Computer Science (arXiv)

Understanding how research areas emerge can help identify nascent areas early and inform research strategy, yet how the intellectual base of a field restructures as an area takes shape remains unclear. We hypothesize that the emergence of a research area is accompanied by the integration of largely separate knowledge communities, observable as a decline in the modularity of its co-citation network, which represents its intellectual base. We propose a framework that tracks this modularity over time, evaluates the statistical robustness of its changes, and identifies the papers highly associated with the decline. We applied it to three areas with different modes of growth: higher-order network science, superstring theory, and graph representation learning. In all three, modularity declined in correspondence with each area's emergence or transformation, and in superstring theory, the decline aligns with an independently documented transition. Further analysis of higher-order network science shows that its decline reflects a cross-disciplinary integration. In graph representation learning, the gradual decline is followed by a rise, which we interpret as a re-differentiation after the emergence period. Our results suggest that a decline in the modularity of a co-citation network can serve as a structural signature that retrospectively characterizes this integrative mode of emergence.

Agent-based social media simulators offer a controlled environment to study content moderation, yet their value hinges on how faithfully they reproduce real platform dynamics. We develop a calibrated extension of SimSoM, an agent-based model of information diffusion on social networks, grounded in a real-world dataset of online vaccine discourse during the COVID-19 pandemic. Our approach replaces ad-hoc parametrisations with empirically fitted distributions, optimised via CMA-ES (Covariance Matrix Adaptation Evolution Strategy) and validated against real data across temporal, distributional, and structural dimensions. Using this validated simulator, we provide three key contributions. First, we show that the calibrated model reproduces key statistical signatures of the empirical data, including activity distributions, post/reshare ratios, and temporal patterns. Second, we apply established misinformation-spreader detection and prevention methods to both empirical and simulated data, progressively removing top-ranked users and showing that the resulting decline in low-quality content is consistent across the two. Third, comparing static (retroactive) and dynamic (in-simulation) moderation across 30 network realisations, we show that static evaluation significantly overestimates the effectiveness of user bans for the most effective detectors: when moderation is applied in real time, compensatory resharing by the remaining users dampens the expected reduction in low-quality content, so static estimates should be read as an upper bound. These findings highlight the necessity of simulation-based evaluation for content moderation policies and contribute a reusable, empirically grounded simulation framework.

Aligning intraoperative biplanar digital subtraction angiography (DSA) to pre-procedural computed tomography angiography (CTA) requires rapid and accurate 3D-to-2D registration. Optimization-based methods are sensitive to initialization and may require hundreds of iterations, whereas learning-based approaches commonly rely on patient-specific training. We propose GeoPose, a population-trained framework that estimates the C-arm pose in a learned canonical frame and transfers it to the native frame of an unseen CTA through projection-space calibration and transform composition. A population-trained residual network refines the pose, followed optionally by low-budget image-driven optimization. GeoPose requires neither patient-specific adaptation nor explicit inter-volume preregistration. On 80 DSA observations from 20 held-out patients, optimization-free GeoPose achieved a carotid mean projected centerline distance (mPCD) of 5.8 mm and a clDice of 0.45, compared with 14.5 mm and 0.28 for the best-performing baseline, while requiring only 0.15 s. After 25 optimization iterations, GeoPose reached an mPCD of 4.6 mm and a clDice of 0.58 in approximately two seconds. Under the same budget, native-initialized optimization achieved 14.6 mm and 0.15, respectively. GeoPose thus provides rapid native-frame registration with fixed population-level weights and the geometric correspondence required for downstream biplanar 3D vascular reconstruction.

In a recent preprint, Simon proposed a polynomial-time quantum algorithm for the Dihedral Coset Problem and rested the analysis on four lemmas. Three of them carry only proof sketches, and this paper gives each of those three a statement that admits a single reading together with a complete proof. Lemma 1 follows from an exact second-moment computation for the subset-sum counts, and it holds with probability tending to one in place of the constant originally claimed. The amplitude bound of Lemma 3 follows from an exact Parseval identity on the cube of measurement outcomes and holds at every threshold with no well-behavedness hypothesis, so that predicate leaves the argument entirely. For Lemma 4, we compute both balls-in-bins covariances exactly and find that the second carries a term a fixed ball count leaves out. The assumption that the distinguished group contains no faulty samples can also be dropped. The two branch amplitudes share a signed prefactor, so the counting estimates control their difference and not the ratio the lemma states. We prove the additive form and show that the closing argument consumes nothing more than that. A single hypothesis survives all of this. It asks that the partition into the two sides be fixed independently of the measured string, and the rule the algorithm gives for choosing that partition does not supply it. Establishing these four lemmas therefore does not by itself establish the correctness of the algorithm.

Context: Developer efficiency is driven by technical, organizational, and personal factors, yet few longitudinal studies explore how these factors evolve over time. Objective: This study investigates the primary factors hindering the perceived efficiency of developers in a consulting and professional development context, analyzing how these factors vary across recurring data collection cycles and how they are described qualitatively. Method: We conducted a mixed-methods longitudinal case study applying the Adaptive Developer Efficiency Monitoring Method (ADEMM) to 27 external software developers, combining twelve waves of periodic surveys with eighteen semi-structured interviews, analyzed through statistical and thematic analysis. Results: The most frequent bottlenecks were organizational dependencies and waiting for external validation, which stayed structurally stable, followed by technical knowledge gaps, which declined as developers adapted. A generative AI usage barrier emerged qualitatively nine waves into the study, was incorporated into the survey instrument, and became the most frequently coded interview theme. Interviews corroborated the quantitative findings, with insufficient requirements documentation and organizational dependencies as the most recurrent themes alongside AI-related challenges. Conclusions: Perceived developer efficiency is highly dynamic and cannot be accurately captured through a single cross-sectional measurement. Adaptive monitoring via ADEMM identified an emerging factor, generative AI usage barriers, that a fixed instrument would have missed, and informed a concrete organizational intervention during the study. For organizations managing external developers, actions should target external dependencies, communication channels, and developers' evolving use of AI tools.

Cancer survival prediction supports treatment planning, risk stratification, and follow-up management. Existing methods use structured clinical variables, whole-slide images, genomic profiles, or multimodal inputs, while patient reports remain underexplored. We study report-centric survival prediction using reports that organize pathological, clinical, and molecular evidence. Large language models (LLMs) can reason over such reports, but case-wise time regression introduces two mismatches. First, a formulation mismatch arises because survival evaluation depends on ordering comparable patients, whereas independent time predictions do not enforce ranking consistency. Second, a supervision mismatch arises because a censored patient's observed time indicates survival beyond that point and cannot serve as an exact regression target, although it still implies orderings relative to patients who died earlier. To address these mismatches, we propose CACSurv, a Concordance-Aligned Comparative framework for report-centric survival prediction. CACSurv reformulates survival modeling as mini-cohort comparative reasoning, where an LLM predicts relative prognostic orderings. We introduce concordance-aligned rewards derived from comparable relations under right censoring, enabling censored outcomes to provide ranking supervision without exact event-time targets. At inference, Monte Carlo Reference Aggregation compares each patient with sampled references and aggregates positions into a cohort-level ranking. We establish TCGA-SurvReport, a benchmark covering six TCGA cancer cohorts. CACSurv achieves the highest C-index on all six cohorts and an average C-index of 0.722, outperforming the strongest published survival model by 6.5 percentage points and the strongest LLM time-regression baseline by 4.2 percentage points. Our code, models, and dataset will be available at https://github.com/xmed-lab/CACSurv.

Presentation-Attack Detection (PAD) for national ID cards is limited by the lack of publicly available genuine samples, making it difficult for systems to generalize across countries. This paper introduces two main innovations: (1) a Prototypical Network head using an EfficientNet-V2-b0 backbone that requires only four genuine samples per class to create reliable prototypes; and (2) an episodic training regime that keeps PAD classes fixed while varying the card domain, allowing the network to learn universal attack cues. Evaluated on a large multi-country dataset and the public DLC-2021 benchmark, this method achieves an average Equal Error Rate of around 9\%, outperforming conventional softmax and CLIP zero-shot baselines even with data from a single source country. This approach provides accurate, privacy-preserving PAD while minimizing data collection, facilitating scalable cross-jurisdictional remote onboarding.

Embodied agents are increasingly used to close the gap left by end-to-end policy models. Yet the agentic path has not realized closed-loop learning in physical execution: existing harnesses remain largely open-loop, following fixed skills during rollout and reflecting only after an episode completes. Such post-hoc reflection cannot govern execution as it unfolds, because physical interaction requires decisions to track rapidly changing robot-environment states at a frequency beyond today's large agentic models. We present Zetta, a closed-loop embodied harness that evolves code-based runtime critics and recovery skills online while keeping the base policy frozen. Through three timescale-separated loops, Zetta provides action-frequency governance, rollout-level critic-recovery proposal, and validation-gated skill updates. Together with Z-Infra, a rollout infrastructure decoupling agent logic from heterogeneous execution resources, Zetta achieves state-of-the-art success on LIBERO-Pro and RoboCasa under our current rollout budget, reaching 90.8% and 93.6%, with an 11.1x inference speedup; success continues to scale with self-exploration experience; learned skills transfer zero-shot, and clear robotic "Aha Moments" emerge. These results show that closed-loop harness self-evolution opens a scaling path for reliable physical intelligence.

Unsupervised Domain Adaptation for Adverse Weather Semantic Segmentation (UDA-ASS) aims to transfer semantic knowledge from labeled normal-weather images to unlabeled adverse environments. Existing approaches implicitly assume that restoration and segmentation provide mutually beneficial guidance. However, under severe degradation and without target-domain supervision, the validity of cross-task optimization directions becomes fundamentally unidentifiable, leading to hallucination-driven error propagation. In this work, we propose a novel Unsupervised Restoration-Segmentation Collaborative Learning Framework (Ultra), which reframes cross-task interaction as direction selection under uncertainty and causal effect estimation, enabling reliable collaboration through candidate direction generation and intervention-based filtering. In detail, we propose CTDN and CMIL. The former exploits complementary visual structures and semantic information to generate candidate optimization directions and performs cooperative direction selection between restoration and segmentation. The latter reformulates cross-task information transfer from correlation-based propagation into causal effect assessment, suppressing hallucination propagation. Extensive experiments on three widely used UDA-ASS benchmarks demonstrate state-of-the-art segmentation performance. Beyond segmentation, our framework achieves better unsupervised restoration results than existing UDA-ASS restoration methods and generalizes to unsupervised restoration and object detection collaboration tasks. Code and models will be available at https://github.com/Wang-Shiqin/Ultra.

Negative sampling is a key component of implicit collaborative filtering (CF), as it enables recommenders to effectively learn user preferences. Existing negative sampling methods mostly follow a two-stage paradigm: they first construct a candidate negative pool for each user and then select negative samples from the pool according to predefined sampling rules. However, these methods usually overlook the hardness variation of candidate negative pools across users, making it difficult to adaptively adjust the hardness and informativeness of negative samples according to candidate-pool conditions. In addition, most existing samplers evaluate candidate negatives mainly through a matching score computed from the final aggregated user and item embeddings, while ignoring the structural differences captured by multi-hop neighborhood aggregation. As a result, the training value of negatives may be insufficiently characterized. To address these issues, we propose SAHC-NS, a Structure-Aware and Hardness-Calibrated Negative Sampling method. Specifically, SAHC-NS uses the mean and standard deviation of layer-wise matching scores to capture the overall matching strength and cross-layer structural discrepancy of candidate negatives, respectively. This enables SAHC-NS to select informative negatives by taking cross-layer structural discrepancy into account, rather than relying solely on final matching scores. Moreover, SAHC-NS introduces a candidate-pool-aware hardness calibration module to dynamically adjust negative augmentation strength according to candidate-pool hardness, producing hardness-controllable negatives. Extensive experiments demonstrate the superiority of SAHC-NS over existing negative sampling methods.

Dense retrieval is commonly evaluated on benchmarks that represent each document with a single embedding, even though real-world retrieval systems often index long documents that require chunking. In these settings, the chosen chunking method not only affects retrieval quality, but also indexing throughput, query latency, and memory usage. Prior comparisons of chunking strategies have mainly focused on retrieval performance, leaving operational trade-offs underexplored. To address these issues, we evaluate eight representative chunking strategies across two scalable corpora, three embedding models, and multiple corpus sizes, measuring both retrieval effectiveness and system-level costs. Our results show that computationally expensive methods rarely provide consistent gains over simpler chunking. Instead, the best performing strategy depends on the embedding model, dataset, corpus size, and target retrieval metric. Methods with similar performance can also differ substantially in operational cost, showing that chunking should be seen as a multi-objective design decision.

Video Diffusion Transformers (DiTs) spend most of their compute inside the Self-Attention operation, whose cost grows quadratically, $\mathcal{O}(n^2)$, with the number of latent tokens $n$. For the task of video generation, the token count is large, so this term dominates runtime and memory, and thereby caps the resolution and duration we can generate. Linear $\mathcal{O}(n)$ and low-rank $\mathcal{O}(nk)$ surrogates of Self-Attention trade the full softmax $QK^T$ for cheaper kernels, but rarely recover the original's expressivity, leaving a stubborn quality gap. Motivated by this, we propose SQuad, a Sub-Quadratic Attention Distillation framework that achieves a complexity of $\mathcal{O}(n\sqrt{n})$ in the resulting distilled Attention, naturally balancing the efficiency v/s expressivity trade-off. Instead of training our own Video DiT from scratch, which is prohibitively expensive, we fit a pretrained full softmax Self-Attention DiT into our proposed SQuad-Attention one by distilling the former in two stages: Flow-Matching Supervised Fine-Tuning (SFT), followed by improved Distribution Matching Distillation (DMD2) which additionally makes the sampling more efficient. On the Wan~2.2 5B text-to-video model, SQuAD matches the quadratic teacher on VBench ($83.20$ v/s $83.08$) while cutting the per-step per-block attention FLOPs by $\sim$$67\times$ and attention latency by $\sim$$11\times$, and end-to-end DiT latency by 2$\times$, all while also generating a video in only $6$ Neural Functional Evaluations (NFEs) instead of the default $100$.

Software-defined networking (SDN) is a novel networking paradigm that enables network programmability and centralized control for network devices. The southbound application programming interface (API) is used to control and manage the underlying data plane devices. The existing southbound API relies on TLS with legacy cryptographic algorithms such as RSA and ECDSA. In this paper, we explore the performance of the southbound API with postquantum TLS (PQ-TLS) support. We present a proof-of-concept of using PQ-TLS in SDN's southbound API. We study the performance of pure and hybrid PQ-TLS modes across different security levels and compare them with legacy TLS in terms of latency and CPU utilization. We also compare the performance of different post-quantum signature and key establishment schemes.

Blueberry ripeness is judged by berry colour, cluster composition, and the distribution of maturity stages within a plant, however, public green house image resources with dense ripeness-stage masks remain limited. We present AerialYield-B2D, where B2D denotes BlueBerry Dataset, acurated real-image resource containing 514 RGB images and 30,195 annotated blueberry instances across five ripeness stages: green immature, pale pink, pink-turns-purple, fully ripe and over-ripe. The release provides class-specific binary masks, overall berry masks, semantic label maps, image-level count tables, SHA-256 hashes, source metadata, recommended train/validation/test splits and technical validations. AerialYield is the broader project name; this release does not provide harvest weight, fruit mass or per-area yield measurements, and the count labels should therefore be interpreted as image-level berry counts rather than yield estimates. The images include 424 smartphone greenhouse images, 67 video-derived frames, and 23 DJI Fly video-frame samples, providing a reproducible dataset for ripeness segmentation, berry counting, and class-imbalance analysis in controlled-environment blueberry production.

Context: Developer efficiency is influenced by technical, organizational, cognitive, and communication-related factors. However, most studies rely on one-time assessments or fixed instruments, limiting the ability to monitor how barriers emerge and change over time, especially in consulting and professional education contexts. Objective: This study proposes and evaluates the Adaptive Developer Efficiency Monitoring Method (ADEMM), an adaptive longitudinal method for monitoring developer efficiency when the monitoring organization does not directly employ the developers. Method: Following Design Science Research and Action Design Research, we conducted a mixed-method longitudinal study with 27 software developers over twelve survey cycles. ADEMM was designed and refined through five iterative cycles, combining recurring surveys, 18 semi-structured interviews, and joint evaluation with a problem owner. Results: The study resulted in ADEMM, a method that supports continuous data collection, mixed-methods integration, and iterative redesign of monitoring instruments. The evaluation produced three design principles: prioritization with the problem owner based on actionability, combination of closed and open data collection, and adaptation of items based on low variance and emerging qualitative signals. Conclusions: ADEMM provides a transferable approach for adaptive longitudinal monitoring of developer efficiency. It helps balance comparability, contextual sensitivity, and practical utility in environments where organizations need to support developers without directly controlling their work contexts.

AI agents increasingly operate as part of interacting systems rather than in isolation. As agents exchange information and jointly make decisions, their interactions can improve collective reasoning but may also produce herding, polarization, or amplify shared biases. Understanding and predicting these collective dynamics is therefore important for designing effective and aligned multi-agent systems. Here, we study over 10,000 communities of language-model agents that repeatedly exchange messages and revise their opinions across objective mathematics questions and subjective political statements. Despite substantial diversity in possible behavior, the individual and group dynamics can be represented by three characteristic regimes: indifference, polarization, and consensus. AI agents start indifferent and build conviction as they interact. On objective questions, communication improves collective accuracy, while on subjective questions it often drifts group opinions toward the right in the political spectrum. We explain these observations with a statistical-mechanics formalism in which agents stochastically favor lower social pressure. Given only initial opinions, our model predicts individual trajectories, outperforms all standard baselines, generalizes to unseen community graphs, and reproduces the observed group archetype distributions. Our fitted model parameters reveal the mechanics underlying our key observations: i) communities operate below the critical social temperature, which explains conviction buildup; ii) attractive ties outweigh repulsive ones, which favors consensus; and iii) agents holding the correct answer exert the strongest pull, which drives truth-seeking. Overall, our results demonstrate that collective behavior of AI agents, like that of other complex systems, follows compact and predictive dynamical laws.

Large language models (LLMs) are deployed globally in high-stakes settings, yet most safety research and alignment efforts remain concentrated on English. Thus, users interacting with LLMs in other languages may encounter weaker safeguards despite relying on the same systems for similarly sensitive tasks. In this work, we investigate whether safety signals learned from a high-resource language, like English, can improve multilingual safety. We propose BabelSteering, an activation steering method that acts as a lightweight inference- time intervention, using refusal directions derived from English safety supervision to generalize across languages. Our evaluation includes eight languages and jointly measures refusal of harmful requests, over-refusal, and general task utility. The results show that BabelSteering increases the refusal of harmful requests across languages, with only a marginal to no reduction in task utility but with some increase in refusal of pseudo-harmful prompts. For example, for Gemma 7B, we see an average increase in the refusal of harmful prompts across languages of 11 percentage points (pp), with individual languages like Bengali seeing an increase of 17 pp, with no loss of utility on Global MMLU, while pseudo-harmful refusals increase by 13 pp on average. We also introduce a multilingual translation-and-evaluation pipeline to facilitate future work on cross-lingual safety interventions. Overall, our findings suggest that activation steering may provide a practical, low- cost mechanism for extending English-derived safety signals to other languages. Warning: this paper contains examples with unsafe content

AI model lifecycles are commonly understood as a series of technical and organizational processes. Yet once a model enters sustained use, subsequent changes can also affect established user practices and user value. Using the Keep4o movement around GPT-4o as a case, this study examines post-deployment AI model lifecycle issues from the user side. We collected 61,846 public original posts on X from August 2025 to March 2026 and, using a systematically developed coding framework and LLM-assisted content analysis, analyzed discussion themes, users' reasons for wanting to keep GPT-4o, and the specific claims they made. Findings show that the Keep4o discussion extended well beyond continued access to the model itself. It covered concrete experiences of use, model behavioral characteristics and how they changed, and management issues across different stages of the model lifecycle. Reasons for keeping GPT-4o reflected interactional and relational value formed through long-term use, as well as judgments about the adequacy of replacement and the reasonableness of related decisions. The corresponding claims further reflected users' specific expectations for model lifecycle arrangements and governance. Overall, the call to "keep GPT-4o" brought together different judgments about user value and governance concerns. These findings suggest that technical version succession does not necessarily amount to effective replacement on the user side. Post-deployment AI model lifecycle management therefore needs to consider whether established user value can be carried forward and how model changes affect actual use. This study thus provides user-side empirical evidence for AI model lifecycle management. It further shows that user experience can provide important information for identifying post-deployment impacts and should be incorporated into lifecycle evaluation and decision-making.

Learning from demonstration is a promising approach for dexterous manipulation, but collecting high-quality contact-critical demonstrations remains difficult with low-cost teleoperation hardware. We present ViHaTeleop, a lightweight (0.7 kg), low-cost (\$550) visual-haptic teleoperation system with SLAM-based wrist tracking, camera-based hand tracking, and finger-wise vibrotactile feedback through Linear Resonant Actuators (LRA). The system includes several design choices (LED illumination, fisheye hand camera, and tactile-aware retargeting constraints) and is deployed on Franka + LEAP Hand + 9DTact in both real and simulated environments. Under matched with/without-haptic conditions with nine participants across six contact-critical tasks, haptics improved success rates across all tasks (+2.2 to +15.6 percentage points), while completion-time effects were task-dependent. Subjective ratings showed significant gains in contact clarity and grasp confidence in both simulation and real-world settings (Wilcoxon signed-rank, $p<0.05$). We also integrate a lightweight depth-camera-based tactile proxy in Isaac Sim, enabling a full pipeline from multi-modal demonstration collection to visual-tactile policy training. Preliminary downstream validation by training visual-tactile policies from collected demonstrations shows tactile cues benefit contact-critical subtasks (peg-in-hole: +17 percentage points over vision-only).

Functional dependencies are an important and well-studied class of database constraints that correspond to a notion expressed by dependence atoms in team logic. In practice, data often contain errors, so in some cases it might be useful to allow the database to have a small number of tuples that violate the desired dependency. Väänänen (2017) studied the axiomatisation of a notion of approximate dependence that specifies for each dependence atom how much of the database can be disregarded. We demonstrate that the interaction of approximate dependence atoms is more complicated than previously thought in the sense that there is a semantic consequence that is not captured by the inference rules introduced before. We show that Väänänen's axiomatisation is still complete in the restricted case of unary dependencies. We also consider the complexity of model checking for approximate dependence: it is NP-complete for disjunctions of two atoms and LOGSPACE-hard for individual atoms.

Accurate reconstruction of long-duration neural recordings is challenging because local field potentials (LFPs) are high-resolution, multichannel, transient, and variable across subjects. We present PCA-DMD, a scalable operator-theoretic framework that segments LFP recordings into overlapping windows, projects them into a compact PCA space, learns linear Koopman evolution in the latent space, and reconstructs continuous signals through inverse projection and overlap-add aggregation. On 200,000-sample hippocampal recordings, PCA-DMD outperformed Classical DMD, SpDMD, MrDMD, and HODMD, achieving KLD=0.0761 and HD=0.0847. In all-pair cross-subject zero-shot generalization at 300,000 samples, correlations were 0.9504-0.9800, with HD=0.0010-0.0072 and KLD=0.0005-0.0022, without target-subject fine-tuning. Out-of-sample temporal prediction showed close one-step agreement on temporally held-out LFP segments across the unseen interval and multiple channels. Scalability analysis from 400,000 to 900,000 samples showed stable zero-shot reconstruction, with mean correlation remaining about 0.965-0.968 while computational cost increased predictably. External validation on an independent 93-channel Allen Neuropixels recording yielded mean and median channel-wise correlations of 0.7427 and 0.7990, respectively. Koopman spectral and mode analyses revealed dominant eigenvalues concentrated near the unit circle. PCA-DMD therefore provides an interpretable, generalizable, and computationally scalable framework for reconstructing high-dimensional neural dynamics.

Neural audio watermarks are increasingly used to attribute and detect AI-generated speech, so their practical value rests on how cheaply an adversary can remove them. Robustness is usually measured by running a fixed battery of distortions blindly against every scheme. We instead make removal diagnostic: from a few clean/watermarked pairs we compute cheap structural probes that reveal where a watermark sits in the signal (its embedding domain), then apply a single domain-matched attack rather than a blind sweep. We further summarize each scheme with one threshold-free fragility score, the area under its accuracy-versus-quality trade-off, which an accuracy-only benchmark cannot provide. Across ten watermarking schemes the probes separate fragile from robust marks: for magnitude and carrier-domain watermarks a single matched attack erases the payload (WavMark, SilentCipher, audiowmark) or removes the detection flag (AudioSeal) at high objective quality (PESQ >= 3.6), whereas latent-domain marks (VoiceMark, WMCodec, AlignMark, AWARE) resist every training-free attack we apply. The same pair-only probe signatures also identify which watermarking scheme is present (84% over ten schemes).

This cumulative habilitation thesis studies probabilistic circuits (PCs) as a powerful and tractable framework for reasoning and learning under uncertainty in artificial intelligence (AI). It first advocates for probability as a core language for AI, emphasizing its connections to logic and information theory; the conceptual simplicity of probabilistic reasoning---based primarily on the sum and product rules; the parallels between probabilistic inference and human cognition; and the role of probability in optimal decision making. However, probability also faces significant computational challenges, as probabilistic inference is NP-hard in almost all probabilistic models. PCs address these challenges through structural constraints that ensure exact computation of a wide range of inference queries in polynomial time, such as marginals, conditionals, most probable explanations, expectations, and more advanced inference tasks. This thesis synthesizes a decade of research across foundations, algorithmic developments, and empirical validation of PCs. Key contributions highlighted in this work are foundational theory of PCs, Bayesian approaches for learning PCs, scalable implementations and integration with deep learning, hybrid models that combine PCs with intractable models, and connections with symbolic machine learning paradigms. This is the first part of my Habilitation Thesis. The second part is omitted, as it comprises the cumulative part of the thesis and has been published at various venues (see Chapter 5).

Machine learning (ML) is a key technology driving innovation today, but ensuring ML safety remains a major challenge for safety-related applications. A promising idea is to build proven-in-use arguments from field data, e.g. by running ML components (MLCs) in shadow mode or within safety envelopes so that their outputs can be monitored as 'safe probes' without affecting safety. These probes can then be used to build a statistical argument about field performance in a Bayesian way. However, many Bayesian field-data approaches in safety engineering model failures as a simple Bernoulli (or binomial) process with a single global failure probability and i.i.d. trials, which is rarely adequate for MLCs whose performance depends strongly on context. Statistical evidence is also about coverage of relevant situations, including edge cases, and building a single integrated statistical model for the entire system is usually not feasible. To address these challenges, this paper introduces CUBICS, a context-modular framework for per-component, situation-aware performance estimation of safety-relevant ML components. CUBICS partitions the operational design domain into situations and, for each safety-relevant component, defines a set of situation-specific assumptions and probabilistic guarantees that are represented and updated in a Bayesian manner using Subjective Logic (SL). By combining these guarantees with beliefs about how often each situation occurs, CUBICS derives an overall risk estimate for each component without requiring a monolithic system-level statistical model, and thus provides a building block for modular, field-data based safety assurance.

We find no evidence of critical scaling in the Schelling segregation model, in either the Moore neighborhood or its dense-spectrum extension to Chebyshev radii up to $r_0 = 6$ ($k = 168$ neighbors). On periodic grids up to $L = 320$ with 50 trials per point (> 12,500 runs), every finite-size scaling diagnostic in the Moore baseline fails: the per-$L$ $T_c$ does not drift, Var$(S) \sim L^{-2.02 \pm 0.09}$ matches trivial averaging, $γ/ν\approx 0$, and the scaling collapse never reaches a finite optimum. The 8-site Moore neighborhood restricts satisfaction to ratios $j/k$ with $k \leq 8$, giving $S(T)$ a staircase structure with 23 rational thresholds; discreteness alone does not forbid criticality (cf. the Ising model), but the scaling evidence rules it out empirically. A branching-ratio calculation predicts subcritical cascades of mean size $1/(1-R)$ and is validated by perturbation experiments to within 15%; the multiscalar dissimilarity length stays finite across the transition. The dense-spectrum extension strengthens the negative verdict: across $r_0 \in {3,4,5,6}$ on $L \in {40,80,160}$ the Binder cumulant has no $L$-curve crossing and the per-$L$ $T_c$ drift is monotonic and unsaturated; at $r_0 = 4$, extending to $L = 320$ gives $α= -2.70$, below the critical boundary $α= -2$, dissolving an apparent $α= +0.81$ signal visible only on $L \in {40,80}$. The mechanism is the absence of long-range correlation in equilibrium plus deterministic high-$k$ dynamics, not the staircase structure. With a Beta-distributed heterogeneous tolerance, the intolerant tail drives segregation even at moderate population-average tolerance. The staircase theorem and cascade mechanism together account for the Schelling transition without invoking critical phenomena.