Papers

Site-specific fertilizer recommendation systems adapt nutrient advice to location, soil properties, crop type, and production targets, but scientific reuse is constrained when recommendation functions remain accessible mainly through interactive interfaces, outputs are not versioned, and trained approximations cannot be independently loaded or benchmarked. This technical report presents the Turba fertilizer machine learning stack, a three-layer open-source implementation for reproducible site-specific fertilizer recommendation in Morocco. turba-client provides programmatic access to publicly accessible site profiles, crop-specific target-yield spaces, and N, P$_2$O$_5$, and K$_2$O recommendation workflows; turba-data distributes analysis-ready snapshots; and turba-models packages crop-specific machine learning surrogates of recommendation outputs. The architecture links upstream retrieval, versioned analytical snapshots, reproducible cross-model benchmarking, and loadable offline surrogates while preserving the distinction between recommendation-system outputs, observed agricultural data, and model-generated predictions. The first dataset was constructed from 44,096 unique ESA WorldCereal locations. Scenario expansion across supported cereal workflows generated 132,017 crop-location recommendation requests under a medium target-yield setting. The resulting 22-variable dataset spans 10 regions, 66 provinces, and 1,149 communes. Nine regression families were evaluated under a fixed deterministic 80/20 protocol, and the current release packages five best-performing crop-specific models. The machine learning task is recommendation-function emulation rather than prediction of observed crop response. The stack provides a reproducible basis for spatial and temporal validation, uncertainty estimation, field-trial comparison, and future integration with additional data.

We prove a weighted comparison principle for local plurisubharmonic potentials in Cegrell's class on an arbitrary complex manifold under suitable boundary and gluing conditions. We obtain domination and uniqueness when the manifold admits a global psh function in $\mathcal E_{\mathrm{loc}}$ whose Monge-Ampère measure has an absolutely continuous part with positive density almost everywhere. We apply these results to plurisubharmonic functions on domains in $\mathbb C^2$ whose singularities satisfy local inequalities involving $\alpha\log\|f\|$ and a Cegrell potential. For such functions, maximality is equivalent to the vanishing of the Monge-Ampère measure of the current remaining after subtraction of the Siu divisorial part. Consequently, maximality is a local property in this class.

InnerSource (IS) is increasingly viewed as a way to improve large-scale software development. However, evidence on its adoption in public sector organizations remains limited. This paper reports a study of IS adoption in a Swedish Public Sector Organization. Drawing on nine semi-structured interviews and an internal workshop with 13 participants, we examine motivations, challenges, and improvement strategies for IS adoption. The results show that the main motivations are improving reuse, strengthening cross-team collaboration, and enabling faster and more cost-efficient development. Key challenges relate to a lack of an open culture and mindset, management of reusable assets (e.g., ownership, documentation, and code visibility), and ways of working (e.g., siloed practices and limited time for contributions). The study also identified improvements to address challenges, including fostering an enabling culture and mindset for IS and knowledge sharing, strengthening management support, and establishing internal platforms and governance for reusable assets.

We investigate the relation between gravitational entropy and holographic entanglement entropy in braneworld gravity with gravitational dynamics localized on the brane. We first consider the DGP brane model, in which an Einstein-Hilbert term is included in the brane action, and derive the condition under which the gravitational entropy of a braneworld black hole agrees with the holographic entanglement entropy. We show that the two entropies coincide when a specific condition involving the extrinsic curvature of the brane is satisfied. In particular, this condition is automatically satisfied for static configurations. We then generalize the brane-localized Einstein-Hilbert term to higher-curvature gravitational theories. For Lovelock gravity and $f(R)$ gravity on the brane, we derive the corresponding conditions for the equality between gravitational entropy and holographic entanglement entropy and show that they take the same form as that obtained for Einstein gravity. Although the equality is not guaranteed for general stationary configurations, the required condition is satisfied for static braneworld black holes, leading to an agreement between the two entropies. These results clarify the relation between gravitational entropy and holographic entanglement entropy in braneworld gravity and contribute to a deeper understanding of holography in braneworld settings.

Ontology documentation, visualization, and validation are usually carried out with separate tools. This split workflow slows down development and makes knowledge transfer harder. We present OntoInk, an open-source MkDocs plugin that brings these activities together. Within a single documentation-as-code pipeline, OntoInk renders interactive ontology diagrams, validates instance data against SHACL shapes, runs OWL\,DL reasoning, and supports inline Turtle editing. General-purpose diagram plugins for MkDocs cannot parse RDF, dereference IRIs, overlay SHACL constraints, or run OWL reasoning. Compared with standalone ontology visualization tools, OntoInk embeds interactive and editable diagrams directly into documentation pages. A live demo and source code are available at \url{https://ise-fizkarlsruhe.github.io/ontoink/}.

Dysarthric speech reconstruction (DSR) typically relies on linguistic or phonetic representations extracted from impaired speech to generate a more intelligible waveform. We investigate a complementary approach that instead intervenes in a representation related to speech production. Starting from a neural analysis–synthesis framework, we introduce a residual mapper that modifies an articulatory-aligned latent space while preserving speaker and prosodic information. The mapper is pretrained on parallel synthetic healthy and artificially dysarthric speech, then adapted to natural dysarthric speech using phoneme-guided and adversarial objectives. We further compare the articulatory bottleneck with a dimension-matched unsupervised representation to assess the benefit of explicit articulatory supervision.

Agent skills package task-specific knowledge and procedures that can be composed to support complex agent tasks, while public marketplaces provide a growing pool of reusable skills. Existing security vetting, however, largely evaluates skills in isolation, leaving composition-induced risks underexplored. Such risks arise because composing benign skills expands the agent's capability space, enabling behaviors unavailable to any skill alone. Interestingly, we find that directly composing benign skills can already induce malicious behaviors, even when every individual skill passes security vetting. We further find that some target malicious behaviors remain difficult to realize through direct composition, even when the selected skills collectively provide the required capabilities. To systematically instantiate these attacks, we present Compositional Risk Induction via Multi skill Execution (CRIME). CRIME first uses the Malicious Plot Casting (MPC) module to decompose a target malicious behavior into complementary requirements and identify suitable benign skill compositions from public skill repositories. For compositions that cannot directly realize the target behavior, the Runaway Reaction Steering (RRS) module uses execution feedback to iteratively refine the selected skills toward the target while requiring each skill to remain benign under standalone vetting. The resulting composition is then passed to the Skill Reaction Chamber (SRC) module, where the skill pair is executed in a sandbox and the resulting environmental consequences are examined to determine whether the target behavior has occurred. Unsuccessful cases are returned to RRS for further refinement. Furthermore, we construct a benchmark of 4,000 public skills across eight cybersecurity behaviors for systematic evaluation of composition-induced vulnerabilities.

Business Process Management (BPM) was built on a foundational assumption that organizations are populated primarily by human actors whose work can be made visible, governable, and improvable through process models. That assumption is depreciating. AI agent ecosystems increasingly execute, coordinate, and adapt organizational work with limited human direction, challenging not only BPM's methods but its core conception of what a process is. We argue that BPM faces a constitutive shift from modeling human work to governing autonomous agents, for which we propose two new concepts: the Process Constitution, a machine-interpretable, value-laden framework that defines the space of admissible agent behavior, and the Process Steward, a governance agent that interprets and enforces it. The central value proposition of this new generation of BPM is not efficiency but organizational legibility: the capacity to keep agentic organizations accountable, contestable, and humanly understandable. We outline what this means and sketch the research agenda it opens.

Superionic conductors enable exceptionally fast ion transport in solids and offer broad technological potential. However, their highly conductive states are typically accessed only above the order-disorder transition of the mobile-ion sublattice, which limits low-temperature operation. Here, electric-field-responsive machine-learning molecular dynamics reveals a nonequilibrium route to superionic transport in quasi-two-dimensional AgCrSe2 driven by an out-of-plane oscillating electric field. At 300 K, far below the transition temperature near 475 K, the field activates fast in-plane Ag+ transport with an ionic conductivity reaching ~1.7 S/cm. This originates from field-driven dynamic disorder of Ag+ across two equivalent sublattices rather than thermally induced disordering. With increasing frequency, the Ag occupation evolves from complete switching to a highly conductive dynamically disordered state and finally to an ordered state with weak field response. These regimes shift to higher frequencies with increasing temperature or field amplitude, consistent with competition between the driving period and Ag-sublattice response time. At strong fields, this timescale is consistent with field-assisted thermal activation, while weak-field deviations suggest additional dynamics beyond single-ion activation. These results suggest timescale-matched periodic driving as a route to transport-active ionic disorder far below an equilibrium superionic transition.

Despite their impressive performance on a wide range of video understanding tasks, current Vision Language Models (VLMs) are predominantly designed for offline scenarios and struggle to handle online streaming videos that demand low latency response. Several studies have explored memory and token compression strategies in an attempt to adapt offline VLMs for streaming video understanding tasks. However, through our probing experiment, we identify that most existing works tend to progressively lose long context information as length of input stream increases. To address this, we propose ReMem, a novel training-free adaptation technique that enables VLMs to process streaming videos of arbitrary lengths while improving their long context information retention capability. ReMem exploits memory from two perspectives, implemented as two core components. The Streaming Context Memory (SCM) continuously compresses historical context with query-independent attention. The Retrieved Vision Memory (RVM) then retrieves the most salient, query-relevant context from memory to augment the VLM's input. Comprehensive experiments demonstrate that the proposed ReMem achieves state-of-the-art (SOTA) performance across a variety of widely used benchmarks, spanning both streaming video and general long video understanding tasks.

This paper establishes the first general structural solution to the problem of state estimation for nonlinear systems driven by unknown inputs. Building upon nonlinear unknown-input observability theory, we show that every such system admits a structurally equivalent representation, referred to as the UID-induced normal form. The proposed representation decomposes the information carried by the unknown inputs into two complementary components: unknown-input directions that are structurally decoupled from the observable dynamics and observable quantities that completely represent the unknown-input information affecting the observable dynamics. As a consequence, the UID-induced normal form provides a unified structural solution to unknown-input decoupling and unknown-input reconstruction, without requiring any model or stochastic assumption on the unknown inputs. The practical significance of the proposed framework is demonstrated through a previously unexplored minimal Structure-from-Motion configuration. The proposed representation enables recursive state estimation from only three point features and a single-axis gyroscope, allowing the recovery of the three-dimensional structure and camera motion up to an unknown global scale factor. Experiments on real-world data validate the proposed framework and demonstrate the feasibility of this minimal sensing configuration.

Let $A$ be the (non-symmetric) adjacency matrix of a uniformly random $d$-regular directed graph on $n$ vertices, and let $\Xi$ be independent of $A$ with i.i.d. Rademacher entries. Suppose that $\min(d,n-d)\ge\lambda n$ for some fixed $\lambda\in(0,1/2]$. We show that there exists $c > 0$, depending only on $\lambda$, such that \[ \mathbf P_{A,\Xi}\{\operatorname{rank}(A\circ\Xi)\le n-k\}\le e^{-c nk},\qquad 1\le k\le n. \] As an ingredient in the proof of the rank bound, we use the case $\kappa=0$ of the following quantitative smallest singular value estimate: \[ \mathbf P_{A,\Xi}\{s_n(A\circ\Xi)\le\kappa\} \le C\kappa\sqrt n+e^{-c' n}, \qquad \kappa\ge0, \] where $C,c'>0$ depend only on $\lambda$.

We prove that a locally maximal plurisubharmonic function on a domain in $\bC^n$, $n\geq2$, is maximal if it has locally analytic singularities with a two-sided locally bounded remainder. Using McAdam's theorem, we show that local maximality forces the analytic spread of each defining ideal to be less than $n$. We use normalized blow-ups and Siu's analyticity theorem to compare singularities through generic divisorial Lelong numbers. We also obtain a criterion for maximality in terms of maximality outside the pole set and the analytic spread of the local defining ideals.

One of the major transformative factors in 6G will be the integration of Artificial Intelligence (AI) to become a native part of Radio Access Network (RAN). While most physical-layer AI features have so far been evaluated in isolation using link-level simulations, their combined behavior in a realistic multi-cell, multi-UE deployment has remained largely unexplored. In this paper, we present system-level performance results when multiple uplink AI features are enabled together, achieved by integrating accurate link-level and system-level simulators. To infer state-of-the-art deep-learning-aided Multiple Input Multiple Output (MIMO) receivers under the dynamic allocations produced by a realistic uplink scheduler, we propose a mirrored data augmentation method that decouples receiver performance from scheduled allocation size. In addition to these Physical Layer (PHY) receiver features, we combine several recent advances in deep reinforcement learning to train uplink power control and link adaptation that outperform a heuristic baseline and further boost the gains obtainable from the AI receiver alone. The system-level results show that the combined AI features improve the mean uplink user throughput by roughly 27% compared to a non-AI baseline, confirming that the individual PHY and Medium Access Control (MAC) AI features provide complementary gains when deployed jointly.

Language models are moving beyond generating answers to pursuing long-horizon goals in interactive environments. Post-training these agents requires long, heterogeneous trajectories, and synchronous systems leave learner engines idle until rollout and verification finish. To squeeze out these pipeline bubbles, asynchronous training overlaps rollout and learning across updates, but comes at the cost of policy staleness. We introduce ThunderSyncRL, which starts gradient computation as soon as all required inputs are fixed, without policy staleness. For group relative policy optimization (GRPO), ThunderSyncRL computes each trajectory's score gradient as soon as the reward for that trajectory arrives, without waiting for the group. For on-policy distillation (OPD), it computes gradients for each completed agentic turn's teacher-scored actions while tool calls run in the sandbox. We prove that gradient streaming produces the same GRPO and OPD updates as batch-synchronous training, without changing either objective. On SWE-bench Verified and Terminal Bench 4.0, we train models to the same performance up to $1.9 \times$ faster than synchronous training. With zero policy staleness, ThunderSyncRL also outperforms asynchronous training at a fixed budget by up to $2.47$ percentage points.

We show that trap rotation overrides Kibble-Zurek (KZ) universality when a 2D Bose-Einstein condensate forms via a finite-time quench. In stochastic Gross-Pitaevskii simulations, the rotation-induced chiral bias drives the vortex number close to the Feynman-Onsager (FO) value, independent of the quench time. Ramping up rotation with the quench restores the KZ scaling at low frequencies, with vortex numbers below the FO value. The vortices form a disordered lattice that a pinning potential can order into an Abrikosov lattice, even at low rotation. Trap rotation thus offers a control knob for critical dynamics.

Let $(M^n,g)$ be a complete connected noncompact smooth Riemannian manifold without boundary, with $n\ge3$ and positive scalar curvature. We prove that $M$ admits a complete smooth metric of uniformly positive scalar curvature if $$ \liminf_{d_g(o,x)\to\infty}d_g(o,x)^2R_g(x)>\frac{n-1}{n} $$ for a fixed point $o\in M$. This extends earlier existence results for orientable manifolds, due to Chen in dimension three and to the authors in dimensions $4\le n\le7$, to every dimension $n\ge3$ without an orientability assumption.

Visual voice cloning requires intelligible, speaker-consistent speech synchronized with visible articulation. However, sequential multimodal conditioning can disrupt previously established temporal and speaker cues, while imbalanced inference guidance can improve linguistic accuracy at the expense of lip synchronization. In this paper, we propose UltraDub, a Unifying Visually-Steered Flow learning and trajectory Guidance Dubbing framework that leverages vision in two ways: as continuous motion for multimodal context aggregation, and as structural rhythm for trajectory rectification. Specifically, we introduce the Motion-guided Dual-context Retrieving (MDR) module, which continually recalibrates linguistic and speaker-style retrieval through shared lip-motion query residuals, utilizing independent time-conditioned gates to regulate their contributions. Furthermore, we propose Rhythm-anchored Trajectory Guidance (RTG), a training-free mechanism that evaluates hierarchical multimodal corrections at a visual-only predictive midpoint, safely strengthening semantic conditioning while better preserving temporal alignment. Finally, we construct DiverseDub, a multi-scenario benchmark to evaluate video dubbing in the wild. Extensive experiments demonstrate that UltraDub achieves state-of-the-art performance across four datasets.

For every integer $r\geq9$, we construct a smooth Griffiths-positive Hermitian metric on $\mathcal O_{\mathbb P^r}(1)^{\oplus r}$ whose top Chern form is pointwise negative on a nonempty open set. This gives counterexamples on compact projective manifolds to Griffiths' conjecture on the positivity of Chern–Weil forms.

We consider monoenergetic particles propagating in three spatial dimensions and undergoing elastic scattering off an external medium. For an arbitrary rotationally invariant scattering kernel, we derive an exact expression for the frequency- and wave-number-dependent diffusivity, thereby resumming the hydrodynamic gradient expansion to all orders. The result is an infinite continued fraction, with each successive level determined by the relaxation rate of a higher angular harmonic of the collision operator. Thus, while ordinary diffusion probes only the first angular harmonic, the resummed constitutive relation retains information about the full scattering kernel. We show how this microscopic information manifests itself in the analytic structure of the diffusivity and in the response to a localized external source. In particular, deviations from Fick's law decay exponentially with distance, at a rate determined by the nearest complex-wavenumber singularity of the inverse diffusivity. Thus, exponentially small departures from hydrodynamics retain detailed information about the underlying scattering process.

We consider a family of non-local and non-convex functionals depending on three real parameters $(\gamma,p,\lambda)$, which was introduced by H. Brezis, A. Seeger, J. Van Schaftingen and P. L. Yung. In the case $\gamma<-p$ we prove a lower bound for the Gamma-liminf of these functionals as $\lambda\to 0^+$ and a Poincaré-type inequality. We also show that such inequality cannot hold if $\gamma>0$, and we establish a modified version in this case. These results answer some open questions posed by these authors and by H. M. Nguyen.

Quantum error correction (QEC) relies on repeated parity extraction, mid-circuit measurement (MCM), reset, feed-forward, and scheduling, yet these primitives are usually assessed either in isolation or through resource-demanding experiments. We introduce a few-sample benchmark of QEC-relevant primitives based on an MCM implementation of the quantum approximate optimization algorithm (QAOA) with a fixed set of linear parameters (LR-QAOA). For a chosen code, the QAOA Hamiltonian is constructed from its check structure, such that the resulting circuit mimics the syndrome-extraction connectivity and MCM pattern while producing a direct algorithmic signal, the approximation-ratio r. The LR-QAOA depth, defined by the number of QAOA layers, plays a role analogous to the number of repeated syndrome-extraction rounds in a QEC experiment. From QPU executions, the decay of r with depth defines an effective hardware error, which we map to an equivalent two-qubit depolarizing rate {\lambda}eff. We compare direct and MCM-mediated implementations across 10 QPUs from IBM, IQM, and Quantinuum, with circuits containing up to 2950 MCM operations. On Quantinuum's Helios-1 and H2-1, we run code-structured LR-QAOA benchmarks for surface-code, triangular color-code, and bivariate-bicycle qLDPC Hamiltonians up to 81, 91, and 48 data qubits, respectively, using up to 480 MCM operations. On IBM ibm_phoenix, we implement the surface-code structure and compare the LR-QAOA response with logical-memory experiments, observing a correlation between the benchmark and the logical performance across different regions of the QPU. Its construction and low resource requirements provide a practical benchmark for comparing hardware generations and QEC implementations before full logical-memory experiments are performed.

demonstrate a frequency-tunable, low-phase-noise terahertz (THz) synthesizer based on a fiber-interferometer-referenced optical frequency comb (OFC), an erbium-doped fiber amplifier (EDFA)-integrated Vernier cavity, and dual optical injection locking (OIL). The Vernier cavity converts the densely spaced modes of a fiber OFC with a repetition frequency of approximately 100 MHz into a filtered comb with gigahertz-scale effective mode spacing, facilitating selective single-mode OIL. Intracavity and extracavity EDFAs in the Vernier cavity enable a side-mode suppression ratio exceeding 70 dB. Independent control of the Vernier-cavity free spectral range and the OFC repetition frequency provides flexible Vernier filtering conditions for selecting master modes for dual OIL. Two selected comb modes are transferred to injection-locked optical carriers and photomixed in a uni-traveling-carrier photodiode to generate continuous-wave THz signals. Frequency-tunable THz generation is demonstrated at 230, 267, and 320 GHz. At 267 GHz, a signal-to-noise ratio exceeding 70 dB and a measured phase noise of -95.7 dBc/Hz at a 10-kHz offset are obtained; the phase-noise measurement is limited by the local-oscillator system. The proposed architecture provides a fiber-based approach to frequency-tunable, low-phase-noise THz synthesis by combining an optical-reference-based OFC with flexible Vernier-assisted comb-mode selection.

The use of run-off triangles (ROTs) is a common industry practice in estimating the loss given default (LGD) risk parameter when predicting credit losses in banking. We benchmark this industry practice using credit card data against a more sophisticated (though classical) regression-based approach, which is able to leverage various types of input variables in producing loan-level LGD-estimates. This regression-based approach can demonstrably recover the typical characteristics of the 'U-shaped' empirical LGD-distribution, which the ROT-based approach cannot do. First, we critically review the ROT-based approach and identify multiple demerits using data-driven diagnostics. We then estimate a two-stage regression-based LGD-model and favourably assess the model performance of each component (or 'stage'). Finally, we aggregate the LGD-estimates produced by each approach over time, and compare each time series to the mean empirical loss rate over time. The ROT-based aggregates diverge substantially from the empirical rate over most time periods, whilst the regression-based aggregates follow the empirical trends much closer. These results underscore the greater prediction accuracy of the regression-based LGD-model, relative to the ROT-based one. By implication, the former approach is probably better than the latter ROT-based approach when estimating the LGD under the IFRS 9 accounting framework, which prioritises accuracy.

Entropy-based selection objectives suffer from a fundamental degeneracy: minimizing Shannon entropy $H(p_A)$ rewards confident selection regardless of whether the selected candidate is informative. We address this limitation with the residual entropy $D = H(p_A) - H(p_\beta)$, where $p_\beta$ is induced by candidate trust weights. We prove the exact identity $D = -\mathrm{KL}(p_A\Vert p_\beta) - \Delta$, where $\Delta$ measures whether the score-induced distribution and trust profile favor the same candidates. Boundary cases establish basic safety: under uniform trust, $D\leq0$ automatically, so an equal-trust, non-starving state is never penalized, while at any one-hot limit, $D\to0$ regardless of the selected candidate. For the intermediate regime where selection occurs, we prove that $D\leq0$ when candidate ordering by trust agrees pairwise with ordering by informativeness, and derive a tighter certificate based on the leading candidate's margin over its competitors. These results are independent of the candidate-scoring function and apply to both stationary and dynamically changing information. Experiments with a gradient-based mixture-of-experts router confirm that the ordering conditions can hold during real optimization and show that correct ordering improves downstream performance when candidates are non-interchangeable and selections are used directly rather than averaged. Beyond routing, margin-based reweighting matches or outperforms fixed-strength baselines in a class-imbalance task, while informative selection in a production video-prediction system reduces MSE by approximately 20$\%$ and transfers to a related species. Residual entropy, therefore, provides a safety criterion for selection and a usable signal for deciding when that selection is informative.

In the general probabilistic model of information theory, the signaling dimension of a given communication protocol (or, equivalently, of its state space) is the smallest possible complexity of a classical communication protocol capable of perfectly simulating the given one in terms of input-output correlations. The signaling dimensions of three-, four-, and five-dimensional cross-polytopes have been known to be 3. We give logarithmic lower and upper bounds for the signaling dimensions of higher dimensional cross-polytopes, pinning down the value up to an additive gap of 2. The lower bound shows that the signaling dimension goes to infinity, in contrast to the case of balls and hypercubes.

Competent agents need precise and verifiable environments, such as sandboxes that are resumable at any stage and evolve from observable evidence. However, most long-horizon work exposes how rare these are: for example, an agent in medical research must ground a finding, classify it, and write a report over dozens of dependent steps, yet recent environments score only the outcome. To address the challenges in stable training, we present VERA, which builds such environments at scale and lets agents evolve on them. VERA builds these environments from initial trajectories: an agent writes rubrics, executable checks, a judge verifies each sandbox, and only those that pass enter the training bank. On these environments, VERA alternates between two updates: train the model with rubric rewards, or edit the harness skills. We also create a verifier which gates model checkpoints and harness edits using explicit development-set acceptance criteria. This attribution distinguishes VERA's co-evolution from single-axis baselines: its updates target not only the cause but the outcome. With an open-source corpus of 9,000+ long-horizon verifiable environments, a 9B model paired with its co-evolved agent beats the strongest baseline by 10.3 and 13.0 points in the two domains. At 27B, it surpasses the baseline on AutoCoWorkBench (71.6) and AutoMedBench (80.7), transfers to unseen workflows, and retains general capabilities.

Evaluating the causal effect of new features is a central goal for online platforms. While recent literature addresses limited testing traffic via centralized portfolio optimization, this perspective abstracts away a critical institutional reality: experimentation is operationally decentralized. The experimenters who develop new features also dictate which hypotheses to test, and they are typically rewarded based on empirical average treatment effects that are prone to upward bias. Left unchecked, this principal-agent conflict can severely erode platform value, a structural failure that conventional centralized levers, such as significance thresholds and traffic budgets, cannot resolve. By reframing experimentation as an incentive design problem, we demonstrate that two practical mechanisms, sample splitting and shrinkage, can effectively bridge this gap. Sample splitting aligns incentives perfectly at a bounded traffic cost, while shrinkage consumes no additional traffic and guarantees that interventions with negative expected effects are strictly unprofitable to field.

In generative models, Optimal Transport (OT) is used to improve Flow Matching (FM) by reducing noise-data coupling cost. However, different noise-to-output assignments can yield nearly equal costs, raising a key question. Is cost alone sufficient to guide coupling design? We address this question by separating transport cost from routing, i.e., the destination reached by each noise sample. We show numerically how FM and OT can differ in routing while remaining close in cost. We examine its consequences in learned neural FM. Using the exact FM routing as an oracle, we further construct a routing-aware training coupling and find that it yields a directionally consistent improvement in generation over a cost-matched, cost-only counterpart. Our findings highlight what cost minimization can overlook and motivate using both cost and routing to evaluate the design of OT-based FM couplings. Code will be released upon acceptance.

Focusing on scalar conservation laws, a TVD theory is developed for the fully discrete, single-stage Active Flux method. The classical choice of degrees of freedom is considered, yielding a third-order method. Starting from an adaptation of the concept of (discrete) total variation to Active Flux, a corresponding adaptation of Harten's lemma is achieved. Finally, the high-order method is endowed with parameters chosen such that they can be efficiently computed to fulfill the conditions of the lemma.