High Energy Physics (Theory)
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Thursday 5 November
We study the unitary Fermi gas in an external trap in the double-scaling limit of large flavor number N and large charge Q at fixed Q/N. At leading order, the problem reduces to the spectral zeta function of a BdG operator in an inhomogeneous pairing background. The grand potential admits a gradient expansion with distinct bulk and edge sectors. We show that when the trap is harmonic, the bulk sector expansion terminates: beyond the first three terms every bulk contribution integrates to zero. This cancellation is general and follows from zeros of the Euler beta function computed in the Son-Wingate effective field theory, independent of the low-energy constants. The truncation persists for anisotropic traps in any odd dimension. All higher-order dependence for harmonic potentials is therefore carried by the droplet edge. For anharmonic traps, the asymptotic bulk series is factorially divergent. Using a heat-kernel WKB analysis, we show that its large-order growth, sign oscillations, and optimal truncation are fixed by complex periodic orbits of the BdG problem, which are analytic continuations of Andreev reflections. We numerically confirm the predicted Borel singularities for quartic and sextic traps.
High Energy Physics (Phenomenology)
Observing Transverse Momentum Broadening from TEECs in Heavy Ion Collisions
We study the azimuthal dependence of transverse energy-energy correlators (TEECs) in dijet events in heavy-ion collisions. In the back-to-back region this observable resolves the energy flow around $\phi=\pi$ and is sensitive to medium-induced momentum broadening and the response of the medium. To isolate final-state effects, we consider a quark-antiquark pair produced in a colour-singlet hard process and propagating through a homogeneous background medium. Starting from the vacuum TMD factorization formula, we combine NLL$'$ resummation with medium modifications at first order in the strong coupling. For the parameters studied, (classical) momentum broadening lowers the back-to-back peak and leaves the distribution elsewhere nearly unchanged, while radiative corrections significantly suppress the peak further and enhance larger angular decorrelations. We additionally include the medium response sourced by each jet, as a mean energy flow. This spreads the deposited energy over the whole azimuth, so that the contribution to the TEEC varies slowly across the back-to-back region, is small at the peak and grows toward larger angles.
Read the full abstract
Computer Vision
HaNDF: Object-Conditioned Geometric Neural Hand Distance Fields
We propose, HaNDF, object-conditional neural distance fields as data-driven priors for modeling the plausible hand-object configurations, in the product manifold of hand articulations. While recent unconditional generative priors showed success in modeling human pose, hands are versatile articulated bodies often found in interaction with other objects. This makes it crucial to develop a generative prior which can model the plausible hand poses conditioned on the object under interaction. To this end, our HaNDF proposes a geometry-aware, object-conditioned neural distance field, representing the plausible articulations in the zero-level set of a conditional field. We leverage tools from Riemannian geometry to (i) represent plausible hands in the zero level-set of an object-induced neural field, and (ii) project a given pose onto this field. Using HaNDF, we can also optimize for a plausible condition, determining the object under interaction. Our extensive evaluations demonstrate that our HaNDF provides a flexible and powerful prior for hand-object interaction, suitable for applications such as pose refinement, reconstruction under occlusion, and physically plausible manipulation synthesis.
Read the full abstract
Robotics
Meituan-Robotics-0: A Vision-Language-Action Foundation Model for Desktop Manipulation
We present Meituan-Robotics-0 (MR0), a vision-language-action foundation model and embodied-agent system for stationary-base desktop manipulation. MR0 combines 147 million vision-language samples with 15,060 hours of curated robot trajectories spanning 12 embodiments. A three-stage training recipe aligns multimodal representations with robot behavior through discrete action-token pre-training, transfers these priors to continuous cross-embodiment control through flow-matching mid-training, and adapts the policy through real-world post-training. We introduce Rollback, which pairs replaced policy segments with human corrections to provide positive and negative supervision for CFG-DAgger without value learning or advantage labeling. A unified end-effector interface supports heterogeneous embodiments, while asynchronous action-chunk merging maintains smooth execution under variable inference latency. An event-driven agent maintains persistent task events and reassesses plans when progress or feasibility changes, enabling cross-task retrieval and conditional task resumption. MR0 achieves 93.6% success on RoboTwin 2.0 and ranks fifth in the RoboDojo VLA track. Under matched HG-DAgger post-training, MR0 achieves higher average success rate and progress score than $\pi_{0.5}$ across three real-robot tasks. On Ethernet insertion, CFG-DAgger trained with Rollback data reaches 70% success and a progress score of 0.90. Qualitative agent deployments demonstrate cross-task event retrieval and recovery from temporarily unsatisfied task conditions.
Read the full abstract
Robotics
Q-OIL: Optimistic Exploration in Learning from Human Interventions
As robots are deployed in the real world, they will inevitably make mistakes. Human interventions provide valuable feedback for improving their behavior, enabling them to adapt online to new scenarios. However, studying human-in-the-loop robot learning is difficult: real-world experiments are slow and expensive, so most algorithm development relies on simulated human models. We show that common human modeling assumptions do not match how real people intervene in practice. Through analyzing intervention data from human participants, we find that humans tend to intervene when agent progress stalls, and that their corrective actions are often suboptimal. Motivated by these findings, we introduce a progress-based model of human intervention that achieves higher rank correlation with real-user algorithm rankings, enabling realistic algorithm benchmarking and development in simulation. Building on these insights, we propose Q-OIL, a method that adds optimistic bonuses to intervention transitions in reinforcement learning and softly guides exploration toward human corrections. We find that decoupled critics are critical for stability because they prevent the bonuses from backing up to pre-intervention states. As a result, Q-OIL learns more efficiently than competitive baselines, achieving 10-50% higher success in simulation and on real robots, and improving pretrained VLAs on two bimanual manipulation tasks. As we ground both evaluation and algorithms in realistic human behavior, this work provides a practical path toward scalable robot learning from interventions.
Read the full abstract
Optics
Tailoring resonator geometry for astigmatism compensation with structured light in birefringent optical cavities
In this work, we experimentally investigate the astigmatism induced by a birefringent crystal inside an optical cavity and demonstrate a geometric compensation strategy for structured modes carrying orbital angular momentum (OAM). We show that the astigmatism observed in the semimonolithic configuration can be fully compensated by introducing a third mirror to fold the cavity, thereby exploiting the geometric astigmatism arising from oblique incidence on the curved mirror. By performing a study of the resonator geometry, both theoretical and experimental results confirm that this approach enables the preservation of OAM between the pump beam and the converted signal and idler modes, providing an effective design strategy for birefringent optical cavities.
Read the full abstract
Computational Complexity
Composition Theorem for Randomized Query Complexity
For bounded-error randomized query complexity, we resolve the composition conjecture up to logarithmic factors. The general upper bound $R(f\circ g) = O\!\left(R(f)R(g)\log(R(f)+1)\right)$ was already known. Although Meiburg recently disproved constant-factor multiplicative composition for total Boolean functions, the conjecture allowing polylogarithmic losses remained open. We settle this question by proving that, for all (possibly partial) Boolean functions $f$ and $g$, \[ R(f\circ g) = \Omega\!\left(\frac{R(f)R(g)}{\log(n+1)}\right), \] where $n$ is the input size for $f$.
Read the full abstract
High Energy Physics (Phenomenology)
Soft-gluon effective coupling for massive quarks
We present a generalisation of the soft-gluon physical coupling, defined in terms of the strength of soft radiation from a hard dipole, to massive hard emitters. We derive the relation between this coupling and the $\overline{\rm MS}$ coupling through two loops, encompassing all configurations in which both, one, or neither of the hard emitters are massive. We show that the presence of massive emitters introduces additional freedom in the definition of the physical coupling scheme, which we characterise at two-loop order. Our results provide new ingredients for precision QCD resummation in processes involving heavy quarks and for the development of higher-order parton-shower algorithms.
Read the full abstract
Artificial Intelligence
Enhancing Academic Advisement with Case-Based Reasoning
This research involved the study of an application of Case-Based Reasoning (CBR) in the field of academic advising. Academic advisement has been recognised as an important function of higher education. However, there is a widespread desire to improve the advisement services provided by many educational institutions (e.g., colleges and universities) around the world. In the last ten years, CBR expert systems have made important contributions in many domains by applying CBR methodology to a variety of areas of human problem solving. However, to date, expert advisement systems have not been successfully applied to academic advisement, which seems to be a very appropriate domain for this technology. Academic advising is an important and time consuming task, and different tools and techniques must be used to make it an effective and efficient process. This thesis describes the design, development, and validation of a prototype case-based academic advisement recommender system that supplements the conventional advising process, and could eventually evolve to replace human academic advisors when its full potential is realised.
Read the full abstract
Neurons and Cognition
Spiking neuromodulation: a linear control perspective
Neuronal excitability is a nonlinear phenomenon, and the complexity of spiking experimental traces sometimes looks defeating. The present paper takes the view point that excitability is a neuromodulated property, and aims to demonstrate that the neuromodulation of spiking is amenable to the classical linear analysis tools of control theory. We illustrate the value of linear feedback analysis as an alternative to bifurcation analysis in the simple yet classical example of spiking modulation by the A-type potassium current.
Read the full abstract
Data Structures and Algorithms
Weighted and Unweighted Online Connectivity Augmentation via Geometric Hitting Set
We study the Online Connectivity Augmentation Problem (OCAP) with cut arrivals. In this problem, we are given a $k$-edge-connected graph $G = (V,E)$ with $|V| = n$ and a set of links $L \subseteq \binom{V}{2}$. In each iteration, a minimum cut arrives to be augmented, and we seek to maintain a smallest set of links crossing all such cuts. This problem is a special case of online hitting set in which the ground set is $L$ and a sequence of sets to be hit arrive from $\mathcal{C} = \{\delta_L(C) : |\delta_E(C)| = k\}$. As such, OCAP admits a $O(\log^2 n)$-competitive algorithm using the classical algorithm for Online Set Cover of Alon, Awerbuch, Azar, Buchbinder, and Naor. If $k$ is odd, then OCAP reduces to the case where $G$ is a tree. This case was studied by Naor, Umboh, and Williamson, who gave an optimal $O(\log n)$-competitive algorithm, which even holds in weighted setting. Recently, Garg and Subramanian gave an $O(\log n)$-competitive algorithm for unweighted OCAP which holds in the popular node-arrival model, but not for cut-arrivals. No better than $O(\log^2 n)$-competitive algorithm is known for weighted OCAP. In this paper, we provide a nearly optimal $O(\log n \log \log n)$-competitive algorithm for weighted OCAP under cut-arrivals and an optimal $O(\log n)$-competitive algorithm for unweighted OCAP with cut-arrivals. We also give an $O(1)$ algorithm for Online Cycle Augmentation under node-arrivals. Since no $o(\log n)$-competitive algorithm is possible on a cycle with cut-arrivals, this establishes a separation between the best competitive ratio achievable in the two arrival models.
Read the full abstract
Machine Learning
Not All Misalignment Is Alike: A Decomposed Benchmark for Emergent Misalignment
Emergent misalignment (EM), in which fine-tuning on a narrow task induces misaligned behaviour on unrelated prompts, has been demonstrated across a wide range of datasets, model families, and training setups. Yet, most studies continue to rely on the evaluation approach introduced in the original EM paper, producing rates that are sensitive to the small set of evaluation questions used, the choice of judge, and the thresholds applied. These evaluations can also conflate misalignment with capability loss and offer little insight into the specific ways misalignment manifests. Differences in datasets, models, and fine-tuning configurations further make findings difficult to compare across studies. We introduce EMERGE, a decomposed benchmark that tracks changes in eight distinct misaligned behaviours from base to fine-tuned models. It evaluates these behaviours on out-of-domain prompts against matched benign controls and tracks capability, coherence, and refusal to account for potential confounders. Applying EMERGE to 8 open-weight models and 13 fine-tuning datasets, we find that models with similar broad misalignment rates can exhibit very different behavioural profiles that a single score cannot capture. We hope EMERGE can help establish a shared and robust standard for measuring EM, making findings more comparable and helping move research from detecting the phenomenon towards understanding it.
Read the full abstract
Quantitative Methods
Personalized health coaching via coachable health age and counterfactual wearable trajectories
Personalized health coaching requires moving beyond retrospective wearable summaries toward quantitative, actionable behavioral prescriptions. We introduce Coachable Health Age (CHA), an interpretable health-age metric built from modifiable physiological and behavioral signals passively captured by consumer wearables. Across the WEAR-ME (N=2,780), All of Us (N=22,748), and longitudinal Fitbit (N=1,464) cohorts, CHA aligns with established cardiometabolic health constructs and improves prognostic discrimination for incident cardiovascular and metabolic disease beyond age, sex, and body mass index. To enable prospective behavioral optimization, we developed Sensor World Model (SensorWM), an action-conditioned wearable world model pretrained on over 2.2 million person-weeks of longitudinal sensor data from approximately 100,000 participants. SensorWM accurately forecasts multi-week physiological responses to candidate lifestyle changes ($R^2=0.664$), outperforming generic time-series foundation models while preserving downstream clinical signal for diabetes, hypertension, and other health outcomes. Finally, we integrate CHA and SensorWM into an agentic counterfactual coaching system to synthesize personalized 12-week coaching plans. In blinded paired reviews, certified lifestyle coaches preferred counterfactual-grounded plans over plans grounded in wearable history alone for overall quality (71.4%, P < 0.001) and anticipated adherence (71.0%, P < 0.001), and board-certified clinicians judged 99.2% of counterfactual-grounded plans to meet clinical safety criteria (vs. 97.6% of comparison plans). Automated evaluations across the full cohort (N=2,780) also favored counterfactual-grounded plans. By closing the loop between longitudinal tracking, trajectory prediction, and counterfactual coaching, our framework provides a scalable path to systematically improve cardiometabolic health.
Read the full abstract
Geophysics
Reconstruction of Global Groundwater Storage Anomaly from GRACE Satellite Gravimetry Using a Physics-Constrained Decomposition LSTM
Groundwater is the largest accessible store of freshwater on Earth, yet its depletion has been difficult to quantify globally, because satellite gravimetry senses total water storage rather than groundwater alone, the record is coarse and interrupted, and the models used to isolate groundwater underestimate the largest trends. Here, a continuous and physically bounded record of global groundwater storage change was reconstructed for April 2002 to December 2022 at half-degree resolution. Total water storage from the GRACE and GRACE-FO missions was separated from its land-surface components, and a physics-constrained decomposition network learned the groundwater signal as the sum of a trend, a seasonal cycle, and a residual. The trend of every cell was held within the net moisture that the atmosphere can supply, and the mass balance of each river basin was enforced during training, so the learned field could not violate water availability. The reconstruction reproduced withheld observations with a root mean squared error of 6.97 centimeters and a Nash-Sutcliffe efficiency of 0.69, and against an independent network of monitoring wells it achieved a lower trend error than an independent hydrological model and a data-assimilation product at every aggregation scale tested. Gross groundwater depletion over significantly declining cells was estimated at 358 +/- 12 cubic kilometers per year, while net storage change across all land cells was -36 +/- 18 cubic kilometers per year, equivalent to 0.10 +/- 0.05 millimeters per year of sea-level rise. The associated mass redistribution implies a drift of the Earth's rotational pole of 6.9 centimeters per year toward 68.6 degrees East, within 4.5 degrees of the direction inferred in an earlier study.
Read the full abstract
Information Retrieval
CascadeRec: Deployed Adaptive LLM Compute Allocation for Multi-Tenant Content Recommendation
Large language models (LLMs) are attractive rerankers for recommendation. Running an expensive model on every request is costly, and for users with sparse interaction histories it often lowers quality. We present CascadeRec, a two-tier LLM reranking cascade deployed in production at a large financial planning and analysis (FP&A) SaaS provider. It serves content recommendations to enterprise tenants on AWS EKS and GKE. A lightweight Tier-1 model reranks every request. A router escalates only the requests where extra inference-time compute is likely to help, and sends them to a more expensive Tier-2 procedure (self-consistency sampling, multi-path reasoning, or a larger model). The system has run in a production A/B experiment for over a month. Compared with uniform use of the expensive tier, CascadeRec has reduced inference cost by about 10%, with minimal latency impact and no quality regression observed to date. Tenant data cannot be released, so we also evaluate the pipeline on public benchmarks (MovieLens-1M and Amazon-Electronics). Segment-level analysis with paired-bootstrap confidence intervals shows that uniform LLM reranking lowers NDCG@10 for sparse-history users by up to 30%, and its only statistically significant gain appears on the segment with the longest histories. Escalating cold users can make the loss larger instead of repairing it. Which segments benefit from escalation depends on the domain, and this motivates the segment-gated rollout we use in production. We describe the deployment architecture, the measured benefits, and lessons learned, including inter-tenant load imbalance as a driver for cascading and the handling of LLM output-parsing failures at the serving layer.
Read the full abstract
Human-Computer Interaction
Seeing Is Not Knowing: The Epistemic Limits of LLM-Based Mental Health Inference from Youth Social Media Conversations
Adolescents increasingly use social media for private conversations, sharing experiences and seeking support. These interactions generate digital traces, prompting efforts to infer mental health and psychosocial states. As large language models become more capable of reasoning about conversations and interpersonal dynamics, a fundamental question remains nonetheless: what can models actually "know" about an individual's psychological experience from what they are able to "see"? We evaluated ten reasoning LLMs using donated Instagram Direct Messages from 80 adolescents with a history of depression or suicidal thoughts and behaviors, comparing predictions with self-reports and analyzing 204 reasoning traces. Models performed better when inferring more observable states than participants' subjective appraisals of social experience, such as perceived social support. Beyond prediction accuracy, our findings highlight the epistemic limits of digital traces for AI design and evaluation: what can be observed, what can reasonably be inferred, and what requires people's own accounts.
Read the full abstract
Quantum Physics
Thermodynamics of Speed and Efficiency with Quantum Feedback
Rapid thermodynamic operations generally entail irreversible dissipation, giving rise to trade-offs between operating speed and thermodynamic cost. Yet, how quantum feedback fundamentally reshapes these trade-offs remains elusive. Here, we derive a finite-time current–dissipation relation for continuously monitored open quantum dynamics with jump feedback. From this fundamental relation, we establish a quantum speed limit and a power-efficiency trade-off for stationary information engines, identifying how feedback affects state transformations and the performance of engines through both informational entropy and direct dynamical contributions. Our results provide a quantitative foundation for fast, low-dissipation quantum thermodynamics in the presence of feedback control, exemplified by attaining the generalized Landauer bound in finite-time information erasure and approaching unit information efficiency at finite power in the large-system limit of a collective engine.
Read the full abstract
Astrophysics of Galaxies
A combined PRIMER+MINERVA study of the evolution of $\mathbf{[O\mbox{III}]+H\beta}$ equivalent width from $\mathbf{z=3}$ to $\mathbf{z=8.5}$
We present an analysis of the evolution of $[O\,III]+\mathrm{H}\beta$ equivalent width ($W_{\lambda}$) over the redshift range $z=3$ to $z=8.5$, based on combining the JWST NIRCam PRIMER and MINERVA imaging in the UDS survey field. We find that the combined medium- and broad-band data set reduces the photometric redshift outlier rate to $\eta = 4.7\%$ at $m_{\rm F444W}\leq28$, a $42\%$ reduction compared with PRIMER imaging alone. We also demonstrate that the combined data set facilitates the accurate recovery of $[O\,III]+\mathrm{H}\beta$ equivalent widths in redshift bins centred on $z\sim3.2,\, 4.0,\, 5.0,\, 7.2,\, 8.3$; corresponding to the F210M, F250M, F300M, F410M and F460M filters. Based on a final sample of $N=1964$ galaxies with $M_{\rm UV}\leq-19.25$, we first provide new constraints on the bright end of the $z=7-8$ $[O\,III]+\mathrm{H}\beta$ luminosity function ($\log_{10}\phi^{*} = -4.3_{-0.5}^{+0.2}$, $\log_{10}L^{*} = -43.4_{-0.2}^{+0.3}$), before proceeding to investigate the redshift evolution of $[O\,III]+\mathrm{H}\beta$ equivalent width. Using SED modelling based on non-parametric continuity star-formation histories, we find that the median $[O\,III]+\mathrm{H}\beta$ equivalent width monotonically increases with redshift, rising from $\mathrm{log}_{10}(W_{\lambda})=2.54^{+0.01}_{-0.01}$ at $z\simeq 3$ to $\mathrm{log}_{10}(W_{\lambda})=3.05^{+0.03}_{-0.03}$ at $z\simeq 8$. In tandem, the distribution of $\mathrm{log}_{10}(W_{\lambda})$ is observed to become increasingly asymmetric with redshift, producing a fraction of extreme emission-line galaxies (EELGs, $W_{\lambda}>1000\,$Å) that reaches $f_{\rm EELG}=0.50\pm{0.10}$ by $z=8.3$. Our SED modelling results indicate that the physical driver of this evolution is increasingly bursty star-formation, with EELGs the result of particularly high specific star-formation rates over the 10 Myr prior to observation.
Read the full abstract
Differential Geometry
Rigidity, Regularity and Compactness of Special Lagrangian Submanifolds under Simons Stability
In this paper, we prove that, for suitable values of $\delta$, every complete connected $\delta$-Simons stable special Lagrangian immersion in ${\mathbb C}^n$, $n\ge3$, is flat. We also prove that a $\delta$-Simons stable special Lagrangian cone $C\subset{\mathbb C}^n$ with $C\cap{\mathbb S}^{2n-1}$ smooth and compact is a finite union of planes if $\delta>(n-2)^2/(4n)$, and that this coefficient is optimal. At equality, we characterize the $\delta$-Simons stable cones. We further establish regularity and compactness for sequences of proper $\delta$-Simons stable special Lagrangian immersions when $n\ge3$ and $\delta>(n-2)/(n+1)$, under suitable restrictions on the initial singular sets, assuming integral current extensions $T_j$ with $\partial T_j=0$ and uniformly bounded mass on compact subsets. In dimension three, we allow initial singular sets $S_j$ satisfying $\mathcal H^1(S_j)=0$. For suitable values of $\delta$, we also prove properness, Euclidean volume growth, and finite total curvature for complete connected immersions of finite $\delta$-index.
Read the full abstract
Computer Vision
Towards Label-Efficient Geospatial Foundation Model Adaptation via Positive-Unlabeled Learning for Building Damage Mapping
Geospatial foundation models (GFMs) are reshaping remote sensing by shifting the field from task-specific modeling toward general-purpose geospatial intelligence. Mapping the damage to buildings stands to benefit substantially from this shift, as GFMs pretrained on large-scale Earth observation data provide generalizable geospatial representations that can facilitate effective responses to sudden-onset disasters. Although parameter-efficient fine-tuning (PEFT) facilitates GFM adaptation, parameter efficiency does not necessarily translate into label efficiency. Building damage mapping remains challenging due to scarce pre-existing labeled data or limited time for annotation. When weak supervision is introduced to alleviate the labeling bottleneck, we identify the difficulty of learning sufficiently discriminative representations as a key challenge for label-efficient GFM adaptation. To address this challenge, we propose a Positive-Unlabeled GFM Adaptation (PUMA) framework that requires annotations for only a small number of damaged buildings. More specifically, PUMA integrates non-contrastive weakly supervised representation refinement to mitigate the impact of noisy supervision on representation learning from a gradient perspective, and latent-state-based label disambiguation to enable reversible supervision refinement without requiring class prior estimation. Extensive experiments on disaster benchmark datasets from around the world demonstrated the effectiveness and robustness of PUMA. A real-world application to the 2025 Los Angeles wildfires further highlighted its potential to accurately map damage under highly limited supervision. More broadly, these findings suggest that weakly supervised PEFT offers a label-efficient pathway from GFMs to real-world applications.
Read the full abstract
Statistics Theory
Minimax designs and estimators for causal inference under network interference
We develop a design-based minimax theory for linear causal estimands under interference. We study the minimax risks of three problems: optimizing an estimator under a fixed design, optimizing an estimator and a design from a restricted class, or optimizing both over all designs. In each problem, we study the minimax risk over all estimators and over design-unbiased estimators separately. We allow any correctly specified exposure mapping, including the identity. Our theory represents causal inference under interference as a constrained sampling design on an expanded finite population of $\sum_i d_i$ potential outcomes, where $d_i$ is the number of exposure levels attainable by unit $i$. We describe the constrained sampling design by a weighted sampling hypergraph: its vertices are the potential outcomes, its hyperedges are the maximal sets of potential outcomes that can be jointly sampled under feasible treatment assignments, and its weights are induced by the randomization design. We show that the minimax risks of the fixed-design, restricted-design, and all-design problems depend, respectively, on the weighted hypergraph induced by the design, the collection of weighted hypergraphs attainable within the restricted class, and the unweighted hypergraph. Our Bayes lower bounds compare potential-outcome tables that differ on jointly missing sets of potential outcomes: joint-missingness probabilities govern the lower bounds over all estimators, whereas the corresponding odds govern the lower bounds over unbiased estimators. We illustrate the theory for average treatment effects without interference, saturation designs under partial interference, total, direct, and contagion effects under neighborhood interference, and bipartite dose-response contrasts. The upper and lower bounds match in several settings.
Read the full abstract
Machine Learning
PINNeval: A Comprehensive Evaluation Standard for Physics-Informed Neural Networks
The Physics-Informed Neural Network (PINN) literature has produced a rapidly expanding catalogue of methods. We divide a PINN training pipeline systematically into four components and observe that many existing benchmarks study these components in isolation. These components are domain-importance adaptation, loss balancing, neural network architecture, and optimization scheme. Moreover, the studies set the surrounding and not considered components of the pipeline to a minimal default that is no longer representative of current best practice approaches. Conclusions drawn under such conditions may not transfer to competitive pipelines. In this work, we propose a new evaluation protocol that (1) treats the four components as an integrated system; (2) crucially requires competitive implementations of the components not under investigation; (3) specifies the metrics and reporting format; and (4) is accompanied by structured documentation that ensures the traceability of evaluations across studies. Based on this comprehensive protocol, we release PINNpipe, an innovative JAX package containing 18 different methods from the current literature as well as nine efficiently implemented benchmark problems. Using PINNpipe, we illustrate our evaluation protocol in relation to established benchmark studies and report results that challenge common assumptions.
Read the full abstract
Machine Learning
Minimum-Sufficient Knowledge and Computation for Wind Energy Predictive Analysis
Resource-constrained cyber-physical decision systems often apply fixed sensing, knowledge, and computational pipelines despite state-dependent information needs. We introduce Minimum-Sufficient Knowledge and Computation (MSKC), a resource-aware computing principle: select the least-costly route meeting decision-risk and deployment constraints, escalating only when cheaper routes are insufficient. As proof of concept, we isolate adaptive knowledge routing, the decision underlying Knowledge-Guided Machine Learning (KGML), which conventionally injects external knowledge unconditionally rather than only where needed. We instantiate this for wind ramp-event predictive analysis, an edge-oriented decision-support task, evaluating local, relational, and physics-informed routes with a state-aware gate on publicly available real data from two wind farms. We establish conditions for feasibility preservation, cost dominance, and strict expected savings over always-on routing. At a 35.9% relational-knowledge injection rate, adaptive routing is statistically indistinguishable from always-on injection on the reported metrics while reducing knowledge calls by 64%, supporting sustainable, Green-IT-aligned deployment. These results support MSKC's premise that uniformly maximal intelligence is unnecessary. This offers intelligent data analysis for big data streams: the analysis itself decides, per state, how much knowledge and computation the data warrant.
Read the full abstract
Classical Physics
An Overview of Classical Scattering by a Lennard-Jones Potential
We present a numerical and conceptual overview of classical scattering under a Lennard-Jones potential. All results were obtained by direct integration of particle trajectories in the two-dimensional plane for varying impact parameters and energies. From these trajectories, we compute deflection functions, identify critical values associated with unstable circular orbits, and visualize non-monotonic behavior in the angular deflection. We construct color maps representing the deflection angle and interaction delay time as functions of energy and impact parameter, and determine the differential cross section from the deflection function, accounting for multi-valued contributions. These results are displayed both as individual plots and as a full color map in the energy-angle plane, providing a detailed picture of the scattering behavior in this central potential.
Read the full abstract
Computational Geometry
The Containment Polytope of a Triangle and a Convex Polygon
The containment polytope $\mathcal{C}$ for a triangle $P$ and a convex $n$-gon $Q$ is a four-dimensional polytope representing the similarity transforms (translation, rotation, scaling) that place $P$ inside $Q$. While it is known that the complexity of $\mathcal{C}$ and its silhouette $\Lambda$, which represents all the largest placements of $P$ inside $Q$ for all orientations, can be quadratic in $n$, we show that the silhouette can be partitioned into a linear number of planar convex chains and steps in $O(n \log n)$ time. Based on the linear number of convex chains and steps, we can compute the complexity of silhouette $\Lambda$ in $O(n \log n)$ time and query the largest placement of $P$ inside $Q$ for a given orientation in $O(\log n)$ time. We explain geometrically how the triangle $P$ rotates while continuously scaling and translating to be maximal inside $Q$, leading to a parameterized algorithm for containment problems.
Read the full abstract
Machine Learning (Statistics)
pESTO: Iterative p-greedy kernel regression with risk-driven Early STOpping
Kernel-based greedy algorithms select suitable approximation nodes by minimizing point-wise error indicators, such as the power function; in this context, they are usually referred to as p-greedy approaches. In the presence of noisy data, deterministic criteria may fail, as the discrepancy between the true solution and its estimate is influenced by both approximation error and noise amplification, and the balance between the two is unknown. This study explicitly separates these two contributions, which are governed by the power function and a Lebesgue-type term, respectively. These bounds motivate the use of the Stein's unbiased risk estimator (SURE) as a proxy of the mean squared error, which in turn provides a principled and computable stopping criterion for the number of nodes to select in the p-greedy procedure. Numerical experiments demonstrate that this approach effectively balances deterministic accuracy and statistical noise contributions.
Read the full abstract
Functional Analysis
Fejer-Riesz factorization for free products
We establish an operator-valued Fejer-Riesz factorization theorem for the algebraic free product A of C[F], the group algebra on the free group F on finitely many generators, with finitely many finite dimensional C-star algebras. With respect to a natural reduced-word length, called extent, we show that every linear combination F of elements of A with coefficients from the algebra of operators on a Hilbert space of extent at most 2d that is positive under all finite-dimensional unital *-representations of A admits a factorization F=B^*B with B of extent at most d. This sharp half-extent bound holds without strict positivity. The proof combines a new finite-dimensional realization theorem for positive truncated kernels with operator-system methods. We also establish factorization results for tensor products with group algebras of arbitrary discrete groups under strict positivity in the maximal C-star-tensor product; for amenable groups, strict positivity in the minimal tensor product suffices.
Read the full abstract
Combinatorics
Algorithms for Short Human-Verifiable Proofs of Non-Word-Representability
A graph is word-representable if it can be represented using alternation of letters in words. The class of word-representable graphs generalizes several important graph families and admits a characterization via semi-transitive orientations. Recognizing word-representability is an NP-complete problem, and a major bottleneck in the theory of word-representable graphs is providing efficient justifications of non-word-representability of graphs, rather than merely referring to software verification. Recently, Kitaev and Sun initiated the development of methods for the automatic generation of human-verifiable proofs of non-word-representability. They introduced three algorithms and evaluated their performance on the 25 connected non-word-representable graphs on seven vertices, which constitute a fairly small sample set. In this paper, we confirm their ranking of the algorithms by testing them on all non-word-representable graphs on eight and nine vertices. We further introduce three new algorithms, one of which significantly outperforms the previously best method on graphs with at most nine vertices, producing substantially shorter proofs. We apply this algorithm to shorten a known proof of non-word-representability of the Shrikhande graph and to produce a proof of non-word-representability of a certain subgraph of the Kneser graph $K(8,3)$. This subgraph is the first known example of a triangle-free non-word-representable graph, presented in the literature with reference only to a computational proof. As a by-product of our study, we correct an error in the repeatedly published list of the 25 connected non-word-representable graphs on seven vertices.
Read the full abstract
Fluid Dynamics
An Analytical Model for Laser-Activated Two-Phase Choking and Microfluidic Suction on Surface Boundary Layers
We present a predictive fluid mechanics model to calculate pressure drops (0.25 atm) and fluid velocities (2.50 m/s) governed by localized phase change and choking phenomena in a capillary-constrained, coaxial microfluidic actuator. Driven by near-infrared thermal absorption, the analytical framework describes how energy deposition drives vapor cavity growth and subsequent cavitation bubble collapse, inducing a transient suction force across a curved, two-radii Young-Laplace capillary boundary. Velocity drops are derived from closed-form analytical equations that explicitly couple vapor volume dynamics with non-linear flow area restriction and fluid drag. The model predicts a maximum boundary layer wall shear stress of 30.06 Pa at the target interface. Crucially, this shear magnitude matches the physical threshold required to mechanically strip surface contaminant boundary layers while remaining below the damage threshold of underlying target substrates. Finally, we formulate a closed-form expression for the maximum allowable liquid inlet temperature required to prevent thermal saturation and sustain stable, continuous microfluidic suction cycles. These equations establish direct engineering design rules for optimizing phase-change microfluidic actuators and next-generation coaxial fluidic systems.
Read the full abstract
Machine Learning
How Small Can We Go? Calibrating Activation-Cache Compression for SAE Training
To train a sparse autoencoder (SAE), you need a massive cache of model activations. We look into the effects that compression has on an SAE in two ways: we keep the original fp32-trained SAE and pass the compressed cache into the same SAE to see how many feature activations stay the same (the frozen-SAE test), and we retrain an SAE on the compressed cache and see how many of the original features still exist (the retraining test). We found that in the retraining test, there is no clear scale for retained features. On a GPT-2 cache, rounding everything to fp16 keeps 60.6% of features, but changing one value in a million by the smallest step fp32 allows (a change eight orders of magnitude smaller) keeps only 70.5%. Meanwhile, changing the random seed keeps 23.0%. For the frozen-SAE test, the frozen SAE's correlation with the original increases in order of how precise the codecs are, but plain $L_2$ distance gives nearly the same ordering. We also see the same frozen reading correspond to different percentages retained at different sites. This leaves the question of the right measurement. We build one by calibrating: retraining on identical bytes sets the ceiling, and a reseed sets the floor. We then measure the effect of each compressed codec against its own site's floor. With TopK SAEs on GPT-2 small and Pythia-1.4B at research-scale budgets, naïve per-token int4 loses more features than a reseed everywhere we tested, as do PCA and random projection on GPT-2. Standardizing each channel before quantizing brings int4 back above the floor in terms of retained features, at the same raw file size. At our smallest budget, a bf16 forward pass, which some pipelines use to save memory, loses slightly more features than rotated 8-bit storage. Calibrating a site costs four training runs and two evaluation passes.