky, Jonathan Larson

Even when instructed to adhere to source material, language models often generate unsubstantiated content – a phenomenon known as “closed-domain hallucination.” This risk is amplified in processes with multiple generative steps (MGS), compared to processes with a single generative step (SGS). However, due to the greater complexity of MGS processes, we argue that detecting hallucinations in their final outputs is necessary but not sufficient: it is equally important to trace where hallucinated content was likely introduced and how faithful content may have been derived from the source material through intermediate outputs. To address this need, we present VeriTrail, the first closed-domain hallucination detection method designed to provide traceability for both MGS and SGS processes. We also introduce the first datasets to include all intermediate outputs as well as human annotations of final outputs’ faithfulness for their respective MGS processes. We demonstrate that VeriTrail outperforms baseline methods on both datasets.

Social Aspects · Accountability, Transparency and Interpretability

Mark Russinovich, Yanan Cai, Ahmed Salem

Growing concerns over the theft and misuse of Large Language Models (LLMs) underscore the need for effective fingerprinting to link a model to its original version and detect misuse. We define five essential properties for a successful fingerprint: Transparency, Efficiency, Persistence, Robustness, and Unforgeability. We present a novel fingerprinting framework that provides verifiable proof of ownership while preserving fingerprint integrity. Our approach makes two main contributions. First, a "chain and hash" technique that cryptographically binds fingerprint prompts to their responses, preventing collisions and enabling irrefutable ownership claims. Second, we address a realistic threat model in which instruction-tuned models' output distribution can be significantly altered through meta-prompts. By incorporating random padding and varied meta-prompt configurations during training, our method maintains robustness even under significant output style changes. Experiments show that our framework securely proves ownership, resists both benign transformations (e.g., fine-tuning) and adversarial fingerprint removal, and extends to fingerprinting LoRA adapters. We release our code at: https://github.com/microsoft/Chain-Hash.

Social Aspects · Accountability, Transparency and Interpretability

Rafid Mahmood

To mitigate the trade-offs between performance and costs, LLM providers route user tasks to different models based on task difficulty and latency. We study the effect of LLM routing with respect to user behavior. We propose a game between an LLM provider with two models (standard and reasoning) and a user who can re-prompt or abandon tasks if the routed model cannot solve them. The user's goal is to maximize their utility minus the delay from using the model, while the provider minimizes the cost of servicing the user. We solve this Stackelberg game by fully characterizing the user best response and simplifying the provider problem. We observe that in nearly all cases, the optimal routing policy involves a static policy with no cascading that depends on the expected utility of the models to the user. Furthermore, we reveal a misalignment gap between the provider-optimal and user-preferred routes when the user's and provider's rankings of the models with respect to utility and cost differ. Finally, we demonstrate conditions for extreme misalignment where providers are incentivized to throttle the latency of the models to minimize their costs, consequently depressing user utility. The results yield simple threshold rules for single-provider, single-user interactions and clarify when routing, cascading, and throttling help or harm.

Social Aspects · Accountability, Transparency and Interpretability

Guan Zhe Hong, Bhavya Vasudeva, Vatsal Sharan, Cyrus Rashtchian, Prabhakar Raghavan, Rina Panigrahy

When large language models (LLMs) use in-context learning (ICL) to solve a new task, they must infer latent concepts from demonstration examples. This raises the question of whether and how transformers represent latent structures as part of their computation. Our work experiments with several controlled tasks, studying this question using mechanistic interpretability. First, we show that in transitive reasoning tasks with a latent, discrete concept, the model successfully identifies the latent concept and does step-by-step concept composition. This builds upon prior work that analyzes single-step reasoning. Then, we consider tasks parameterized by a latent numerical concept. We discover low-dimensional subspaces in the model's representation space, where the geometry cleanly reflects the underlying parameterization. Overall, we show that small and large models can indeed disentangle and utilize latent concepts that they learn in-context from a handful of abbreviated demonstrations.

Social Aspects · Accountability, Transparency and Interpretability

Jingyi Qiu, Hong Chen, Zongyi Li

The rise of AI has fueled growing concerns about "hype" in machine learning papers, yet a reliable way to quantify rhetorical style independently of substantive content has remained elusive. Because bold language can stem from either strong empirical results or mere rhetorical style, it is often difficult to distinguish between the two. To disentangle rhetorical style from substantive content, we introduce a counterfactual, LLM-based framework: multiple LLM rhetorical personas generate counterfactual writings from the same substantive content, an LLM judge compares them through pairwise evaluations, and the outcomes are aggregated using a Bradley--Terry model. Applying this method to 8,485 ICLR submissions sampled from 2017 to 2025, we generate more than 250,000 counterfactual writings and provide a large-scale quantification of rhetorical style in ML papers. We find that visionary framing significantly predicts downstream attention, including citations and media attention, even after controlling for peer-review evaluations. We also observe a sharp rise in rhetorical strength after 2023, and provide empirical evidence showing that this increase is largely driven by the adoption of LLM-based writing assistance. The reliability of our framework is validated by its robustness to the choice of personas and the high correlation between LLM judgments and human annotations. Our work demonstrates that LLMs can serve as instruments to measure and improve scientific evaluation.

Social Aspects · Accountability, Transparency and Interpretability

Anjali Parashar, Yingke Li, Eric Yu, Fei Chen, James Neidhoefer, devesh upadhyay, Chuchu Fan

As autonomous systems such as drones, become increasingly deployed in high-stakes, human-centric domains, it is critical to evaluate the ethical alignment since failure to do so imposes imminent danger to human lives, and long term bias in decision-making. Automated ethical benchmarking of these systems is understudied due to the lack of ubiquitous, well-defined metrics for evaluation, and stakeholder-specific subjectivity, which cannot be modeled analytically. To address these challenges, we propose SEED-SET, a Bayesian experimental design framework that incorporates domain-specific objective evaluations, and subjective value judgments from stakeholders. SEED-SET models both evaluation types separately with hierarchical Gaussian Processes, and uses a novel acquisition strategy to propose interesting test candidates based on learnt qualitative preferences and objectives that align with the stakeholder preferences. We validate our approach for ethical benchmarking of autonomous agents on two applications and find our method to perform the best. Our method provides an interpretable and efficient trade-off between exploration and exploitation, by generating up to $2\times$ optimal test candidates compared to baselines, with $1.25\times$ improvement in coverage of high dimensional search spaces.

Social Aspects · Accountability, Transparency and Interpretability

Shiyuan Guo, Henry Sleight, Fabien Roger

Detecting harmful AI actions is important as AI agents gain adoption. Chain-of-thought (CoT) monitoring is one method widely used to detect adversarial attacks and AI misalignment. However, attackers and misaligned models might evade CoT monitoring through *ciphered reasoning*: reasoning hidden in encrypted, translated, or compressed text. To assess this risk, we test whether models can perform ciphered reasoning. For each of 28 different ciphers, we fine-tune and prompt up to 10 models to reason in that cipher. We measure model accuracy on math problems as a proxy for reasoning ability. Across the models we test, we find an asymmetry: model accuracy can drop significantly when reasoning in ciphered text, even though models demonstrate comprehension of ciphered text by being able to translate it accurately to English. Even frontier models struggle with lesser-known ciphers, although they can reason accurately in well-known ciphers like rot13. We show that ciphered reasoning capability correlates with cipher prevalence in pretraining data. We also identify scaling laws showing that ciphered reasoning capability improves slowly with additional fine-tuning data. Our work suggests that evading CoT monitoring using ciphered reasoning may be an ineffective tactic for current models and offers guidance on constraining the development of this capability in future frontier models.

Social Aspects · Accountability, Transparency and Interpretability

Nikhil Singh, Manuel Cherep, Pattie Maes

The fidelity with which neural networks can now generate content such as music presents a scientific opportunity: these systems appear to have learned implicit theories of such content's structure through statistical learning alone. This offers a potentially new lens on theories of human-generated media. When internal representations align with traditional constructs (e.g. chord progressions in music), they show how such categories can emerge from statistical regularities; when they diverge, they expose limits of existing frameworks and patterns we may have overlooked but that nonetheless carry explanatory power. In this paper, focusing on autoregressive music generators, we introduce a method for discovering interpretable concepts using sparse autoencoders (SAEs), extracting interpretable features from the residual stream of a transformer model. We make this approach scalable and evaluable using automated labeling and validation pipelines. Our results reveal both familiar musical concepts and coherent but uncodified patterns lacking clear counterparts in theory or language. As an extension, we show such concepts can be used to steer model generations. Beyond improving model transparency, our work provides an empirical tool for uncovering organizing principles that have eluded traditional methods of analysis and synthesis.

Social Aspects · Accountability, Transparency and Interpretability

Julian Minder, Clément Dumas, Stewart Slocum, Helena Casademunt, Cameron Holmes, Robert West, Neel Nanda

Finetuning on narrow domains has become an essential tool to adapt Large Language Models (LLMs) to specific tasks and to create models with known unusual properties that are useful for safety research. Model diffing--the study of differences between base and finetuned models--is a promising approach for understanding how finetuning modifies neural networks. In this paper, we show that narrow finetuning creates easily readable biases in LLM activations that can be detected using simple model diffing tools, suggesting that the finetuning data is overrepresented in the model's activations. In particular, analyzing activation differences between base and finetuned models on the first few tokens of random text and steering with this difference allows us to recover the format and general content of the finetuning data. We call this the Activation Difference Lens (ADL). We demonstrate that these analyses significantly enhance an LLM-based interpretability agent's ability to identify subtle finetuning objectives through interaction with base and finetuned models. Our analysis spans synthetic document finetuning for false facts, emergent misalignment, subliminal learning, and taboo guessing game models across different architectures (Gemma, LLaMA, Qwen) and scales (1B to 32B parameters). Our work: (1) demonstrates that researchers should be aware that narrow finetuned models will represent their training data and objective very saliently, (2) warns AI safety and mechanistic interpretability researchers that these models might not be a realistic proxy for studying broader finetuning, despite current literature widely using them. While we show that mixing pretraining data into the finetuning corpus is enough to remove this bias, a deeper investigation is needed to understand the side effects of narrow finetuning and develop truly realistic case studies for model-diffing, safety and interpretability research.

Social Aspects · Accountability, Transparency and Interpretability

Ziyang Guo, Yifan Wu, Jason Hartline, Jessica Hullman

Multiple agents are increasingly combined to make decisions with the expectation of achieving *complementary performance*, where the decisions they make together outperform those made individually. However, knowing how to improve the performance of collaborating agents requires knowing what information and strategies each agent employs. With a focus on human-AI pairings, we contribute a decision-theoretic framework for characterizing the value of information. By defining complementary information, our approach identifies opportunities for agents to better exploit available information in AI-assisted decision workflows. We present a novel explanation technique (ILIV-SHAP) that adapts SHAP explanations to highlight human-complementing information. We validate the effectiveness of our framework and ILIV-SHAP through a study of human-AI decision-making, and demonstrate the framework on examples from chest X-ray diagnosis and deepfake detection. We find that presenting ILIV-SHAP with AI predictions leads to reliably greater reductions in error over non-AI assisted decisions more than vanilla SHAP.

Social Aspects · Accountability, Transparency and Interpretability

Manuel Cherep, Chengtian Ma, Abigail Xu, Maya Shaked, Pattie Maes, Nikhil Singh

Environments built for people are increasingly operated by a new class of economic actors: LLM-powered software agents making decisions on our behalf. These decisions range from our purchases to travel plans to medical treatment selection. Current evaluations of these agents largely focus on task competence, but we argue for a deeper assessment: how these agents choose when faced with realistic decisions. We introduce ABxLab, a framework for systematically probing agentic choice through controlled manipulations of option attributes and persuasive cues. We apply this to a realistic web-based shopping environment, where we vary prices, ratings, and psychological nudges, all of which are factors long known to shape human choice. We find that agent decisions shift predictably and substantially in response, revealing that agents are strongly biased choosers even without being subject to the cognitive constraints that shape human biases. This susceptibility reveals both risk and opportunity: risk, because agentic consumers may inherit and amplify human biases; opportunity, because consumer choice provides a powerful testbed for a behavioral science of AI agents, just as it has for the study of human behavior. We release our framework as an open benchmark for rigorous, scalable evaluation of agent decision-making.

Social Aspects · Accountability, Transparency and Interpretability

Jiaqi Li, Zihan You, Ruoyan Shen, Shenyu Zhang, Songlin Zhai, Yongrui Chen, Chuanyi Zhang, Jiahui Geng, Fakhri Karray, Sheng Bi 等

Multimodal Large Language Models (MLLMs) achieve remarkable cross-modal understanding by training on vast web-scale datasets, but inadvertently internalize sensitive personal and proprietary information. Existing machine unlearning methods address this by irreversibly altering model parameters to permanently erase knowledge. This destructive paradigm conflicts with modern privacy regulations that mandate auditable, reversible, and user-controllable data management. To address these challenges, we propose Knowledge Externalization, a novel framework for reversible and modular knowledge management in MLLMs. We first propose Dual-Stream Memory Tuning, a method that transfers targeted knowledge from a model's internal parameters into external memory tokens. To mitigate gradient interference when externalizing multiple concepts, we further introduce Soft Orthogonal Weighting, a technique that preserves the independence of each token. Our resulting framework demonstrates three key capabilities: (i) It achieves effective forgetting of target concepts within the base model, while enabling high-fidelity knowledge restoration using the corresponding memory token. (ii) It supports continuous knowledge editing, allowing the information stored within an external token to be dynamically updated post-externalization. (iii) It displays a remarkable emergent ability for compositionality, where multiple memory tokens (including edited ones) can be freely combined to simultaneously recover knowledge corresponding to each concept. Our source code will be released in the near future.