Search research

Deep reinforcement learning (DRL) agents achieve strong performance in complex environments, yet their decision-making processes remain difficult to interpret. We introduce SPOT (Sampling Policy Observation Tree), a novel model-agnostic, sampling-based framework for interpreting DRL policies. Given access to the policy and an environment simulator, SPOT constructs an interpretable finite-horizon tree by sampling actions and recursively simulating the resulting successor states. The tree provides an empirical representation of the policy's action preferences and their possible downstream evolution. We provide formal guarantees establishing SPOT's asymptotic recovery of the policy's unique most probable action and characterizing its disagreement behavior under high-entropy policies. We demonstrate SPOT in the SUMO-RL traffic-signal control domain. The case study illustrates how its tree-based representation can be used to inspect policy preferences, compare alternative future trajectories, and reveal downstream behaviors that are not visible through single-timestep feature-attribution methods.

In safety-critical sectors such as robotics and automotive engineering, the deployment of Deep Reinforcement Learning (DRL) is often hindered by the black-box nature of deep neural networks. This lack of transparency poses significant challenges for regulatory compliance and human-agent trust. This paper presents an experimental study aimed at making high-performance continuous control DRL systems interpretable. A policy distillation framework is implemented using the classic Inverted Pendulum benchmark. A high-performance Twin Delayed DDPG (TD3) agent serves as an opaque, continuous teacher model, whose policy is distilled into an interpretable student surrogate based on a shallow Decision Tree. By leveraging a custom physics-aware feature and "Noisy Oracle Rollouts" for dataset generation, the distillation process achieves performance equivalent to the expert teacher. Furthermore, comparative control theory analysis reveals a fundamental trade-off: transitioning from continuous to discrete rule-based control induces high-frequency Bang-Bang actuation and a stable bimodal limit cycle. Simulation results indicate that Bounded-Input Bounded-Output (BIBO) stability is maintained while providing both global and local interpretability for safe autonomous systems.

The wide-scale use of sparse autoencoders (SAEs) as interpretability tools is limited by inconsistent links between SAE features and model behavior. Features with clear activation descriptions may have weak or unexpected causal effects; steering can vary across prompts or oppose the intended direction; and activation-based feature selection can miss features that produce the desired output change. Prior work has studied feature geometry inside the model, where features are computed. We instead study the geometry of changes in model logits caused by feature interventions. We introduce Feature-Effect Geometry Analysis (FEGA), an unsupervised framework that removes the same active SAE feature across contexts and analyzes the resulting cloud of logit changes. Across SAE variants, consistent one-dimensional effects are rare: few features behave like reusable directions. To interpret this variation, we distinguish value-like features, tied to static information such as factual attributes, from pointer-like features, associated with context-dependent operations. Value-like features more often exhibit structured, low-dimensional effects, although these effects typically span several directions. Pointer-like features, by contrast, predominantly exhibit diffuse effects. Our results show that a feature can be interpretable and causally relevant without providing a stable direction for steering.

Advanced Persistent Threats (APTs) are difficult to detect and interpret due to their multi-stage and stealthy nature. While recent autonomous defense systems leverage provenance graphs and learning-based models for detection and mitigation, their outputs remain largely machine-oriented and difficult for analysts to interpret. Large language models (LLMs) offer a promising interface for report generation, but often produce hallucinated or weakly grounded content. In this paper, we propose DeepFaith, an evidence-grounded framework for faithful incident reporting in multi-stage APT defense. DeepFaith transforms structured outputs from autonomous defense and explainability modules into natural-language reports that are explicitly aligned with underlying system evidence. The framework integrates a unified evidence representation, evidence-grounded prompting, faithfulness-aware generation, and post-generation verification to ensure that all generated statements are supported. Experiments in a realistic enterprise testbed demonstrate that DeepFaith improves faithfulness from 0.68 to 0.92, reduces unsupported claims from 0.32 to 0.08, and increases temporal consistency from 0.6 to 0.88, while maintaining concise reports and lower error rates than existing template-based and LLM-based solutions. These results show that evidence-grounded generation enables reliable, interpretable, and actionable reporting for security operations centers.

Recent Vision-Language Models (VLMs) have achieved remarkable success in visual understanding, driven by the growing availability of high-quality image-text pairs. However, the performance of VLMs often degrades in the video domain due to the essential need for temporal understanding and the scarcity of large-scale annotated video data. In this work, we focus on marine video understanding, which brings further challenges: first, it requires substantial domain expertise; and video VLMs usually struggle with localizing and interpreting critical information from marine videos, as the informative events are typically sparse, unpredictable, and unevenly distributed. To address these challenges, we carefully curate the first event-centric marine video understanding dataset called MarineEVT, which features 20K multi-task, video-level visual question-answering pairs spanning multiple dimensions of marine understanding and analysis. Meanwhile, based on MarineEVT, we decompose marine video understanding as an Event-centric Visual Tool-integrated Reasoning process EVT-R1 for short, where we leverage powerful visual tools to drive the model to localize and interpret critical information aligned with visual questions and human intent. To demonstrate its effectiveness, we compare EVT-R1 against 11 SOTA VLMs in different settings. EVT-R1 outperforms the top open-source and top commercial models by 5.22 and 11.09, respectively. MarineEVT and EVT-R1 lay the foundation for ecological discovery and marine education, fostering the development of VLMs capable of interpreting marine dynamics, reasoning about ecological interactions, and supporting sustainable ocean video understanding and analysis.

Digitized musical heritage collections offer new opportunities to examine how stylistic traditions change over historical time, but computational analyses often reduce musical works to static classifications or similarity scores. This article proposes a representation-to-dynamics framework for studying cross-cultural harmonic change in Western art music. Starting from symbolic chord sequences, we derive contextual chord embeddings, project work-level representations into a shared harmonic space, and reconstruct country-level trajectories through temporally causal Kalman filtering. These trajectories are then modeled with DeGroot and Friedkin--Johnsen dynamics, yielding interpretable influence-like networks and estimates of stylistic anchoring. We apply the framework to a curated corpus of 480 dated works from 1875 to 1940 across Russia, France, Germany, Austria, and a heterogeneous ''Others'' group. Models fitted on 1875--1925 are evaluated through recursive forecasts for 1930--1940, testing whether the estimated dependency structure remains informative across a potentially changing historical and stylistic regime. The estimated trajectories and influence patterns are broadly consistent with established music-historical accounts of late Romantic and early modernist exchange, including the close relationship between German and Austrian traditions and historically plausible cross-currents between Russian and French traditions. PCA-variance-weighted estimation provides modest improvements while preserving a single interpretable influence network. Rather than treating the estimated matrices as direct causal evidence, the framework offers a reproducible quantitative layer for cultural heritage research, complementing archival and musicological interpretation.

This paper develops an online, off-policy policy-iteration framework for reinforcement learning (RL), based on sparse Gaussian-mixture-model Q-functions (S-GMM-QFs). The framework reconciles streaming, non-stationary data with the Riemannian structure of the parameter space while handling distributional mismatch through experience replay. S-GMM-QFs are introduced via Hadamard overparametrization, enabling interpretable sparsification through smooth regularization that facilitates Riemannian-based optimization. Overparametrization allows the framework to adaptively identify meaningful components from a large initial pool, yielding sparse models where interpretability emerges naturally from geometry: each component's parameters (means and covariances) explicitly encode its geometric role in the ambient state-action space. These geometric roles are learned through online gradient descent on a smooth objective over a (Cartesian-product) Riemannian manifold. Numerical tests demonstrate that S-GMM-QFs match or exceed deep RL methods while using substantially fewer parameters and achieving faster improvement per observed transition. Notably, parameter efficiency and interpretability combine to maintain strong generalization in low-parameter regimes where sparsified deep RL approaches degrade.

Dysphagia is a debilitating late effect of head and neck cancer (HNC) treatment, yet timely identification of at-risk patients remains challenging in survivorship care. Definitive assessment relies on videofluoroscopic imaging, as captured by the Dynamic Imaging Grade of Swallowing Toxicity (CTCAE-DIGEST), which, while validated, requires specialized equipment, trained personnel, and significant patient burden, limiting its routine use in surveillance. Patient-reported outcomes (PROs), by contrast, are low-cost, scalable, and easily collected at any clinical encounter, making them an attractive alternative signal for identifying patients who may warrant further evaluation. However, a clear clinical framework for translating PRO responses into actionable interventions is still evolving. In particular, uncertainty remains regarding when a patient's self-reported symptom burden should prompt escalation of care. This study addresses this gap by formulating a single-visit PRO-clinical prediction framework and introducing a clinically interpretable two-stage stacking model to predict swallowing impairment risk using PRO responses and structured clinical variables, without requiring videofluoroscopic imaging. The proposed framework quantifies the independent contributions of patient-reported symptoms and clinical factors within a unified and interpretable risk assessment model. Our findings demonstrate that individual MDADI responses contain predictive information beyond that captured by composite or global summary scores, while interpretability analyses reveal symptom patterns and clinical risk factors associated with swallowing impairment. Together, these results support the use of structured PRO-clinical integration as a practical, imaging-free approach for dysphagia risk stratification in HNC survivorship.

Industrial health management increasingly relies on heterogeneous information sources, including condition monitoring systems, supervisory control and data acquisition systems, maintenance records, inspection results, and prognostic models. Although large language models provide new opportunities for cross-source reasoning, industrial data and analytical outputs differ substantially in structure, temporal resolution, physical meaning, and reliability. Directly integrating such heterogeneous information into a monolithic model may reduce interpretability, traceability, and adaptability to equipment and data changes. This paper introduces Industrial Tokenization, a conceptual interface for transforming source-specific analytical outputs into structured and machine-interpretable units of industrial evidence, termed Industrial Tokens. Unlike numerical tokens used to encode raw time-series data, Industrial Tokens represent domain-grounded evidence together with source, temporal scope, operating context, analytical meaning, quality or confidence information, and provenance. Based on this concept, a federated industrial architecture is proposed, where heterogeneous analytical subsystems retain autonomy while exposing standardized Industrial Tokens to a central reasoning layer. As an initial implementation, this study presents an end-to-end DiagnosisToken pathway based on vibration-diagnostic outputs, rule-based event aggregation, structured textual token generation, and LLM-based interpretation. Other Industrial Tokens, including SCADA-based condition-monitoring tokens, maintenance tokens, and prognostic tokens, are reserved as future extensions. The proposed framework positions Industrial Tokenization as a semantic interface between domain-specific industrial intelligence and LLM- or agent-based reasoning, rather than another method for encoding raw industrial data.

As autonomous driving systems move toward real-world deployment, interpretable, behavior-level decision-making is essential for safety, trust, and regulation. We introduce CommandLM, a multimodal large language model that generates concise, human-readable behavior descriptions for ego vehicles from fused multi-sensor data. Our model processes temporally fused bird's-eye view representations from LiDAR and multi-camera inputs via a Q-Former adapter connected to a quantized, LoRA-fine-tuned large language model. Trained on our CommandLM-nuScenes dataset, CommandLM produces intent-aware, interpretable captions suitable for planner supervision and safety auditing. Experiments demonstrate strong linguistic and behavioral alignment, achieving CIDEr 0.67, and BERT-F1 0.88, substantially outperforming the BLIP-2 baseline (CIDEr 0.52, BERT-F1 0.86). In human evaluation, 58% of the generated descriptions were rated accurate, efficient and rule-compliant, confirming their real-world plausibility. While the remaining descriptions may not always select the most efficient, goal-oriented behavior, CommandLM's interpretable outputs enable downstream validation systems to identify and correct such cases, making it an effective tool for transparent behavior auditing. These results show that integrating multimodal fusion with language reasoning yields efficient and transparent behavior-level understanding for autonomous driving. We release our code and dataset at: https://github.com/b-tok/CommandLM

Few-shot class-incremental learning (FSCIL) aims to incrementally learn novel classes with only a few samples while avoiding forgetting base classes. However, current methods show a tendency to misclassify novel-class samples into base classes, which we find to be caused by the excessive focus on base-class-discriminative regions on novel-class samples. In this work, we aim to explore the underlying mechanism for an interpretation and solution. We first provide a compositional view to analyze the transferred and reused spatial patterns on novel-class samples. Then, through extensive experiments and theoretical analysis, we identify both empirically and theoretically that a shortcut exists in the model's base-class training, which naturally forms the excessive focus on only the most discriminative regions (primitives), which we term as the regional shortcut. Finally, based on this interpretation, to address this problem, we propose a compositional-learning-based method to learn two primitive sets (a common set and a discriminative set), which alleviates the regional shortcut by constraining the model to learn and utilize the common primitive set for base- and novel-class recognition. Extensive experiments on standard FSCIL benchmarks demonstrate the effectiveness of our approach, yielding consistent improvements over existing state-of-the-art methods in both accuracy and interpretability.

Neural network compression and interpretability remain open challenges in modern deep learn- ing, where billion-parameter architectures deliver impressive accuracy at the cost of trans- parency, computational efficiency, and reliable uncertainty quantification. This paper introduces Neural Atom Prevalence (NAP), a principled Bayesian framework for structured node-level model selection in feedforward neural networks. NAP introduces the neural atom (activation unit) and functions as a hybrid method operating through a four-phase pipeline: Bayesian Lottery Ticket (BLT) identification via Iterative Magnitude Pruning (IMP), soft variational training of the Spike and Slab Independent Gaussian (SS-IG) model, Poisson-Binomial (PB) optimal layer-size selection, and Bayesian fine-tuning to produce a sparse, stable, interpretable, and accurate model. Extensive empirical validation across simulated nonlinear regression, two UCI benchmark datasets (Concrete, YearPredictionMSD), and the MNIST image classification task demonstrates that NAP achieves state-of-the-art structural sparsity, reducing active nodes to as few as 8% of the original dense architecture on MNIST, while well-calibrated probabilisti- cally: the aleatoric-epistemic uncertainty decomposition reveals that model ignorance accounts for only 3 to 4% of total predictive variance across all experiments, and regression reliability diagrams confirm a near-nominal predictive interval coverage (93.4% observed against a 95% target). These results establish NAP as a reliable, theoretically grounded, and computation- ally tractable solution to the simultaneous pursuit of sparsity, accuracy, interpretability, and uncertainty quantification in Bayesian neural networks.

The electrocardiogram (ECG) is a cornerstone of cardiac as- sessment, yet clinical deployment of deep learning models remains con- strained by limited interpretability and the hallucination risk of large language models (LLMs). Existing CNN+Grad-CAM+multimodal LLM frameworks can generate ECG reports, but their explanations are often only weakly grounded in established diagnostic criteria, reducing trust- worthiness and reproducibility. We propose a guide-grounded multimodal framework that explicitly anchors report generation in curated clinical knowledge. A convolutional neural network (CNN) and Grad-CAM first produce class probabilities and class-specific heatmaps from 12-lead ECG images. In parallel, authoritative ECG textbooks and guideline materials are distilled offline into a structured ECG Interpretation Guide, which is injected as a fixed knowledge block for every sample. Conditioned on the ECG image, Grad-CAM overlay, CNN-derived fact pack, and the in- jected guide, a multimodal LLM generates structured diagnostic reports with guideline-consistent terminology and criteria usage. Experiments on the full PTB-XL test set demonstrate that guide grounding improves se- mantic quality and perceived consistency of generated reports while pre- serving competitive classification performance. In particular, our method increases the average BERTScore of generated impressions from 0.818 to 0.953 relative to a strong CNN+Grad-CAM+MLLM baseline, indicat- ing closer alignment with reference reports. These findings suggest that injecting a distilled interpretation guide into the multimodal prompting pipeline offers a practical pathway to reduce hallucinations and enhance the clinical plausibility of LLM-based ECG explanations, bringing ex- plainable cardiac diagnosis closer to real-world deployment.

Malware analysis demands rapid interpretation of complex detonation reports spanning filesystem, network, and process behaviours. While large language models (LLMs) demonstrate impressive capabilities for technical artifact interpretation, the opacity and escalating API costs of closed-weight frontier models motivate exploration of open-weight alternatives. However, many open-weight models are large, demanding significant compute resources and incurring non-trivial hosting costs that place them beyond reach for resource-constrained deployments. This paper investigates whether orchestrated ensembles of small language models (SLMs) can match or exceed single LLM performance on structured questions about malware detonation reports. We established baselines by testing eleven open-weight SLMs, three cyber security pre-trained models, and six frontier LLMs on Meta's CyberSecEval Malware Analysis benchmark. We then designed and evaluated four orchestration architectures: (i) a multi-agent pipeline that decomposes analysis into structured evidence-collection and reasoning stages, (ii) an adversarial debate framework in which two agents iteratively critique each other's reasoning, (iii) a hierarchical consultation system that pairs a general-purpose SLM with a cyber-specialised expert model, and (iv) a hybrid architecture that combines evidence-grounded pipelines with adversarial debate reasoning. The hybrid system (Qwen3-4B with Foundation-Sec-8B) achieved 35.30% overall accuracy, exceeding the strongest cyber-specialised baseline (22.54%) and the strongest ungrounded frontier baseline (34.77%); when given the same evidence pipeline, grounded Gemini remained the strongest configuration at 38.22%. These findings show that evidence-grounded orchestration can substantially improve the performance of collaborative SLMs for supporting interpretation of malware detonation reports.

Highly nonlinear chaotic dynamical systems remain difficult to model due to fundamental trade-offs between complexity, expressivity, and data efficiency. Modern machine learning methods achieve strong predictive performance but often rely on a-priori system knowledge or curated data with limited interpretability. Koopman operator theory offers a promising direction via linear representation in an infinite-dimensional observable space. However, many data-driven Koopman methods seek globally valid operators for which useful finite-dimensional spectral embeddings remain difficult to identify under these constraints. To overcome associated limitations, we introduce Fuzzy Spectral Region Decomposition (fSRD), a fully automated learning framework for estimating finite Koopman representation via multiple operators. The proposed method realizes a data-adaptive framework for assembling locally invariant embeddings, termed Invariant Decomposition. fSRD achieves highly accurate linear reconstructions of nonlinear systems while learning finite-dimensional representations of their induced evolution operators, bridging interpretable operator-theoretic models with expressive data-driven sequence learning. These embeddings are adaptively constructed via a global fuzzy tree model, drawing inspiration from fuzzy neural architectures to learn the induced dynamics while prioritizing parsimonious solutions. Empirical results across canonical chaotic systems (e.g., Lorenz and Duffing) and high-dimensional real-world data demonstrate strong predictive accuracy, interpretability, and robust expressivity across data-rich and data-limited regimes, highlighting the method's generality.

Approximately 3,000 of the 4,500 oracle bone script (OBS) characters remain undeciphered due to fragmentary inscriptions and sparse evidence. Current AI approaches fail to replicate expert workflows that integrate form analysis, contextual semantics, and philological reasoning. We introduce AlphaOracle, a human-workflow-inspired framework that systematizes OBS decipherment using the largest digitized corpus to date. Its multi-stage pipeline comprises: (i) rubbing parsing; (ii) radical-based morphological analysis with diachronic modeling; (iii) contextual retrieval with semantic alignment; and (iv) philological validation against classical sources. Each stage generates explicit, confidence-weighted evidence chains, culminating in interpretable reports for scholarly verification. Across multiple test characters, AlphaOracle's readings strongly agreed with expert interpretations. In a study of 86 domain specialists, it reduced analysis time by 64% and 79% of participants rated it highly useful. Notably, AlphaOracle resolves the character "Lao" as a toponymic or clan designation, offering concrete revisions to Shang administrative and social interpretations. These results suggest that computational methods aligned with philological practice can facilitate OBS research and provide a conceptual reference for studies of other undeciphered scripts.

A Bayes-filtered transformer (BFT) is a transformer trained on sequences that are generated in two steps: first a latent task is drawn from a prior, then observations are drawn conditional on that task. Trained under autoregressive log loss, the BFT's next-token prediction, in the idealized limit, is the Bayesian posterior predictive distribution (PPD) induced by that prior and that conditional law. In practice the trained BFT is only an approximation of this ideal PPD, raising an interpretive question: what prior and posterior over the latent task has the trained BFT actually internalized? Existing work answers this question by comparing the trained BFT's predictions against the predictions of various "reference" posteriors, each standing in for a different candidate algorithm or computation the BFT might be implementing. This prediction-space comparison is fragile: different posteriors can share the same posterior-mean predictions. We use predictive Monte Carlo (PMC) as a general interpretability tool for any BFT: using only next-token generation, PMC returns an approximation to the implicit prior and posterior over the latent task, answering the interpretive question directly in latent space. We apply PMC to three stylized task families spanning 0-Markov and 1-Markov exchangeability. The phenomena previously reported in these settings remain visible in latent space. Code is available at https://github.com/afiq-aswadi/bft-pmc

Autonomous rendezvous and proximity operations (RPO) in adversarial orbital environments require guidance architectures balancing target pursuit, safety preservation, and real-time adaptability under dynamically evolving interaction conditions. Although learning-based approaches show promise, their application to safety-critical orbital robotics remains limited by concerns regarding interpretability, robustness, and constraint awareness. This work presents an adaptive Model Predictive Control (MPC) framework for autonomous spacecraft RPO in multi-agent adversarial scenarios. The proposed architecture combines a constrained receding-horizon MPC formulation with a data-driven supervisory tuning layer that adjusts controller parameters from offline closed-loop evaluation and online interaction geometry. Relative motion follows Clohessy-Wiltshire (CW) dynamics, enabling computationally efficient finite-horizon prediction and real-time quadratic optimization. The MPC formulation incorporates actuator limits, predictive keep-out-zone constraints, slack-variable feasibility handling, and optional Control Barrier Function (CBF) safety filtering. Rather than generating thrust commands directly, the adaptive layer modifies interpretable MPC parameters, including tracking weights, safety penalties, minimum-separation objectives, and keep-out-zone objectives. The framework was evaluated in the official Kerbal Space Program Differential Game (KSPDG) Capture-the-Satellite environment through Monte Carlo simulations. Results demonstrate improved closed-loop robustness, adaptive maneuvering behavior, and rendezvous performance compared with fixed-parameter MPC while preserving safety-aware operation and real-time feasibility, providing a modular, interpretable foundation for adaptive spacecraft RPO.

AI governance increasingly requires judgments about whether an AI system remains adequately trustworthy over time, whether observed changes are tolerable, and how such judgments should be documented in a transparent and contestable way. Existing approaches remain either too high-level to support lifecycle monitoring and reassessment or too narrowly metric-driven to connect multidimensional trustworthiness evidence with governance decisions. We propose a lightweight methodology centered on \emph{trustworthiness level functions}: auditable rules that map measured trustworthiness profiles to governance-relevant levels. The methodology separates the underlying trustworthiness evidence from the governance rule used to interpret it and treats that rule as a lifecycle governance object. The rule may remain expert-defined or, when available evidence warrants empirical learning, be approximated by an interpretable candidate model. An AI lifecycle governance procedure embeds this choice in explicit decision gates for determining whether learning should be attempted and whether a learned candidate should become operative. The resulting rule supports lifecycle monitoring through level transitions, boundary margins, and profile drift, with explicit human responsibilities for validation, approval, and reassessment. We illustrate the methodology on synthetic AI lifecycle scenarios involving degradation, shocks, updates, heterogeneous monitoring cadences, and system comparison. Our methodology does not replace expert or legal judgment, but makes the governance interpretation of trustworthiness evidence more explicit, auditable, and contestable over time.

Multimodal large language models (MLLMs) are increasingly used to interpret visualizations, yet current evaluations remain largely chart-centric and provide limited evidence of understanding of scientific visualization (SciVis). We benchmark six MLLMs on the scientific visualization literacy assessment test, a standardized SciVis literacy assessment comprising 49 items based on 18 scientific visualizations and illustrations, spanning 8 techniques and 11 task types. We evaluate three closed-source and three open-source models under a closed-world protocol and compare their performance using data from 485 human participants. Results show that current MLLMs do not exhibit uniform SciVis literacy. Gemini is the strongest model overall, exceeding the human mean across the evaluated subsets, whereas the open-source models remain below the human baseline. Performance is highly uneven across techniques and tasks: models perform best on scientific illustration, search, and spatial understanding, but struggle on texture-based and integration-based visualizations and on quantitative estimation. Error analysis reveals recurring failures in fine-grained quantitative estimation, flow-direction interpretation, and grounded encoding interpretation. These findings position SciVis literacy as a necessary benchmark dimension for evaluating multimodal AI systems. Our code and model outputs are publicly available at https://github.com/patdmp/mllm-scivis-lit-benchmark.

In this paper, we study the connection between an LLM's output distribution and the data used to train it. Specifically, we study the degree to which an LLM's next-token distribution agrees with the empirical next-token distribution (ENTD) given the context in the training data. The ENTD is an appealing target because it is the unrestricted global minimizer of the next-token cross entropy loss used for pretraining, as well as an easily interpretable function of the pretraining corpus. We find that for a significant fraction of inputs, the LLM's distribution agrees with the ENTD almost perfectly, and the average agreement increases with model scale and training compute. Nevertheless, there is a long tail of input sequences where the LLM and ENTD differ significantly, and we examine several possible sources of this discrepancy across the transformer architecture, training procedure, and finite-sample noise in the ENTD estimate itself. More broadly, we hope our findings will encourage more work on ``data-centric mechanistic interpretability,'' a complement to standard mechanistic interpretability that opens the black box of how model behaviors arise from the data, rather than how they are encoded in the learned weights.

Music-driven dance generation aims to produce human motion that is both rhythmically synchronized and semantically consistent with music. While recent neural approaches have achieved impressive visual realism, they typically model motion as a continuous signal and neglect its compositional nature, making generated dances structurally incoherent and difficult to control. In this work, we introduce a structure-aware framework that models choreography as a sequence of atomic movements-semantically interpretable motion events that serve as the building blocks of dance. To construct this atomic movement vocabulary, we first segment large-scale dance data and cluster them into atomic movement groups. We then employ a large language model to semantically relabel and refine the clusters, yielding a set of interpretable and reusable atomic movements. Based on these atomic movement annotations, we design a two-stage generation framework that mirrors the human choreography process. In the atomic movement planning stage, the model predicts the type, duration, and timing of atomic movements conditioned on the input music, forming a symbolic dance allocation. In the completion stage, a transition-aware generator synthesizes smooth and stylistically coherent motion conditioned on the planned structure. Extensive experiments demonstrate that our method produces dances with significantly improved structural coherence, rhythmic alignment, and perceptual naturalness compared to existing baselines, while providing enhanced interpretability and controllable editing through explicit structural representation.

Contrastive Language-Image Pretraining (CLIP) representations form a semantic embedding space governed by cosine similarity, reflecting an intrinsic hyperspherical geometry. However, existing probabilistic interpretations typically rely on Gaussian assumptions, which fail to capture this directional and multimodal structure. We propose a principled density model for the CLIP latent space based on Mixtures of von Mises-Fisher (MovMF) distributions defined on the unit hypersphere. Using the Expectation-Maximization (EM) algorithm, we efficiently learn a probabilistic model in which each mixture component corresponds to a coherent semantic concept. This formulation yields a closed-form likelihood naturally aligned with hyperspherical geometry, enabling accurate and interpretable density estimation. Empirically, our model significantly improves long-tailed and out-of-distribution detection and provides a natural semantic decomposition, representing each embedding as a sparse probabilistic combination of interpretable concepts. These results suggest that CLIP latent space is more faithfully characterized as a hyperspherical semantic mixture rather than an isotropic Gaussian, establishing a simple and geometrically consistent probabilistic framework for modeling and understanding multimodal representations. Project page is available at https://xiaoyuzhizi.github.io/movmf-clip/.

Direct human interaction with autonomous UAV systems can be enabled through modalities such as speech, gestures, and graphical interfaces. However, interpreting such inputs as directly executable commands introduces safety risks in dynamic environments. Operator requests may conflict with terrain constraints, inter-UAV separation requirements, or flight-envelope limitations. In this paper, we present a requirements-governed maneuver-response model that mediates multi-modal human intent into safe UAV maneuvers by treating operator inputs as bounded maneuver requests rather than direct commands. Requested maneuvers are mapped to constrained motion primitives and processed through a structured request-evaluate-execute pipeline. Each request is interpreted with associated confidence, validated against terrain, separation, workspace, and flight-envelope constraints, and either constrained, rejected, or executed under continuous runtime monitoring. We further formalize the approach as a requirements-based specification model in which maneuver primitives are associated with explicit preconditions, invariants, guard conditions, and postconditions governing admissibility, execution safety, and emergency handling. These requirements support runtime verification and future reactive synthesis approaches. We present an initial lab-based validation demonstrating that voice and GUI-based inputs can be reliably interpreted and safely executed as constrained maneuver requests.

Active flow control involves nonlinear dynamics, partial observations, and computationally expensive simulations, making controller design particularly challenging. Deep reinforcement learning (DRL) has emerged as a powerful framework for such problems, but its success typically relies on large numbers of simulator interactions and produces neural-network policies whose decision process often remains difficult to interpret. In this work, we investigate a different paradigm: instead of optimizing neural-network parameters, we use modern coding agents to search directly for explicit executable feedback laws. We introduce a constrained heuristic-learning protocol in which an agent iteratively proposes, evaluates, and revises controller implementations while interacting exclusively through the public benchmark interface. The proposed framework is evaluated on 13 active flow-control benchmarks spanning one, two, and three-dimensional problems and compared against the strongest available DRL baselines under identical simulation budgets. The discovered heuristic controllers match or outperform the best DRL policy in 10 of the 13 environments while remaining compact, interpretable, and directly inspectable. Beyond aggregate performance, the resulting controllers reveal physically meaningful feedback mechanisms, transfer successfully across more challenging configurations, and remain competitive under varying Reynolds and Rayleigh numbers, actuator counts, and observation sparsity. These results suggest that heuristic learning through coding agents constitutes a credible and complementary alternative to conventional reinforcement learning, combining competitive performance with physically interpretable controller representations. Prompts and source code are available at https://github.com/DonsetPG/fluid-heuristic-learning.

Automatically retrieving videos from large camera-trap datasets remains challenging. Text-to-Video retrieval (TVR) methods based on large video-language models (VLMs) have potential to retrieve events of interest by describing them with simple text queries. However, current methods often lack spatiotemporal understanding and do not generalize well to ecological data. In this work, we introduce Prompting-MammAlps, the first camera-trap TVR benchmark, and propose a fine-grained and interpretable TVR method. Specifically, we trained a vision transformer to perform spatiotemporal action localization, and convert its output to structured text, describing each video. Independently, ethology-inspired queries are processed by a Large-Language Model (LLM) based coding agent to parse the structured text per video and retrieve videos accordingly. We harnessed the LLM to use functions from a custom parsing library to minimize the risk of LLM hallucinations and to improve method interpretability. This retrieval approach applied on the Prompting-MammAlps benchmark achieved a set-based F1-score of 34% on a test set of 135 ecologically-relevant queries and 775 candidate videos. In comparison the best zero-shot VLM achieved a F1-score of 18%, while also lacking interpretability. Project page: https://cnai.epfl.ch/prompting-mammalps

The current state of the art in AI/ML rests on deep neural architectures, which, in general, suffer from a lack of interpretability. Interpretability is crucial to gleaning insights while analyzing experimental data, where quantitative predictions may not be adequate for a scientist. We present a three layer neural architecture, SAMPAT (Smooth Approximation via Multivariate Polynomials and Analytic Transformations), that can provably learn a continuous, everywhere differentiable function, that can approximate any smooth function arbitrarily closely. SAMPAT's approximant can be expressed as a closed and compact algebraic, analytic expression, providing complete interpretability. Experiments on synthetic and benchmark datasets indicate that SAMPAT yields competitive performance with simpler representations. For many tasks, a two layer SAMPAT suffices. By imposing restrictions on the connectivity between neurons, SAMPAT may be used to provide a range of approximants, including regular and trigonometric polynomials, rational expressions, Gaussians, mixtures of Gaussians, as well as arbitrary combinations of the same; without restrictions, it learns a suitable structure. SAMPAT may be used to factorize polynomials and model nonlinear systems. With the addition of skip connections, a 4 to 6 layer SAMPAT is adequate to represent a substantive range of methods widely used in AI/ML, allowing the choice of a model's family, not just its parameters, to also be optimized as part of the learning process.

Integrating complex, multi-omics data presents significant challenges. Existing approaches often face a trade-off between model interpretability and representational capacity, with most either relying on post-hoc interpretation or use linear models that may overlook complex interactions. We report Pathway Activity Autoencoders for the multi-omics setting, which embed prior knowledge via pathway-informed architectural constraints, fostering interpretability, while preserving representational power. Our multi-omic framework is applied in the context of breast cancer and is evaluated in survival prediction and subtype classification with results indicating a positive effect of integration. We conduct analysis of individual omics layer impact on end-task performance, revealing that gene, protein, and microRNA expression layers provide the strongest contribution. Repeatability studies indicate that, while dropout improves model robustness and consistency, excessive regularisation can reduce predictive performance. Finally, visualizations of the learned feature space illustrate the framework's intrinsic transparency and clinical relevance. The results underscore the value of multi-omic integration and delineate the impact of individual omics layers, establishing practical guidelines for integration within our framework. Overall, our pathway activity autoencoder frameworks yield superior latent representations that are biologically meaningful and are directly translatable into clinically relevant insights.

We characterize observability in representation learning through Platonic Projection Structures (PPS), an operator-theoretic framework for analyzing representation accessibility under partial observation. Rather than treating observable outputs as direct reflections of latent representations, PPS models observation through a self-adjoint positive semidefinite operator acting on a latent representation space. A system is represented as a triple $(H, Π, O)$, where $H$ is a latent representation space, $Π\succeq 0$ is an observation operator, and $O(v)=\langle v,Πv\rangle$ defines an induced scalar observable. Observability is characterized by the quotient geometry $H/\ker(Π)$, representing equivalence classes of latent states indistinguishable under observation. We show that quantum measurement and representation inference under linear observation models share this operator-theoretic structure while differing in the algebraic properties of their observation operators; the correspondence is structural rather than physical. Representation transfer and knowledge distillation can likewise be interpreted as approximate preservation of observable geometry through $ΦΠ_T \approx Π_S Φ$. PPS also reveals a structural limitation of output-based interpretability: latent components in $\ker(Π)$ are inaccessible from induced observables, imposing intrinsic constraints on attribution and explanation methods. Controlled empirical validations demonstrate kernel-invariant observability, projection-induced attribution gaps, and rank-controlled observable geometry in latent representation spaces. PPS thus provides an explicit characterization of observability through operator-induced quotient geometry and a unified perspective on representation accessibility, interpretability, and projection-mediated inference.

Depression screening from large-scale behavioral data is challenged by fragmented circadian indicators, limited interpretability, and the lack of intervention-oriented analysis. Existing approaches typically analyze sleep, activity, and social behaviors in isolation, failing to capture their joint circadian structure. To address this limitation, we first propose the Circadian Rhythm Score (CRS), a composite index that compresses multi-domain daily behaviors into a unified representation of circadian rhythm. CRS is constructed to maximize discriminative power for depression screening while preserving behavioral semantics through non-negativity constraints. Empirical results demonstrate near-lossless compression, where a single CRS retains almost the full predictive capability compared with multiple raw behavioral indicators. Building upon CRS, we develop an interpretable depression screening framework based on gradient-boosted trees and SHAP analysis, revealing nonlinear and saturation-like associations between circadian rhythm and depression risk. Beyond risk prediction, we further integrate interaction modeling and counterfactual regression to estimate heterogeneous and dose-dependent behavioral effects, enabling intervention-oriented reasoning under different circadian contexts. Experiments on the China Health and Retirement Longitudinal Study (CHARLS, n=15,233), demonstrate robust screening performance (ROC-AUC=0.825) and identify actionable behavioral thresholds, including a minimum effective exercise dose of approximately 300 MET-min/week and an optimal restorative nap duration of approximately 65 minutes for sleep-deprived individuals. By bridging supervised representation learning and interpretable modeling, this work provides a scalable framework for depression screening and intervention-aware healthcare data mining.