论文检索

Social Aspects · Accountability, Transparency, and Interpretability

Aaron Mueller, Atticus Geiger, Sarah Wiegreffe, Dana Arad, Iván Arcuschin, Adam Belfki, Yik Siu Chan, Jaden Fiotto-Kaufman, Tal Haklay, Michael Hanna 等

How can we know whether new mechanistic interpretability methods achieve real improvements?In pursuit of lasting evaluation standards, we propose MIB, a Mechanistic Interpretability Benchmark, with two tracks spanning four tasks and five models. MIB favors methods that precisely and concisely recover relevant causal pathways or causal variables in neural language models. The circuit localization track compares methods that locate the model components---and connections between them---most important for performing a task (e.g., attribution patching or information flow routes). The causal variable track compares methods that featurize a hidden vector, e.g., sparse autoencoders (SAE) or distributed alignment search (DAS), and align those features to a task-relevant causal variable. Using MIB, we find that attribution and mask optimization methods perform best on circuit localization. For causal variable localization, we find that the supervised DAS method performs best, while SAEs features are not better than neurons, i.e., non-featurized hidden vectors. These findings illustrate that MIB enables meaningful comparisons, and increases our confidence that there has been real progress in the field.

Social Aspects · Accountability, Transparency, and Interpretability

Justin S. Kang, Landon Butler, Abhineet Agarwal, Yigit Efe Erginbas, Ramtin Pedarsani, Bin Yu, Kannan Ramchandran

Large language models (LLMs) have revolutionized machine learning due to their ability to capture complex interactions between input features. Popular post-hoc explanation methods like SHAP provide *marginal* feature attributions, while their extensions to interaction importances only scale to small input lengths ($\approx 20$). We propose *Spectral Explainer* (SPEX), a model-agnostic interaction attribution algorithm that efficiently scales to large input lengths ($\approx 1000)$. SPEX exploits underlying natural sparsity among interactions—common in real-world data—and applies a sparse Fourier transform using a channel decoding algorithm to efficiently identify important interactions. We perform experiments across three difficult long-context datasets that require LLMs to utilize interactions between inputs to complete the task. For large inputs, SPEX outperforms marginal attribution methods by up to 20\% in terms of faithfully reconstructing LLM outputs. Further, SPEX successfully identifies key features and interactions that strongly influence model output. For one of our datasets, *HotpotQA*, SPEX provides interactions that align with human annotations. Finally, we use our model-agnostic approach to generate explanations to demonstrate abstract reasoning in closed-source LLMs (*GPT-4o mini*) and compositional reasoning in vision-language models.

General Machine Learning · Causality

Kevin Xia, Elias Bareinboim

The study of causal abstractions bridges two integral components of human intelligence: the ability to determine cause and effect, and the ability to interpret complex patterns into abstract concepts. Formally, causal abstraction frameworks define connections between complicated low-level causal models and simple high-level ones. One major limitation of most existing definitions is that they are not well-defined when considering lossy abstraction functions in which multiple low-level interventions can have different effects while mapping to the same high-level intervention (an assumption called the abstract invariance condition). In this paper, we introduce a new type of abstractions called projected abstractions that generalize existing definitions to accommodate lossy representations. We show how to construct a projected abstraction from the low-level model and how it translates equivalent observational, interventional, and counterfactual causal queries from low to high-level. Given that the true model is rarely available in practice we prove a new graphical criteria for identifying and estimating high-level causal queries from limited low-level data. Finally, we experimentally show the effectiveness of projected abstraction models in high-dimensional image settings.

Deep Learning · Large Language Models

Weihua Du, Yiming Yang, Sean Welleck

Multi-sample aggregation strategies, such as majority voting and best-of-N sampling, are widely used in contemporary large language models (LLMs) to enhance predictive accuracy across various tasks. A key challenge in this process is temperature selection, which significantly impacts model performance. Existing approaches either rely on a fixed default temperature or require labeled validation data for tuning, which are often scarce and difficult to obtain. This paper addresses the challenge of automatically identifying the (near)-optimal temperature for different LLMs using multi-sample aggregation strategies, without relying on task-specific validation data. We provide a comprehensive analysis of temperature’s role in performance optimization, considering variations in model architectures, datasets, task types, model sizes, and predictive accuracy. Furthermore, we propose a novel entropy-based metric for automated temperature optimization, which consistently outperforms fixed-temperature baselines. Additionally, we incorporate a stochastic process model to enhance interpretability, offering deeper insights into the relationship between temperature and model performance.

Social Aspects · Accountability, Transparency, and Interpretability

Jiahai Feng, Stuart Russell, Jacob Steinhardt

Pretrained language models (LMs) can generalize to implications of facts that they are finetuned on. For example, if finetuned on "John Doe lives in Tokyo," LMs correctly answer "What language do the people in John Doe's city speak?'' with "Japanese''. However, little is known about the mechanisms that enable this generalization or how they are learned during pretraining.We introduce extractive structures as a framework for describing how components in LMs (e.g., MLPs or attention heads) coordinate to enable this generalization. The structures consist of informative components that store training facts as weight changes, and upstream and downstream extractive components that query and process the stored information to produce the correct implication. We hypothesize that extractive structures are learned during pretraining when encountering implications of previously known facts. This yields two predictions: a data ordering effect where extractive structures can be learned only if facts precede their implications, and a weight grafting effect where extractive structures can be grafted to predict counterfactual implications.We empirically show these effects in the OLMo-7b, Llama 3-8b, Gemma 2-9b, and Qwen 2-7b models.Of independent interest, our results also indicate that fact learning can occur at both early and late layers, which lead to different forms of generalization.

Position · Methodology

Oliver Eberle, Thomas McGee, Hamza Giaffar, Taylor Webb, Ida Momennejad

What algorithms do LLMs actually learn and use to solve problems? Studies addressing this question are sparse, as research priorities are focused on improving performance through scale, leaving a theoretical and empirical gap in understanding emergent algorithms. This position paper proposes AlgEval: a framework for systematic research into the algorithms that LLMs learn and use. AlgEval aims to uncover algorithmic primitives, reflected in latent representations, attention, and inference-time compute, and their algorithmic composition to solve task-specific problems. We highlight potential methodological paths and a case study toward this goal, focusing on emergent search algorithms. Our case study illustrates both the formation of top-down hypotheses about candidate algorithms, and bottom-up tests of these hypotheses via circuit-level analysis of attention patterns and hidden states. The rigorous, systematic evaluation of how LLMs actually solve tasks provides an alternative to resource-intensive scaling, reorienting the field toward a principled understanding of underlying computations. Such algorithmic explanations offer a pathway to human-understandable interpretability, enabling comprehension of the model's internal reasoning performance measures. This can in turn lead to more sample-efficient methods for training and improving performance, as well as novel architectures for end-to-end and multi-agent systems.

Position · Understanding

John Hewitt, Robert Geirhos, Been Kim

This position paper argues that, in order to understand AI, we cannot rely on our existing vocabulary of human words. Instead, we shouldstrive to develop neologisms: new words thatrepresent precise human concepts that we wantto teach machines, or machine concepts that weneed to learn. We start from the premise thathumans and machines have differing concepts.This means interpretability can be framed as acommunication problem: humans must be able toreference and control machine concepts, and communicate human concepts to machines. Creatinga shared human-machine language through developing neologisms, we believe, could solve thiscommunication problem. Successful neologismsachieve a useful amount of abstraction: not toodetailed, so they’re reusable in many contexts, andnot too high-level, so they convey precise information. As a proof of concept, we demonstrate howa “length neologism” enables controlling LLMresponse length, while a “diversity neologism” allows sampling more variable responses. Takentogether, we argue that we cannot understand AIusing our existing vocabulary, and expanding itthrough neologisms creates opportunities for bothcontrolling and understanding machines better.

Position · Methodology

Yunke Wang, Yanxi Li, Chang Xu

AI Scaling has traditionally been synonymous with Scaling Up, which builds larger and more powerful models. However, the growing demand for efficiency, adaptability, and collaboration across diverse applications necessitates a broader perspective. This position paper presents a holistic framework for AI scaling, encompassing Scaling Up, Scaling Down, and Scaling Out. It argues that while Scaling Up of models faces inherent bottlenecks, the future trajectory of AI scaling lies in Scaling Down and Scaling Out. These paradigms address critical technical and societal challenges, such as reducing carbon footprint, ensuring equitable access, and enhancing cross-domain collaboration. We explore transformative applications in healthcare, smart manufacturing, and content creation, demonstrating how AI Scaling can enable breakthroughs in efficiency, personalization, and global connectivity. Additionally, we highlight key challenges, including balancing model complexity with interpretability, managing resource constraints, and fostering ethical development. By synthesizing these approaches, we propose a unified roadmap that redefines the future of AI research and application, paving the way for advancements toward Artificial General Intelligence (AGI).

Applications · Everything Else

Herman Chau, Helen Jenne, Davis Brown, Jesse He, Mark Raugas, Sara Billey, Henry Kvinge

With recent dramatic increases in AI system capabilities, there has been growing interest in utilizing machine learning for reasoning-heavy, quantitative tasks, particularly mathematics. While there are many resources capturing mathematics at the high-school, undergraduate, and graduate level, there are far fewer resources available that align with the level of difficulty and open endedness encountered by professional mathematicians working on open problems. To address this, we introduce a new collection of datasets, the Algebraic Combinatorics Dataset Repository (ACD Repo), representing either foundational results or open problems in algebraic combinatorics, a subfield of mathematics that studies discrete structures arising from abstract algebra. Further differentiating our dataset collection is the fact that it aims at the conjecturing process. Each dataset includes an open-ended research level question and a large collection of examples (up to 10M in some cases) from which conjectures should be generated. We describe all nine datasets, the different ways machine learning models can be applied to them (e.g., training with narrow models followed by interpretability analysis or program synthesis with LLMs), and discuss some of the challenges involved in designing datasets like these.

Deep Learning · Everything Else

Qian-Yuan Tang, Yufei Gu, Yunfeng Cai, Mingming Sun, Ping Li, zhou Xun, Zeke Xie

It is well-known that the Hessian of deep loss landscape matters to optimization and generalization of deep learning. Previous studies reported a rough Hessian structure in deep learning, which consists of two components, a small number of large eigenvalues and a large number of nearly-zero eigenvalues. To the best of our knowledge, we are the first to report that a simple but overlooked power-law Hessian structure exists in well-trained deep neural networks, including Convolutional Neural Networks (CNNs) and Large Language Models (LLMs). Moreover, we provide a maximum-entropy theoretical interpretation for the power-law Hessian structure and theoretically demonstrate the existence of robust and low-dimensional subspace of deep neural networks. Our extensive experiments using the proposed power-law spectral method demonstrate that the power-law Hessian spectra critically relate to multiple important behaviors of deep learning, including optimization, generalization, and overparameterization. Notably, we discover that the power-law Hessian structure of a given LLM can effectively predict generalization during training, while conventional sharpness-based generalization measures that often works well on CNNs become nearly useless for as a generalization predictor of LLMs.

Social Aspects · Accountability, Transparency, and Interpretability

Shahaf Bassan, Yizhak Elboher, Tobias Ladner, Matthias Althoff, Guy Katz

Despite significant advancements in post-hoc explainability techniques for neural networks, many current methods rely on heuristics and do not provide formally provable guarantees over the explanations provided. Recent work has shown that it is possible to obtain explanations with formal guarantees by identifying subsets of input features that are sufficient to determine that predictions remain unchangedusing neural network verification techniques.Despite the appeal of these explanations, their computation faces significant scalability challenges. In this work, we address this gap by proposing a novel abstraction-refinement technique for efficiently computing provably sufficient explanations of neural network predictions. Our method *abstracts* the original large neural network by constructing a substantially reduced network, where a sufficient explanation of the reduced network is also *provably sufficient* for the original network, hence significantly speeding up the verification process. If the explanation is insufficient on the reduced network, we iteratively *refine* the network size by gradually increasing it until convergence.Our experiments demonstrate that our approach enhances the efficiency of obtaining provably sufficient explanations for neural network predictions while additionally providing a fine-grained interpretation of the network's predictions across different abstraction levels.

Social Aspects · Accountability, Transparency, and Interpretability

Bartosz Cywiński, Kamil Deja

Diffusion models, while powerful, can inadvertently generate harmful or undesirable content, raising significant ethical and safety concerns. Recent machine unlearning approaches offer potential solutions but often lack transparency, making it difficult to understand the changes they introduce to the base model. In this work, we introduce SAeUron, a novel method leveraging features learned by sparse autoencoders (SAEs) to remove unwanted concepts in text-to-image diffusion models. First, we demonstrate that SAEs, trained in an unsupervised manner on activations from multiple denoising timesteps of the diffusion model, capture sparse and interpretable features corresponding to specific concepts. Building on this, we propose a feature selection method that enables precise interventions on model activations to block targeted content while preserving overall performance. Our evaluation shows that SAeUron outperforms existing approaches on the UnlearnCanvas benchmark for concepts and style unlearning, and effectively eliminates nudity when evaluated with I2P. Moreover, we show that with a single SAE, we can remove multiple concepts simultaneously and that in contrast to other methods, SAeUron mitigates the possibility of generating unwanted content under adversarial attack. Code and checkpoints are available at [GitHub]( https://github.com/cywinski/SAeUron).

Deep Learning · Robustness

Peimeng Guan, Mark Davenport

Inverse problems aim to reconstruct unseen data from corrupted or perturbed measurements. While most work focuses on improving reconstruction quality, generalization accuracy and robustness are equally important, especially for safety-critical applications. Model-based architectures (MBAs), such as loop unrolling methods, are considered more interpretable and achieve better reconstructions. Empirical evidence suggests that MBAs are more robust to perturbations than black-box solvers, but the accuracy-robustness tradeoff in MBAs remains underexplored. In this work, we propose a simple yet effective training scheme for MBAs, called SGD jittering, which injects noise iteration-wise during reconstruction. We theoretically demonstrate that SGD jittering not only generalizes better than the standard mean squared error training but is also more robust to average-case attacks. We validate SGD jittering using denoising toy examples, seismic deconvolution, and single-coil MRI reconstruction. Both SGD jittering and its SPGD extension yield cleaner reconstructions for out-of-distribution data and demonstrates enhanced robustness against adversarial attacks.

Deep Learning · Large Language Models

Xiang Li, Neil Chowdhury, Daniel Johnson, Tatsunori Hashimoto, Percy Liang, Sarah Schwettmann, Jacob Steinhardt

Language models exhibit complex, diverse behaviors when prompted with free-form text, making it hard to characterize the space of possible outputs. We study the problem of behavioral elicitation, where the goal is to search for prompts that induce specific target behaviors (e.g., hallucinations, harmful responses) from a target language model. To navigate the exponentially large space of possible prompts, we train amortized investigator models to emulate the posterior distribution over the prompts, conditioned on the target behavior. Specifically, we first fit a reverse model and then use reinforcement learning to optimize likelihood of generating the target behavior. To improve the diversity of the prompt distribution, we further propose a novel iterative training objective based on the Frank-Wolfe algorithm that encourages each iteration to discover different sets of prompts not captured by previous iterations. Our investigator models produce prompts that exhibit a variety of effective and human-interpretable strategies for behavior elicitation, obtaining a 100% attack success rate on AdvBench (Harmful Behaviors) and an 85% hallucination rate.

Probabilistic Methods · Bayesian Models and Methods

Jake Snell, Thomas Griffiths

As machine learning-based prediction systems are increasingly used in high-stakes situations, it is important to understand how such predictive models will perform upon deployment. Distribution-free uncertainty quantification techniques such as conformal prediction provide guarantees about the loss black-box models will incur even when the details of the models are hidden. However, such methods are based on frequentist probability, which unduly limits their applicability. We revisit the central aspects of conformal prediction from a Bayesian perspective and thereby illuminate the shortcomings of frequentist guarantees. We propose a practical alternative based on Bayesian quadrature that provides interpretable guarantees and offers a richer representation of the likely range of losses to be observed at test time.

General Machine Learning · Transfer, Multitask and Meta-learning

Zitao Wang, Ziyuan Wang, Molei Liu, Nian Si

Wasserstein Distributionally Robust Optimization (WDRO) is a principled framework for robust estimation under distributional uncertainty. However, its standard formulation can be overly conservative, particularly in small-sample regimes. We propose a novel knowledge-guided WDRO (KG-WDRO) framework for transfer learning, which adaptively incorporates multiple sources of external knowledge to improve generalization accuracy. Our method constructs smaller Wasserstein ambiguity sets by controlling the transportation along directions informed by the source knowledge. This strategy can alleviate perturbations on the predictive projection of the covariates and protect against information loss. Theoretically, we establish the equivalence between our WDRO formulation and the knowledge-guided shrinkage estimation based on collinear similarity, ensuring tractability and geometrizing the feasible set. This also reveals a novel and general interpretation for recent shrinkage-based transfer learning approaches from the perspective of distributional robustness. In addition, our framework can adjust for scaling differences in the regression models between the source and target and accommodates general types of regularization such as lasso and ridge. Extensive simulations demonstrate the superior performance and adaptivity of KG-WDRO in enhancing small-sample transfer learning.

Applications · Language, Speech and Dialog

Martin Andrews, Sam Witteveen

Cryptic crossword clues are challenging language tasks for which new test sets are released daily by major newspapers on a global basis. Each cryptic clue contains both the definition of the answer to be placed in the crossword grid (in common with regular crosswords), and ‘wordplay’ that *proves* that the answer is correct (i.e. a human solver can be confident that an answer is correct without needing crossing words as confirmation). This work describes an LLM-based reasoning system built from open-licensed components that solves cryptic clues by (i) hypothesising answers; (ii) proposing wordplay explanations; and (iii) using a verifier system that operates on codified reasoning steps. Overall, this system establishes a new state-of-the-art performance on the challenging Cryptonite dataset of clues from The Times and The Telegraph newspapers in the UK. Because each proved solution is expressed in Python, interpretable wordplay reasoning for proven answers is available for inspection.

Deep Learning · Graph Neural Networks

Cheng Xin, Fan Xu, Xin Ding, Jie Gao, Jiaxin Ding

Graph Neural Networks (GNNs) have shown remarkable success across various scientific fields,yet their adoption in critical decision-making is often hindered by a lack of interpretability. Recently,intrinsic interpretable GNNs have been studied to provide insights into model predictions by identifying rationale substructures in graphs. However, existing methods face challenges when the underlying rationale subgraphs are complex and varied. In this work, we propose TopInG: Topologically Interpretable Graph Learning, a novel topological framework that leverages persistent homology to identify persistent rationale subgraphs. TopInG employs a rationale filtration learning approach to model an autoregressive generating process of rationale subgraphs, and introduces a self-adjusted topological constraint, termed topological discrepancy, to enforce a persistent topological distinction between rationale subgraphs and irrelevant counterparts. We provide theoretical guarantees that our loss function is uniquely optimized by the ground truth under specific conditions. Extensive experiments demonstrate TopInG's effectiveness in tackling key challenges, such as handling variform rationale subgraphs, balancing predictive performance with interpretability, and mitigating spurious correlations. Results show that our approach improves upon state-of-the-artmethods on both predictive accuracy and interpretation quality.

Social Aspects · Accountability, Transparency, and Interpretability

Qingyu Yin, Chak Tou Leong, Hongbo Zhang, Minjun Zhu, Hanqi Yan, Qiang Zhang, Yulan He, Wenjie Li, Jun Wang, Yue Zhang 等

The alignment of large language models (LLMs) with human preferences remains a key challenge. While post-training techniques like Reinforcement Learning from Human Feedback (RLHF) and Direct Preference Optimization (DPO) have achieved notable success, they often experience computational inefficiencies and training instability. In this paper, we propose \textbf{F}eature-level constrained \textbf{P}reference \textbf{O}ptimization (FPO), a novel method designed to simplify the alignment process while ensuring stability. FPO leverages pre-trained Sparse Autoencoders (SAEs) and introduces feature-level constraints, allowing for efficient, sparsity-enforced alignment. Our approach enjoys efficiency by using sparse features activated in a well-trained sparse autoencoder and the quality of sequential KL divergence by using the feature-level offline reference. Experimental results on benchmark datasets demonstrate that FPO achieves an above 5\% absolute improvement in win rate with much lower computational cost compared to state-of-the-art baselines, making it a promising solution for efficient and controllable LLM alignments.

General Machine Learning · Transfer, Multitask and Meta-learning

Sarthak Mittal, Eric Elmoznino, Léo Gagnon, Sangnie Bhardwaj, Guillaume Lajoie, Dhanya Sridhar

Large autoregressive models like Transformers can solve tasks through in-context learning (ICL) without learning new weights, suggesting avenues for efficiently solving new tasks. For many tasks, e.g., linear regression, the data factorizes: examples are independent given a task latent that generates the data, e.g., linear coefficients. While an optimal predictor leverages this factorization by inferring task latents, it is unclear if Transformers implicitly do so or instead exploit heuristics and statistical shortcuts through attention layers. In this paper, we systematically investigate the effect of explicitly inferring task latents by minimally modifying the Transformer architecture with a bottleneck to prevent shortcuts and incentivize structured solutions. We compare it against standard Transformers across various ICL tasks and find that contrary to intuition and recent works, there is little discernible difference between the two; biasing towards task-relevant latent variables does not lead to better out-of-distribution performance, in general. Curiously, we find that while the bottleneck effectively learns to extract latent task variables from context, downstream processing struggles to utilize them for robust prediction. Our study highlights the intrinsic limitations of Transformers in achieving structured ICL solutions that generalize, and shows that while inferring the right latents aids interpretability, it is not sufficient to alleviate this problem.