arXiv:2607.09135v1 Announce Type: new
Abstract: Medical images require comprehensive and accurate interpretation to support the diagnosis of diverse clincial conditions. Recent vision-language generalist models offer broad task coverage and promising zero-shot capabilities, yet often lack fine-grained anatomical and lesion awareness for reliable diagnosis and spatial interpretability. In contrast, supervised specialist models achieve strong performance on specific tasks but typically lack generalization across diseases and anatomies. In this work, we present SuG, a Super-Generalist framework that unifies generalist vision-language learning with specialist objectives, enabling both broad generalization and specialist-level diagnostic capability. We perform specialist-enhanced vision-language alignment in SuG by incorporating spatial priors from multiple segmentation experts, including anatomy, class-specific lesion and class-agnostic lesion segmentors that captures lesions beyond anatomies annotated during training. To improve lesion grounding capability, we leverage lesion masks as spatial priors to calibrate text-conditioned visual attention, encouraging disease-related semantics to focus on clinically relevant regions. We evaluate SuG on extensive chest and abdominal CT benchmarks, including CT-RATE, Merlin, MedVL-CT69K, and several in-house tumor datasets. SuG achieves state-of-the-art performance across a wide range of disease diagnosis tasks and surpasses specialist models on several critical tumor diagnosis benchmarks. Furthermore, SuG demonstrates strong lesion grounding capability, including robust generalization to lesion types lacking class-specific supervision.
Science Journals
arXiv:2510.19588v2 Announce Type: replace-cross
Abstract: Valley, as a ubiquitous degree of freedom in lattices, has found wide applications in both electronic and classical-wave devices in recent years. However, achieving valley-polarized states, a prerequisite for valley-based operations, still remains challenging. Here, we propose and experimentally demonstrate a universal non-Hermitian mechanism for valley filtering using only uniform background dissipation, which creates a propagation length contrast between valleys through their intrinsic group velocity differences. We implement this concept in an acoustic crystal, observing switchable and robust valley polarization of sound through large-scale field mapping. Remarkably, our approach is solely based on uniform loss, without the need for any special lattice structures, tailored excitations, or external fields. We further provide designs of our non-Hermitian valley filter on photonic and electronic platforms. Our results offer a simple and effective solution to valley-polarized state generation and may advance the development of novel valley-based devices in both classical and quantum regimes.
arXiv:2605.05046v2 Announce Type: replace
Abstract: We study the sampling problem for simultaneous edge colorings. Given a pair of graphs $G_1=(V,E_1)$ and $G_2=(V,E_2)$ which are on the same vertex set $V$, a simultaneous edge coloring is an edge coloring of $G_1\cup G_2$ so that each of the individual graphs is properly colored. When each of $G_1$ and $G_2$ are of maximum degree $\Delta$, then it is conjectured that $\Delta+2$ colors suffice, and recent work asymptotically establishes the conjecture.
We study Markov chains for randomly sampling from the uniform distribution over simultaneous edge colorings. Straightforward applications of Jerrum's classical coupling argument establish rapid mixing of the Glauber dynamics on the corresponding line graph when $k>8\Delta$. We present a simple weighted Hamming distance for which Jerrum's coupling yields optimal mixing time (up to constant factors) of $O(m\log{n})$ when $k>(6+\delta)\Delta$ for any fixed $\delta>0$. Moreover, utilizing the flip dynamics with our new metric, we obtain $O(m\log{n})$ mixing of the flip dynamics when $k\geq 5.948\Delta$, using a local choice of flip parameters which only flips bounded-size components. The proof adapts previous coupling analyses for the flip dynamics to the setting of simultaneous edge colorings.
arXiv:2605.05129v3 Announce Type: replace
Abstract: We consider the Landau-Lifshitz-Gilbert equation (LLG), which models time-dependent micromagnetic phenomena. We analyze a fully discrete scheme that combines first-order finite elements in space with a BDF2 method in time. The method requires the solution of only one linear system of equations per time step and does not enforce the pointwise unit-length constraint of the magnetization. While unconditional weak convergence has been analyzed in an earlier work, we now prove optimal-order convergence rates under sufficient regularity assumptions on the exact solution and the external field. In combination with our previous work, this establishes the first higher-order-in-time and linear integrator that converges both to weak and strong solutions of LLG. Numerical experiments confirm first-order convergence in space and second-order convergence in time.
arXiv:2605.12920v3 Announce Type: replace
Abstract: Effective collaboration between embodied agents requires more than acting in a shared environment; it demands communication grounded in each agent's evolving understanding of the world. When agents can only partially observe their surroundings, coordination without communication is provably hard, but communication can, in principle, bridge this gap by allowing agents to share observations and align their world models. In this work, we examine whether LLM-based embodied agents actually realize the ability to communicate. We extend PARTNR, a benchmark for collaborative household robotics, with a natural-language dialogue channel that enables two agents with partial observability to communicate during task execution. To evaluate whether dialogue leads to genuine world-model alignment rather than superficial coordination, we propose a framework for measuring world-model alignment defined over per-agent world graphs: observation convergence (do private world models align over time?), information novelty (do messages convey what the partner lacks?), and belief-sensitive messaging (do agents model what their partner knows?). Our experiments across three LLMs reveal that dialogue reduces action conflicts 40 to 83 percentage points but degrades task success relative to silent coordination. Using our metrics, we characterize the gap between superficial coordination and genuine world-model alignment, and identify where current models fall on this spectrum. Project Website: https://uiuc-conversational-ai-lab.github.io/partnr-dial-wmd/
arXiv:2607.09267v1 Announce Type: new
Abstract: Mobile robots operating indoors must re-identify previously observed objects after long temporal gaps, significant viewpoint changes, and severe illumination variations. This remains a challenging problem: multi-object tracking methods are optimized for short-term association of pedestrians and vehicles at video rates, person and vehicle re-identification approaches lack persistent memory mechanisms, and state-of-the-art video object segmentation techniques rely on reactive distractor filtering rather than enforcing global identity consistency.
To address these limitations, we present REMIND, an online tracker designed for long-term multi-object re-identification of generic indoor objects from monocular RGB imagery, requiring neither camera pose nor depth. Motivated by evidence from visual cognition that humans rely on accumulated appearance familiarity and spatial context rather than explicit self-localization, REMIND combines frozen DINOv3 features with a dual-bank multi-prototype appearance memory, part- and background-level descriptors, a neighbour-context reasoning module exploiting spatial co-occurrence, and joint Hungarian assignment with ambiguity-aware safeguards. On a purpose-built indoor dataset featuring controlled revisits and dense same-class clutter, REMIND reaches 90.35% IDF1, nearly 20 points above a state-of-the-art video object segmentation baseline and more than 36 above a strong tracking-by-detection baseline. On ScanNet++, it attains the highest IDF1 in every setting but one, end-to-end detection over all scenes, where the tracking-by-detection baseline is marginally ahead while REMIND still associates and recovers identities more accurately; it also completes every scene, whereas the video object segmentation baseline exhausts GPU memory on 66.9% under YOLO detections. The complete system, evaluation framework, and dataset are publicly released.
arXiv:2607.09614v1 Announce Type: new
Abstract: Predicting where floodwater goes and how deep it gets, at high resolution and across large domains, remains computationally expensive with conventional hydraulic solvers, while purely data-driven surrogates are fast but lack physical guarantees and generalize poorly beyond their training events. We present Inunda, a GPU-native flood inundation model that solves the two-dimensional shallow water equations. Inunda uses a mass-conservative local-inertial scheme and runs multi-day events over millions of cells in minutes on a single GPU. Because every operator is autograd-compatible, the solver is differentiable by construction: model parameters can be estimated by gradient descent against gage observations through reverse-mode automatic differentiation of the full simulation. We demonstrate Inunda on three case studies. For a hindcast of Hurricane Harvey (2017) in Harris County, Texas, Inunda matches surveyed high-water marks to a mean absolute error of 0.67 m, competitive with or better than a suite of established flood models, and reaches a median gage water-level Nash Sutcliffe efficiency of +0.72, more than double the +0.31 of the operational National Water Model v3.0. For the July 2025 Central Texas flash flood, Inunda is driven by an 18-member 1-km convection-allowing precipitation ensemble to produce probabilistic flood forecasts whose skill improves systematically as lead time to the crest shortens. For a post-fire flash-flood application in the Rio Ruidoso burn scar, differentiable calibration recovers the saturated hydraulic conductivity as a spatially explicit field at the model's own resolution and traces its multi-year post-fire recovery. Inunda provides an open, end-to-end pipeline for real-event flood modeling that couples the accuracy of physics-based hydraulics with the calibration and coupling advantages of modern differentiable programming.
arXiv:2606.21747v2 Announce Type: replace
Abstract: Motivated by the need to extrapolate the results from ground-based experiments to the conditions of high-speed flight, we present the Lie equivalence symmetry analysis of the hypersonic stagnation-point boundary layers. We demonstrate the application of the equivalence symmetry on the set of coupled ODEs which are physically relevant in hypersonics. By allowing the property laws to transform along with the independent and dependent variables, the invariants derived within the similarity-reduced stagnation-point ODE formulation identify families of non-linear maps that can be used to extrapolate the laboratory-scale predictions to flight. In practice, implementing these maps requires the laboratory-scale ODE solution together with the lab- and flight-side thermochemical property data, which are generally available from existing databases. The maps are derived for the similarity-reduced non-dimensional temperature across the boundary layer and are shown to {collapse with independently computed flight solutions} for a range of relevant cases.
arXiv:2606.23964v2 Announce Type: replace
Abstract: Self-supervised learning in fluorescence microscopy often relies on 2D projections, despite the inherently three-dimensional nature of cells. We present a systematic comparison of 2D and 3D masked autoencoders (MAE-2D vs. MAE-3D) on volumetric microscopy data. Under matched architectures and training protocols, MAE-3D consistently outperforms 2D max-projection and slice-based variants on downstream single-cell tasks. We further align visual representations with a pretrained protein language model (ESM2) and show that cross-modal supervision yields larger gains for volumetric models. Channel cross-attention and frequency-domain regularization are critical for leveraging 3D spatial context. On protein--protein interaction prediction, our best model achieves a ROC--AUC of 0.86, while on protein localization it reaches an AUC$_{\text{micro}}$ of 0.95 and an F1$_{\text{micro}}$ of 0.74, demonstrating competitive performance on both tasks. Overall, our findings highlight the potential of volumetric modeling and multimodal alignment for representation learning in single-cell microscopy.
arXiv:2604.20690v2 Announce Type: replace
Abstract: We study unidirectional transverse scattering in a two-dimensional acoustic dimer composed of two circular subwavelength scatterers. Using a coupled multipole model, we show that interparticle coupling enables effective monopole--dipole interference and supports a transverse Kerker effect under plane-wave excitation. In contrast to a single non-absorbing rotationally symmetric particle, where Kerker-type directional scattering is only approached in the weak-scattering limit, the dimer can combine pronounced directionality with strong overall scattering. This regime is promising for compact acoustic beam steering and directional wave routing.
arXiv:2602.10111v2 Announce Type: replace
Abstract: Learning-based controllers have achieved impressive performance in agile quadrotor flight but typically rely on massive training in simulation, necessitating accurate system identification for effective Sim2Real transfer. However, even with precise modeling, fixed policies remain susceptible to out-of-distribution scenarios, ranging from external aerodynamic disturbances to internal hardware degradation. To ensure safety under these evolving uncertainties, such controllers are forced to operate with conservative safety margins, inherently constraining their agility outside of controlled settings. While online adaptation offers a potential remedy, safely exploring physical limits remains a critical bottleneck due to data scarcity and safety risks. To bridge this gap, we propose a self-adaptive framework that eliminates the need for precise system identification or offline Sim2Real transfer. We introduce Adaptive Temporal Scaling (ATS) to actively explore platform physical limits, and employ online residual learning to augment a simple nominal model. {Based on the learned hybrid model, we further propose Real-world Anchored Short-horizon Backpropagation Through Time (RASH-BPTT) to achieve efficient and robust in-flight policy updates. Extensive experiments demonstrate that our quadrotor reliably executes agile maneuvers near actuator saturation limits. The system evolves a conservative base policy with a peak speed of 1.9 m/s to 7.3 m/s within approximately 100 seconds of flight time. These findings underscore that real-world adaptation serves not merely to compensate for modeling errors, but as a practical mechanism for sustained performance improvement in aggressive flight regimes.
arXiv:2607.04261v2 Announce Type: replace
Abstract: Current Legal Judgment Prediction (LJP) is constrained by its reliance on post-hoc judicial materials, increasing the likelihood that models perform retrospective classification rather than true forecasting. This paper empirically investigates shortcut learning in this context by studying claim-level outcome prediction in UK Employment Tribunal (UKET) decisions. Using a corpus of 33,158 individual claims, we predict outcomes from claim texts and LLM-extracted case summaries, evaluating models ranging from interpretable TF-IDF-based classifiers to black-box LLMs. While headline predictive performance figures appear strong, we demonstrate that such performance in LJP systems trained on post-hoc judicial text can be driven by the retrospective nature of the source material. Stratifying the test data by human judgments of leakage reveals that performance increases where outcome-revealing cues are embedded in the narrative. Moreover, a model trained on just the 4% of features identified as leakage achieves high performance, outperforming human experts. These findings substantiate concerns that LJP performance may be exaggerated by linguistic artefacts. Yet this vulnerability is not fatal to the research agenda. Instead, post-hoc judgments might be treated as potentially contaminated texts, requiring active auditing. Retraining models after masking leakage features results in only a negligible reduction in Macro-F1. Hence, while models will opportunistically exploit shortcuts when available, they remain capable of extracting useful predictive signals when these artefacts are removed.
arXiv:2606.12187v2 Announce Type: replace
Abstract: We consider the strategic facility location problem in $\ell_q(\mathbb R^d)$ spaces where the social cost is defined by an arbitrary $p$-norm of the individual costs. While the optimal approximation ratios for deterministic strategyproof mechanisms are well established in the $d = 1$ setting, the guarantees for multi-dimensional spaces under an arbitrary $p$-norm are less understood.
In this work, we analyze the well-studied, strategyproof coordinate-wise median (CM) mechanism and provide approximation guarantees for these generalized social costs.
* We show that the CM mechanism is in fact robust to a broader class of social objectives: for every monotone symmetric norm objective, including all $p$-norm social costs, its approximation ratio never exceeds $3$ in arbitrary $\ell_q(\mathbb R^d)$ spaces, regardless of the dimension.
* For $d = 2$, we establish tight approximation ratios for all $p, q \geq 1$ in $\ell_q(\mathbb R^2)$. In particular, we show that the CM mechanism is a $2^{1-1/\max(p,q)}$-approximation, resolving the conjecture of Goel and Hann-Caruthers (Social Choice and Welfare, 2023) in the Euclidean case and extending the guarantee to arbitrary $\ell_q$ distances.
* For $d\geq 3$, we refine the dimension-independent approximation guarantee for $p$-norm social costs in $\ell_q(\mathbb R^d)$ spaces, giving upper bounds that depend on the relationship between the social-cost norm $p$ and the underlying distance norm $q$. This generalizes the recent result of Gravin and Jia (STOC, 2025) for the utilitarian social cost.
arXiv:2606.13402v2 Announce Type: replace
Abstract: We present a robust, ultrafast and highly efficient setup for distinguishing molecular enantiomers by combining ultrafast techniques with vector beams, a type of topological light with azimuthally varying polarization. An infrared vector beam generates high-order harmonics in a sample of randomly oriented chiral molecules, resulting in the emission of an ultraviolet vector beam whose intensity profile carries information about the handedness of the chiral molecules. Our approach allows for spatial discrimination of molecular enantiomers, opening a new route for studying ultrafast chirality.
arXiv:2607.06988v2 Announce Type: replace
Abstract: Steering robot foundation models (RFMs) toward new task variants or user-preferred behaviors remains challenging, often requiring additional robot demonstrations, task-specific fine-tuning, or long-context conditioning. We present WAM-TTT, a test-time training framework for steering world action models from raw human videos. Rather than treating human videos as trajectories to imitate, WAM-TTT absorbs them into a lightweight adaptive memory inside a frozen WAM through self-supervised video prediction. To make this memory useful for control, we introduce a meta-training stage that aligns human demonstrations with robot behaviors using paired human-robot data and a key--value memory reconstruction objective. At test time, only unlabeled human videos are required to adapt the memory, while the pretrained WAM remains frozen. This enables efficient and reusable steering without robot actions, human-side annotations, or task-specific fine-tuning, while preserving the generalization ability of the foundation model. Extensive experiments show that WAM-TTT consistently outperforms in-context human-video conditioning baselines across diverse manipulation tasks and generalization settings.
arXiv:2606.20777v2 Announce Type: replace
Abstract: Submodular sparsification generalizes the classical sparsification problems of graphs and matrices to summations of submodular functions. Given the summation $F(S):=f_1(S)+\cdots+f_m(S)$ of $m$ submodular functions $f_1,\ldots,f_m:\{0,1\}^n \to \mathbb{R}_{\ge 0}$. An size-$s$ sparsification of $F$ is a weight vector $w \in \mathbb{R}^m_{\ge 0}$ such that $w_1 f_1(S) + \cdots w_m f_m(S) \approx F(S)$ for every subset $S \subset [n]$. Motivated by the wide applications of submodular functions in data mining and economics, submodular sparsification has been studied in the last few years. For general submodular functions, Kenneth and Krauthgamer provided an efficient construction of size $O(n^3)$. Although several families of submodular functions admit sparsifiers of size $\tilde{O}(n)$, there is a lower bound $\Omega(n^2)$ on the size of sparsifiers by Cohen et al.
In this work, we study whether cardinality constraints, such as restricting $S$ to subsets of size at most $k$, could reduce the size of sparsifiers or not. Namely, if the guaranty is $w_1 f_1(S) + \cdots w_m f_m(S) \approx F(S)$ for every $S$ in $[n]$ of cardinality at most $k$, are there sparsifiers of size smaller than $o(n^2)$? Our main result shows an efficient construction of size-$O(n k^2 \log n)$ sparsifiers for summations of arbitrary submodular functions. This improves the $\Omega(n^2)$ bound for the general setting. Then we consider the existence of size-$(k \log n)^{O(1)}$ sparsifiers under the constraint of cardinality at most $k$ and show several natural families do not admit such a small sparsifier.
Technically, our algorithm applies the Lov\'{a}sz extension and Edmonds' greedy algorithm to extend Kenneth and Krauthgamer's approach. In particular, we provide an efficient algorithm to provide a tight estimate (up to a constant) of the sensitivity of each $f_i$ under cardinality constraints.
arXiv:2606.21250v2 Announce Type: replace
Abstract: The anti-ferromagnetic multi-state Potts model, a generalization of the Ising model, is one of the most fundamental models in statistical physics. It was conjectured by Koteck\'{y} (Phys. Rev. B, 1985) that the model undergoes a phase transition from a disordered phase at infinite temperature to an ordered phase at sufficiently low temperature on lattices. Such phase transitions are believed to play an important role in computational complexity theory and remain closely connected to the problem of approximating the partition function of the system. For proper three-coloring models (corresponding to the zero-temperature), torpid mixing of a family of local-update Markov chains on lattices was established by Galvin, Kahn, Randall and Sorkin (SIDMA, 2015), coinciding with the presence of phase coexistence following shown by Feldheim and Spinka (J. Eur. Math. Soc., 2019).
In this work, we study approximating the partition function of the anti-ferromagnetic multi-state Potts model at low temperature on random regular bipartite graphs, which are with high probability good bipartite expanders. On the negative side, we generalize the result by Geisler, Kang, Sarantis and Wdowinski (arXiv, 2026) for anti-ferromagnetic Ising models to show that when the temperature is sufficiently low relative to the degree of the underlying graph, the celebrated single-site Glauber dynamics has exponentially slow mixing time. On the positive side, we design a deterministic algorithm that yields an approximation to the partition function of the model via the framework of abstract polymer models as Jenssen, Keevash and Perkins (SICOMP, 2020), Liao, Lin, Lu and Mao (Theor. Comput. Sci., 2022), Galanis, Goldberg and Stewart (TOCT, 2021) and Geisler, Kang, Sarantis and Wdowinski (arXiv, 2026).
arXiv:2603.24190v2 Announce Type: replace-cross
Abstract: We study the nonlinear chaotic dynamics in a system of linear oscillators coupled by social network links with an additional stratification of oscillator energies, or frequencies, and supplementary nonlinear interactions. It is argued that this system can be viewed as a model of social stratification in a society with nonlinear interacting agents with energies playing a role of wealth states of society. The Hamiltonian evolution is characterized by two integrals of motion being energy and probability norm. Above a certain chaos border the chaotic dynamics leads to dynamical thermalization with the Rayleigh-Jeans (RJ) distribution over states with given energy or wealth. At low energies, this distribution has RJ condensation of norm at low energy modes. We point out a similarity of this condensation with the wealth inequality in the world countries where about a half of population owns only a couple of percent of the total wealth. In the presence of energy pumping and absorption, the system reveals features of the Kolmogorov-Zakharov turbulence of nonlinear waves.
arXiv:2604.16602v2 Announce Type: replace-cross
Abstract: Searches for QCD axions with masses in the neV/$c^2$ mass range are strongly motivated by new physics at the GUT scale and by well-motivated pre-inflationary axion symmetry breaking scales. This parameter space is challenging to probe due to the small axion-photon couplings, which typically require large, high-field magnets with substantial stored energy. In this paper, we propose a new experimental geometry based on a narrow-bore, segmented solenoid that optimizes the collection of the axion-induced signal using LC resonators outside the high-field region of the magnet bore. This alternative optimization significantly reduces the required stored magnetic energy while preserving sensitivity, enabling a near-term experiment in the 30-200 MHz (120-830 neV/$c^2$) range, with a cost-effective, staged scaling to a GUT-scale experiment in the 100 kHz-30 MHz (0.4-120 neV/$c^2$) range.
arXiv:2606.23969v2 Announce Type: replace
Abstract: GPU Confidential Computing (GPU-CC) now preserves GPU-local performance: on NVIDIA B300, BF16 matmul runs at 0.998x of non-confidential performance. Yet LLM serving under Intel TDX plus GPU-CC still loses 13-27% of throughput, and KV-cache restore latency can more than double. This paper studies that gap on two Blackwell platforms, RTX Pro 6000 and B300 HGX, and identifies its dominant cause: the confidential VM-GPU bridge, not GPU compute.
We find that GPU-CC turns host/device movement into a serialized, high-setup-cost channel. Secure copies do not gain CUDA-stream concurrency within a context, asynchronous transfers block at the runtime boundary, and small crossings pay a fixed toll. This violates the assumptions of modern inference runtimes, where DMA is expected to be cheap, concurrent, and asynchronous. In vLLM dense decode, the gap closes around 44x-slower small alloc-and-copy operations; targeted patches reject alternative explanations. A scheduling flag recovers 57% of the gap, while a worker-thread drain recovers up to 92% in qualified high-concurrency runs. The same bridge model explains a +131% KV-restore penalty and a 34x model-load slowdown.
Blackwell also changes the confidential tenancy unit. We qualify confidential multi-GPU NVSwitch tenants on B300, including 510 GB/s NVLink P2P inside a CVM and concurrent isolated tenants, and identify the remaining fabric-attestation gap for production confidential AI platforms.
arXiv:2607.00118v4 Announce Type: replace
Abstract: The Minimum Temporal Path Cover (TPC) and Minimum Temporally Disjoint Path Cover (TDPC) problems were introduced by [Chakraborty, Dailly, Foucaud, Klasing, MFCS '24]. Both were shown to be NP-hard on temporal DAGs, while the latter is also NP-hard on temporal oriented trees. All tractable cases for T(D)PC established in that paper satisfy a temporal Dilworth property, namely that the size of the minimum T(D)PC is equal to the size of the maximum antichain. This raises a natural question: is T(D)PC polynomial-time solvable under the promise that the respective Dilworth property holds? In this work, we answer this question in the affirmative for both problems, proving in fact that, under the respective promise, the size of the minimum T(D)PC is exactly equal to the Lov\'asz number of the connectivity graph.
In another direction, we establish parameterized algorithms and hardness results for TPC and TDPC. Our main result is that TPC is W[1]-hard parameterized by the deletion distance to linear forest even for temporal graphs with two time-steps, answering in the negative an open question by Chakraborty et al. about whether an XP algorithm parameterized by treewidth plus number of time-steps can be improved to FPT. On the other hand, we prove that an FPT algorithm does exist if the vertex cover number is used as parameter instead of the treewidth in the above parameterization. We complement this with a proof that including the number of time-steps in the parameter is necessary to yield tractability, as, otherwise, both TPC and TDPC remain NP-hard even for constant vertex cover size. Along the way, we establish various other para-NP-hardness results involving structural parameters such as the pathwidth and the maximum degree of the underlying graph.
arXiv:2607.04527v2 Announce Type: replace-cross
Abstract: Biological systems exhibit a hierarchical structure, characterised by directed flow from upstream regulators to downstream effects. Although this ordering provides a natural scaffold for causal inference, most causal discovery and GRN methods either ignore the tiered organisation or condition on all upstream variables, which becomes infeasible for high-dimensional omics data. We present ASCEND (Ancestral Scalable Causal discovEry via iNherited Descent), a constraint-based framework that leverages known two-tiered structure to enable genome-scale causal discovery. ASCEND introduces a divide-and-conquer strategy that maintains dynamically updated ancestral conditioning sets for each downstream variable, dramatically reducing the number of conditional independence tests required, and achieves polynomial-time complexity where traditional approaches face exponential blow-up. Through extensive simulations and real biological data, we demonstrate that ASCEND accurately recovers ancestral relationships, scales properly and much faster, and outperforms existing gene regulatory network inference methods in both causal precision and computational efficiency. The algorithm's ability to resolve directionality makes it particularly suited for integrating multi-omic data where upstream regulators (e.g., SNPs, methylation sites) and downstream responses (e.g., gene expression) are measured jointly.
arXiv:2607.00728v2 Announce Type: replace
Abstract: Graph approximate-nearest-neighbor (ANN) indexes (HNSW, DiskANN/Vamana) lose recall under insert/delete churn, because deletions orphan the greedy-search paths that route through removed nodes. Production systems restore navigability by repairing the graph on a fixed schedule (consolidate every X operations). We asked whether triggering local edge repair on a measured navigability-degradation signal, rather than a blind clock, spends a fixed repair budget better. At matched repair budget, it does not. On two real ANN datasets (SIFT-128 and Fashion-MNIST-784) under a bursty churn stream, compared against a fixed-cadence baseline actually run at the triggered policy's realized consolidation count, the tail-recall advantage is indistinguishable from zero at every operating point, graph degree, and index scale; at several points the clock is better. We trace our earlier positive result to an interpolated baseline: recall is sharply concave in repair budget -- one consolidation captures over half of all achievable gain -- so reading the baseline off a straight line between zero and four passes understates it by more than the effect claimed. Evaluated by the statistic we pre-registered -- correlation with the subsequent recall drop rather than with the concurrent recall level -- the probe signal is also not a leading indicator. What remains is useful: an exact live-set recall oracle, a reproducible churn harness, a drift-severity regime map, and a budget-parity protocol that makes this error detectable. We report the negative result and the trap, because the trap generalizes: any "at matched budget X" comparison whose baseline is read off an interpolated curve, at the scarce end of a concave response, will manufacture an effect favouring the proposal.
arXiv:2607.02134v2 Announce Type: replace
Abstract: Scientific machine learning papers typically make computational claims, e.g., that the relative mean square error is less than 5% or that the 95% predictive credible interval covers the test data. A coding agent can be prompted to replicate those claims from paper materials alone, but the prompt does not by itself reliably preserve progress or check whether generated evidence supports the paper's claims. We introduce Paper-replication, a workflow that makes each selected paper claim a target with recorded evidence, and implement it as a coding-agent skill. The workflow makes the agent record those targets, reconstruct the paper's method, run computational experiments, link generated outputs to provenance and comparisons with the paper's claims, record where matched evidence appears in the replication report, and pass validation checks before completion. We evaluate Paper-replication on twelve independent runs across four scientific machine learning papers. All twelve workspaces pass the completion gate, and all 158 recorded targets are matched with report coverage. Even in this completed workspace state, repeated runs differ in how papers are divided into targets, in numerical fidelity to the source papers, in elapsed replication time, in the number of intermediate executions replaced before final evidence is accepted, and in the rules used to accept evidence. Paper-replication makes completion depend on workspace evidence and validation checks rather than on the agent's final message.
arXiv:2607.08936v1 Announce Type: new
Abstract: We present a tensor-train discontinuous Galerkin (TT-DG) formulation for the Vlasov--Maxwell system that combines a modal DG discretization with low-rank tensor representations of the phase-space solution and discrete operators. The formulation exploits the tensor-product structure of the DG discretization to perform quadrature, differentiation, nonlinear upwind flux evaluation, and time integration directly in compressed form.
The method is evaluated on several standard 1D2V Vlasov--Maxwell benchmark problems, including the streaming Weibel instability, weak Landau damping, and two-stream instability problems. Across these problems, the TT formulation reproduces the accuracy and conservation behavior of the underlying full-grid DG discretization while substantially reducing memory usage and runtime. For weakly nonlinear problems, compression ratios exceeding $10^4$ are obtained together with significant speedups relative to the full-grid solver. For the strongly nonlinear two-stream instability problem, the TT formulation remains effective despite reduced compressibility caused by fine-scale phase-space filamentation.
These results demonstrate that tensor-train representations provide an effective approach for reducing the computational cost of deterministic DG-based kinetic plasma simulations while retaining the favorable numerical properties of the underlying discretization.