arXiv:2603.22006v2 Announce Type: replace-cross
Abstract: Upcoming stage-IV surveys such as Euclid and Rubin will deliver vast amounts of high-precision data, opening new opportunities to constrain cosmological models with unprecedented accuracy. A key step in this process is the reconstruction of the dark matter distribution from noisy weak-lensing shear measurements. Current deep-learning-based mass-mapping methods achieve high reconstruction accuracy, but either require retraining a model for each new observed sky region (limiting practicality) or rely on slow Markov chain Monte Carlo sampling. Efficient exploitation of future survey data therefore calls for a new method that is accurate, flexible, and fast at inference. In addition, an uncertainty quantification with coverage guarantees is essential for a reliable cosmological parameter estimation. We introduce PnPMass, a plug-and-play approach for weak-lensing mass mapping. The algorithm produces point estimates by alternating between a gradient descent step with a carefully chosen data fidelity term and a denoising step implemented with a single deep-learning model trained on simulated data corrupted by Gaussian white noise. We also propose a fast sampling-free uncertainty quantification scheme based on moment networks, with calibrated error bars obtained through conformal prediction to ensure coverage guarantees. Finally, we benchmark PnPMass against model-driven and data-driven mass-mapping techniques. PnPMass achieves a performance close to that of the currently best deep-learning methods while offering fast inference. It converges in just a few iterations, and it requires only a single training phase, regardless of the noise covariance of the observations. It therefore combines flexibility, efficiency, and reconstruction accuracy while delivering tighter error bars than existing approaches, making it well suited for upcoming weak-lensing surveys.
Science Journals
arXiv:2603.16103v2 Announce Type: replace
Abstract: 3D Gaussian Splat (3DGS) enables high-fidelity, real-time novel view synthesis by representing scenes with large sets of anisotropic primitives, but often requires millions of Splats, incurring significant storage and transmission costs. Most existing compression methods rely on GPU-intensive post-training optimization with calibrated images, limiting practical deployment. We introduce \textbf{NanoGS}, a training-free and lightweight framework for Gaussian Splat simplification. Instead of relying on image-based rendering supervision, NanoGS formulates simplification as local pairwise merging over a sparse spatial graph. The method approximates a pair of Gaussians with a single primitive using mass preserved moment matching and evaluates merge quality through a principled merge cost between the original mixture and its approximation. By restricting merge candidates to local neighborhoods and selecting compatible pairs efficiently, NanoGS produces compact Gaussian representations while preserving scene structure and appearance. NanoGS operates directly on existing Gaussian Splat models, runs efficiently on CPU, and preserves the standard 3DGS parameterization, enabling seamless integration with existing rendering pipelines. Experiments demonstrate that NanoGS substantially reduces primitive count while maintaining high rendering fidelity, providing an efficient and practical solution for Gaussian Splat simplification. Our project website is available at \href{https://saliteta.github.io/NanoGS/}{https://saliteta.github.io/NanoGS/}.
arXiv:2607.13610v1 Announce Type: new
Abstract: Large-scale gas-network scenario evaluation is a computational bottleneck in integrated energy-system planning, particularly when gas infrastructure interacts with power, heat, hydrogen, and sector-coupling pathways. Conventional nonlinear hydraulic solvers provide reliable feasibility assessment but are costly for stochastic screening, whereas unconstrained learning-based surrogates may produce hydraulically infeasible states. This study develops a physics-informed graph neural network surrogate for steady-state gas-network simulation and feasibility screening. The model uses an edge-centric architecture to predict pipe-level squared-pressure differences and flows. A differentiable projection layer enforces nodal mass conservation on predicted flows, while a Laplacian reconstruction maps edge pressure differences to topologically consistent nodal pressures.
The framework is evaluated on GasLib-134, GasLib-135, and GasLib-582 using stochastically generated operating scenarios. On the meshed 582-node benchmark trained with 5000 scenarios, the surrogate achieves a pressure mean absolute error of 1.05~bar, corresponding to 1.3\% of the realized pressure range, with $R^2 = 0.981$. Projected-flow predictions reach $R^2 = 0.972$, and mass-balance residuals are reduced to numerical precision, on the order of $10^{-5}$--$10^{-4}$~Nm$^3$/s. Compared with the MYNTS reference solver, inference is reduced from seconds to milliseconds, with the largest benchmark evaluated in less than 40~ms. Loadability and out-of-distribution stress-test evaluations demonstrate robust feasibility screening under high-load conditions, while strongly localized demand concentrations are identified as cases requiring solver-based verification near feasibility limits. The framework provides a physically constrained planning accelerator for high-volume scenario screening and prioritization.
arXiv:2607.13737v1 Announce Type: new
Abstract: For low-data and resource-constrained regimes typical of quantum chemistry, parameter-efficient learning is a key objective. Here, we propose a topology-aligned inductive bias in which the model architecture mirrors the molecular bond graph: atoms map to a fixed register of computational units, and bonds determine which pairs interact through shared learnable parameters. This principle is instantiated in two architectures: a variational quantum circuit (Iso-QGNN), and a parameter-matched classical message-passing model (Iso-CGNN). The models are benchmarked on HOMO-LUMO and dipole moment binary classification tasks over the QM9 benchmark. With 64 trainable parameters, the implementations achieve test AUCs of approximately 0.88 (quantum) and 0.91 (classical) on the gap task, and close to 0.78 (both) on the dipole task. The models reach 90% of asymptotic performance within about 250 training molecules and gradient norms remain stable throughout training. These results indicate that the topology-aligned inductive bias is the active ingredient driving parameter efficiency at QM9 scale, with implications for matched-baseline benchmarking in quantum machine learning.
arXiv:2607.13612v1 Announce Type: new
Abstract: Joint-Embedding Predictive Architectures (JEPAs) are the dominant design for latent world models, yet they are usually justified by empirical performance rather than a normative principle. We show that the choice of anti-collapse regulariser determines whether a JEPA's training objective, a prediction loss plus a weighted embedding regulariser, is a valid Active Inference (AIF) variational free energy. We organise four non-contrastive regularisers (VICReg, LogDet, PairDist, and SIGReg) into an entropy-estimator hierarchy indexed by a prior-miscalibration gap, and show that the gap's sign, whether the estimator bounds the latent entropy from above or below, decides whether the AIF surprise bound survives: VICReg and LogDet are unsafe upper bounds, PairDist a safe lower bound, and SIGReg eliminates the gap. We then prove a correspondence theorem: under the standard constant-noise encoder model and successful SIGReg enforcement (isotropic-Gaussian embeddings), the gap vanishes, the objective becomes an exact information bottleneck, the surprise bound is preserved, and the latent goal cost becomes an exact proxy for AIF pragmatic value, whereas VICReg leaves an irreducible second-order anisotropy term. We extend the correspondence to multi-step expected free energy, ensemble epistemic value, and a learned-policy regime, and we identify the one AIF term no current JEPA world model computes: the state-epistemic value, a future-state coverage signal. The predictions differ in kind, not degree, and are stated here as theoretical consequences left for empirical test in separate work; full proofs are in Appendix A, and the algebraic core of every result is machine-verified in Lean 4 (Appendix D).
arXiv:2607.13869v1 Announce Type: new
Abstract: To show clinical feasibility of a previously proposed probabilistic planning approach that can precisely optimize for clinical goals with patient-specific acceptance probabilities on a neuro-oncological patient group, we compared probabilistic plans with (automated) robust plans for one patient (group A) that could achieve sufficient clinical target coverage and for four patients (group B) where target coverage had to be compromised due to organ-at-risk (OAR) dose constraints. The probabilistic approach is percentile-based and uses the fact that a (dose) percentile can be approximated as a linear combination of its expected value and standard deviation. The optimization has a nested structure: the inner optimization optimizes the beam weights for a given percentile estimate, while an outer loop iteratively updates and improves the accuracy of the percentile estimate. For every outer iteration, the optimization is warm-started from the previous iteration. Percentiles are efficiently calculated by sampling a polynomial chaos expansion of the dose-influence matrix. The patient in group A achieved cumulative OAR dose reductions (of OAR-related DVH-metrics) of 19 GyRBE, for identical target coverage. Target coverage improved for all patients in group B (the 10th percentile of $D_{99.8\%}$ increased up to 0.93 GyRBE), at the same time reaching cumulative OAR dose reductions (of OAR-related DVH-metrics) up to 33 GyRBE. Probabilistic plans were optimized in 44h to 141h. For two representative patients, eliminating warm-starting (i.e., the outer loop) from the approach reduced total optimization times to below 10h (which took originally 80h and 141h). Compared to robust optimization methods, the probabilistic approach achieves improved trade-offs between probabilistic target coverage and OAR sparing, potentially leading to better treatments.
arXiv:2607.13062v1 Announce Type: cross
Abstract: Speculative window decoders hide quantum error-correction decoder latency by guessing the cross-boundary decisions that link adjacent decoding windows, running downstream work on the guess, and verifying lazily. SWIPER and ARTERY each build one predictor, about 90% accurate; neither built the verifier side. We build it on a reconstructed SWIPER harness (Stim rotated surface code, minimum-weight matching). A predictor-only bracket shows the cross-boundary decision is local, the achievable accuracy reaching about 0.999 within three rounds, with small, diffuse headroom over SWIPER. We establish a worst-case temporal blast-radius bound, its probability core machine-checked in Lean4 and conditional on a modeling reduction we then test: a misprediction's effect decays exponentially in the commit width, so the radius is one and speculation adds no error floor. We falsify that reduction shot by shot and find the real mechanism, clearest at near-threshold noise, is a global minimum-weight re-pairing. A compiler pass derives SWIPER's restart policy from these numbers; a runtime executor confirms on the harness that the loop recovers exactly and removes the serial commit-chain stall up to a small penalty. A second decoder (union-find) settles which results are decoder-agnostic: the predict-verify-recover wrapper and the structural phenomenology, while the absolute magnitudes and the min-weight mechanism are matching-specific.
arXiv:2607.13550v1 Announce Type: cross
Abstract: Boosting is one of the most successful learning techniques for standard classification and regression tasks. Its extension to multi-output prediction problems has found an increasing number of applications in recent years. Among them is the prediction of entire conditional distributions rather than single functionals, which can often be framed as a multi-output regression problem, for example multiple quantile regression. Addressing such problems with classical implementations of boosting is computationally challenging, because usually one base model is trained for each target at every iteration. More efficient variants of boosting have been proposed to speed up training, but they tend to be tied to specific loss functions and classes of base learners, usually decision trees. In this work, we study a modification of the gradient boosting algorithm, which we call parallel gradient boosting, designed to circumvent all these limitations. The core idea is to use a common descent direction for all training observations. By doing so, only one base model is needed at each iteration, regardless of the number of targets, which allows for considerable performance gains. We establish sufficient conditions for the convergence of the algorithm, whose practical use is introduced via the multiple quantile regression setting. We show that in such a setting, it provides predictions of similar quality to state-of-the-art boosting libraries such as XGBoost, while being faster by several orders of magnitude. Then, we evaluate the properties of the resulting conditional distribution estimator, which is shown empirically to outperform other nonparametric and semiparametric estimators, especially in high-dimensional settings and in the presence of mixed and/or missing covariates.
Genre Bias or Aesthetic Perception? Identifying and Mitigating Shortcut Learning in Music Evaluation
arXiv:2607.13903v1 Announce Type: new
Abstract: Music aesthetics scoring plays a critical role in applications such as dataset curation, generative model evaluation, and reward modeling for music generation. Recent approaches rely on deep neural networks trained on human-annotated ratings, but these models may exploit spurious correlations rather than capturing perceptually meaningful aesthetics. In this work, we identify a previously underexplored failure mode in music evaluation models: genre-induced shortcut learning. Through a systematic analysis of SongEval, we show that biases in training data lead to strong correlations between genre-related features and predicted scores, causing the model to use them as a proxy for aesthetics. This results in systematic overestimation of pop music and undervaluation of high-quality samples from other genres, leading to predictions that are inconsistent with human preferences. To address this issue, we propose a training objective that jointly reweights hard samples and regularizes group-level performance, encouraging the model to learn genre-invariant representations of musicality. Experimental results demonstrate that our method reduces genre-dependent bias and improves alignment with human preferences, as reflected by gains in both cross-genre and within-genre preference alignment.
arXiv:2607.13651v1 Announce Type: new
Abstract: The bottleneck of Earth Observation processing chains is not the arrival of new imagery but whether the surface is actually visible when the image arrives. We study this as an observability forecasting problem on EarthNet2021. Given recent multispectral imagery and exogenous weather drivers, the goal is to predict whether the next acquisition will be usable and, if not, when a usable view is likely to return. To do this, we adapt LeWorldModel, a joint-embedding predictive architecture world model, to cloud-aware Earth Observation sequences. The final pipeline converts raw minicubes into episodic HDF5 sequences with five image channels (blue, green, red, near-infrared, cloud mask) and eight meteorological and calendar covariates. The resulting model has 18.0M trainable parameters and is trained from scratch on 23,904 training episodes. The trained leWorldModel is evaluated under a locked protocol: linear probes are fit on train only, calibration choices are set on an internal validation split, and the fitted heads are then frozen for valsplit, IID, OOD, and extreme evaluation. On the full frozen-bundle observability benchmark, LeWorldModel consistently outperforms persistence. For next-step usability, balanced accuracy ranges from 0.769 to 0.887, compared with 0.493 to 0.556 for persistence. For exact first-usable-horizon prediction, accuracy ranges from 0.602 to 0.806, compared with 0.120 to 0.369 for persistence. Against a frozen LightGBM baseline fit on the same training windows, LeWorldModel is better on continuous clear/cloud regression and on exact recovery timing on valsplit, IID, and extreme, while LightGBM is stronger on the simpler binary any-usable-within-six task and is more robust on OOD. In separate sampled diagnostic analyses, LeWM also produces strong ranking-based anomaly signals under synthetic temporal inconsistencies.
arXiv:2607.13905v1 Announce Type: new
Abstract: The International StepUP Competition Series was launched to advance research in pressure-based footstep biometrics through a standardized and challenging evaluation framework. Using the large-scale StepUP-P150 dataset (with more than 200,000 high-resolution dynamic footsteps from 150 individuals) and a previously unreleased test set, the 2nd edition of the competition addressed three key challenges: (1) generalization to unseen users with limited enrollment data, (2) robustness to domain shift caused by variations in footwear and walking speed and (3) effective fusion of paired left-right footsteps. While the first two challenges built on the inaugural competition, this edition introduced more extreme cross-domain conditions and moved beyond isolated footsteps to stride-level verification, enabling new opportunities for representation learning and inter-step information fusion. The competition attracted 26 registrants from academia and industry, with a best equal error rate of 8.00% achieved by the ArogyaPandit Research Team using a spatiotemporal CNN combined with an ensemble-based scoring strategy. The top solutions showcase the value of harnessing temporal patterns and of incorporating inference-time normalization and calibration strategies to improve scoring. However, the results also reveal that recognizing users in unseen personal footwear remains a challenge, especially in the presence of distractors with similar characteristics.
arXiv:2606.23531v4 Announce Type: replace
Abstract: Endoscopic retrograde cholangiopancreatography (ERCP) demands precise endoscopic navigation and stable biliary cannulation within a narrow monocular field characterized by specular reflections, partial occlusions, and frequent tissue contact. Although recent robotic systems and vision-based assistance techniques improve operator ergonomics and provide perceptual cues, their performance degrades under pronounced anatomical variability and safety-critical visual artifacts, which hinders reliable autonomy in cannulation-grade procedures. Here, we present BiliVLA, a scene-aware Vision-Language-Action (VLA) framework that formulates biliary endoscopic navigation as an instruction-conditioned visuomotor learning problem. Given an endoscopic observation and a stage-specific language instruction, BiliVLA jointly predicts the target category, a grounded bounding box, and a discrete three-degree-of-freedom (3-DoF) motor command for a continuum endoscope. The proposed framework incorporates scene-aware supervision to improve semantic target consistency and safety-aware recovery supervision to induce conservative retreat behaviors under luminal wall contact. A key component of BiliVLA is a two-stage training paradigm that combines grounding-enhanced supervised fine-tuning (SFT) with Group Relative Policy Optimization (GRPO), thereby improving action reliability and decision consistency during closed-loop navigation. Across three ERCP subtasks, BiliVLA achieves the best overall performance in physical phantom experiments, with a total mIoU of 0.9625, an overall action precision of 91.96\%, and an overall success rate (SR) of 84.85\%. These results indicate that integrating semantic grounding, scene-aware learning, and reward-guided optimization strengthens perception--action alignment and enables more robust autonomous biliary endoscopic navigation.
arXiv:2607.13898v1 Announce Type: new
Abstract: Modern deep learning workloads increasingly rely on narrow numerical formats to improve efficiency and reduce memory footprint. The recently standardized microscaling floating-point (MXFP) family of formats, including MXFP8, MXFP6, and MXFP4, offers a practical approach to low-precision inference, yet the digital signal processing (DSP) blocks in current FPGA architectures offer limited native support for these formats. In this work, we first present a comprehensive characterization of MXFP dot product implementations on Altera Agilex-5 FPGAs, exploring a range of strategies spanning pure soft logic, DSP blocks in fixed-point, floating-point, and tensor modes. Our results show that while the tensor mode delivers the highest arithmetic density for MXFP4 (E2M1) and MXFP6 (E2M3), it cannot implement MXFP6 (E3M2) or any MXFP8 precisions, forcing designers to fall back to lower-density alternatives. Motivated by this gap, we propose targeted modifications to the DSP block's internal tensor-mode architecture that enable native support for all MXFP precisions while retaining backward compatibility. We estimate the area cost of these modifications using a simplified version of the Agilex-5 DSP block core implemented using the open-source ASAP7 PDK. We evaluate a variety of modified DSP block designs that present a tradeoff between format coverage, arithmetic density, and area overhead. Our preferred design point increases the DSP tile area by 36%, corresponding to only 1.8\% of the total FPGA die area. We evaluate the device-level impact of our enhanced DSP block by comparing systolic array matrix multiplier implementations across all MXFP precisions, contrasting the best-available strategies on the existing architecture against designs leveraging our modified DSP block. Our results demonstrate an average throughput improvement of 4.2x across all supported MXFP formats.
arXiv:2607.13928v1 Announce Type: new
Abstract: Whistleblowers are a key safeguard against organizational wrongdoing, but the threat of retaliation deters reporting. Existing whistleblower-protection proposals lack formal privacy guarantees, and existing differential privacy mechanisms do not directly target the natural threat model -- one in which the audited organization itself observes auditor selection decisions and uses them to identify reporters. We formalize protection against a strong-adversary threat model as per-report $(0, \delta)$-differential privacy on the transcript of audit selections. Within this framework we prove that a natural approach -- randomized response applied at the selection step -- can never outperform uniform random auditing by more than $\delta$ at any horizon. We then give a generic mechanism that reduces private auditing to private continual counting: any $(0, \delta)$-DP continual counter plugs in by post-processing, and the audit transcript inherits the same per-report guarantee. Instantiating the reduction with a recent work in continual counting yields per-report $(0, \delta)$-DP with noise scaling as $O(\sqrt{\log T})$ across a horizon of $T$ audit decisions. A utility theorem shows that the selection error vanishes whenever the noisy report gap between the most-reported organization and the runner-up grows faster than $\sqrt{\log T}$. Simulations show a substantial improvement over randomized response.
arXiv:2606.26738v2 Announce Type: replace
Abstract: While recent advancements in generative image editing models have achieved stunning visual fidelity, it remains an open question whether these systems possess an intrinsic knowledge of real-world lighting. Existing benchmarks typically evaluate high-level plausibility of perceptual light transport on curated internet imagery, using VLMs or human judgement, or they rely on synthetically generated datasets. In this work, we introduce the 3D-anchored Light Probe (3DLP) benchmark, for which we have captured a new high-fidelity HDR dataset of real-world lighting changes. The dataset consists of 1K image pairs of diverse indoor scenery in which light probes are physically turned on and off. To allow for a granular performance analysis, we annotated specific image regions such as cast shadows or metallic surfaces. With this data, we evaluate a range of state-of-the-art image editing models by measuring how well their light probe edits align with reality. The evaluation uses two new scores to compensate for AI-generated photographic effects, such as adjusted white balance. Our results show that the overall performance of models differs considerably, with differences slightly less pronounced for specular highlights. The best image editing models are remarkably consistent with real-world physics, however, they still leave room for improvement. We observe that image regions that receive less light from the light probe are more prone to errors for all models. Furthermore, building on their success in evaluating macroscopic lighting plausibility, we test VLMs on our task but find that they are unsuitable for pixel-level light transport analysis. We will make the benchmark, together with the real-world dataset, publicly available to encourage future research on this topic.
arXiv:2607.13335v1 Announce Type: cross
Abstract: We study the deterministic query complexity of minimizing a convex Lipschitz function over a $d$-dimensional Euclidean ball using only exact function values. At accuracy $\Theta(d^{-1/2})$, the previously applicable lower bound was $\Omega(d)$, inherited from the stronger full first-order oracle, while an upper bound from Protasov's value-only method requires $O(d^2\log^2 d)$ evaluations. By providing a lower bound of $\Omega(\,\frac{d^2}{\log(d+1)})$ on the oracle complexity in this setting, we thereby close this gap dating back to 1996, up to polylogarithmic factors. Furthermore, we are able to lift this result to the mixed-integer setting: Mixed-integer convex optimization with $d$ continuous and $n$ discrete variables using function values requires $\tilde{\Omega}(d^2\cdot 2^n)$ queries.
arXiv:2607.13401v1 Announce Type: cross
Abstract: We prove cutoff in total variation distance for the discrete-time Kac walk on $S^{n-1}$ started from a coordinate vector. The cutoff occurs at $
C_{\mathrm{BRW}}n\log n$, where $C_{\mathrm{BRW}} \approx 3.8916$ is an explicit constant determined by the speed of the leftmost particle in a branching random walk. In particular, the cutoff location is not at the conjectured time $ 2n\log n$.
arXiv:2511.03573v2 Announce Type: replace
Abstract: We report on the radiation tolerance characterisation of a p-type silicon pad sensor indigenously designed by BARC-VECC and fabricated at Bharat Electronics Limited (BEL), India -- representing a significant step toward establishing a domestic silicon sensor manufacturing capability for high-energy physics applications. Single-pad test structures were irradiated with neutrons over a range of fluences from $\sim10^{7}$ to $\sim2.5\times10^{14}$~1~MeV~$n_{\mathrm{eq}}$/cm$^{2}$, spanning the operational regime relevant to forward calorimetry in high-luminosity heavy-ion collider experiments. Post-irradiation performance was characterised through systematic measurement of leakage current evolution and calorimetric response as functions of accumulated neutron fluence. A single-exponential annealing model is introduced to describe the time dependence of leakage current within the observation window, and the current-related damage constant $\alpha$ is extracted and compared with the RD48 reference value. The results demonstrate that the sensors survive the target fluence with measurable but recoverable degradation, validating the fabrication process and providing a baseline for future qualification of this indigenous sensor production chain.
arXiv:2607.13414v1 Announce Type: cross
Abstract: Non-expansive two-time-scale stochastic approximation is governed by a slow stochastic Krasnoselskii--Mann fixed-point iteration rather than by contraction to a unique equilibrium. We study this regime under a contractive fast map and a non-expansive reduced slow map. We first prove a finite-horizon lower bound showing that, for any prescribed slow stepsize schedule $(\beta_k)$, the classical KM residual scale $(\sum_{i<N}\beta_i(1-\beta_i))^{-1}$ is worst-case sharp for the corresponding unregularized KM update. Combined with the raw fast-tracking leakage scale, this explains the previously observed $k^{-1/4+o(1)}$ last-iterate mean-square residual exponent.
We then introduce a residual-preconditioned slow oracle that cancels the first-order dependence on the fast tracking error. In a nested Tikhonov-KM algorithm, the uncorrected oracle yields total-sample rate $T^{-1/4+o(1)}$, while the corrected oracle yields $T^{-1/3+o(1)}$. This improvement comes from changing the slow-oracle bias from first order to second order in the fast error after all inner-loop samples are counted.
Finally, we show that the repeated inner-loop cost of the nested method can be avoided in a smooth derivative-oracle model. A single-loop algorithm that tracks both the fast equilibrium and the leakage preconditioner online achieves $T^{-1/2+o(1)}$ with $O(1)$ primitive samples per iteration.
arXiv:2603.25237v2 Announce Type: replace
Abstract: In complex molecular systems, the reaction coordinate (RC) that characterizes transition pathways is essential to understand underlying molecular mechanisms. This review surveys a framework for identifying the RC by applying deep learning to the committor, which provides the most reliable measure of the progress along a transition path. The inputs to the neural network are collective variables (CVs) expressed as functions of atomic coordinates of the system, and the corresponding RC is predicted as the output by training the network on the committor as the learning target. Because deep learning models typically operate in a black-box manner, it is difficult to determine which input variables govern the predictions. The incorporation of eXplainable Artificial Intelligence (XAI) techniques enables quantitative assessment of the contributions of individual input variables to the predictions. This approach allows the identification of CVs that play dominant roles and demonstrates that the committor distribution on the surface using important CVs is separated by well-defined boundaries. The framework provides an explainable deep learning strategy for assigning a molecular mechanism from the RC and is applicable to a wide range of complex molecular systems.
arXiv:2607.13555v1 Announce Type: cross
Abstract: Bioacoustic call-type classification relies on costly expert annotation. Active learning can reduce this burden by selecting a small batch of segments for expert annotation and using the labeled segments for training the classifier. The setting is hard: the target calls are extremely sparse and the call-type distribution is long-tailed, so a tight budget must be spent on the few rare, informative segments. We propose BADGE-Greedy-DPP, a deterministic batch selector that greedily adds the segment whose BADGE gradient embedding most enlarges the volume spanned by the batch; because this log-volume objective is submodular, the greedy rule guarantees a batch value at least a (1-1/e) fraction of the optimum of this objective, a guarantee not provided by BADGE's existing k-means++ and MCMC DPP sampling heuristics. There is also a temporal granularity mismatch in the task. The acquisition function scores whole segments, yet the informative frames inside them are few. Uniform averaging therefore washes them out. We show that the BADGE construction naturally addresses this mismatch when applied frame-wise, as prediction residuals weight the aggregated pseudo-gradient, so confidently predicted no-call frames contribute little while a single uncertain rare-call frame can still set the segment's direction. Across 10 runs on a sparse, imbalanced hyena call-type dataset, BADGE-Greedy-DPP achieves the best overall and rare-call-type performance among all compared query strategies, including MFFT, the strongest non-BADGE baseline, and the two vanilla BADGE traversals.
arXiv:2603.14473v2 Announce Type: replace
Abstract: Scientific discovery depends on expert judgement and foresight, which we call scientific taste: the ability to judge and propose research ideas with potential for long-term scientific impact. Whether AI can learn this ability remains an open question. Here we provide evidence that artificial intelligence can learn judgement and ideation. We introduce Reinforcement Learning from Community Feedback (RLCF), a training paradigm that uses large-scale signals from scientific community as supervision. We first train Scientific Judge on field- and time-matched pairs of high- vs. low-citation papers to judge ideas. We then train a Scientific Thinker, to propose research ideas with high potential impact. Experiments show that the 30B Scientific Judge variant outperforms strong LLM baselines (e.g., GPT-5.4 Thinking), while Scientific Judge generalizes across future-year papers, unseen fields, and other community metrics. Furthermore, Scientific Thinker proposes research ideas with higher potential impact than baselines. These results suggest that AI can learn scientific taste, marking an important step towards AI systems that could help accelerate scientific discovery.
arXiv:2603.14604v2 Announce Type: replace
Abstract: We propose TacFiLM, a lightweight modality-fusion approach that integrates visual-tactile signals into vision-language-action (VLA) models. While advances in VLAs have introduced robot policies that are both generalizable and semantically grounded, these models mainly rely on vision-based perception. Vision alone, however, cannot capture the complex interaction dynamics that occur during contact-rich manipulation, including contact forces, surface friction, compliance, and shear. While recent attempts to integrate tactile signals into VLA models often increase complexity through token concatenation or large-scale pretraining, the heavy computational demands of behaviour models necessitate lightweight fusion strategies. To address these challenges, TacFiLM outlines a post-training finetuning approach that conditions intermediate visual features on pretrained tactile representations using feature-wise linear modulation (FiLM). Experimental results on insertion and drawer opening tasks demonstrate consistent improvements in success rate, direct task performance, completion time, and force stability across both in-distribution and out-of-distribution tasks. Together, these results support our method as an effective approach to integrating tactile signals into VLA models, improving contact-rich manipulation behaviours. Project page: https://charliem7.github.io/projects/TacFilm/
arXiv:2607.13560v1 Announce Type: cross
Abstract: Recent advances in generative and embodied AI have been driven by large-scale predictive learning over multimodal data. However, the resulting systems remain largely based on passive training regimes where linguistic regularities create the scaffold onto which information from other modalities is attached. Conversely, neuroscience and cognitive science suggest that biological intelligence is organized in the opposite way, where grounded world models acquired through interaction with the environment provide the semantic scaffold to which language is attached. Here, we illustrate five examples of neural circuits supporting grounded world modelling, which underlie navigation in physical and conceptual spaces, affordance-based perception and interaction with objects, active perception and exploratory learning, allostatic control and emotion, and the distinction between self- and world-generated outcomes. These examples highlight several features largely missing from current embodied AI, including the role of intrinsic dynamics as a foundation for learning, the centrality of action in aligning these dynamics with the external world, the prominence of autonomous experience and open-ended learning over passive assimilation of externally provided data, and the fact that early predictive and control mechanisms scaffold higher cognitive abilities such as reasoning, conceptual navigation, planning, imagination, understanding others' minds, and communication. Finally, we discuss whether and how principles derived from biological systems may inform future embodied AI, including training regimes based on social interaction to construct world models that are not only grounded but also socially shared and aligned with human norms and values.
arXiv:2607.13913v1 Announce Type: cross
Abstract: We present long-baseline Very Low Frequency (VLF) observations of solar flare-induced ionospheric disturbances obtained at the Bulgarian Polar Astronomical Observatory (St. Kliment Ohridski Base) on Livingston island, Antarctica. Using continuous VLF transmissions at 21.4 kHz (NPM, Hawaii) and 24.0 kHz (NAA, Maine), propagating over trans-hemispheric paths exceeding 11000 km, we analyze observations of solar flares during the period 24 January--8 February 2025. After removing the strong diurnal signal via superposed epoch analysis, we analyse the flare-related perturbations in VLF amplitude and their correlation with GOES soft X-ray flux for 250 flares of C and M class. The long propagation paths provide enhanced sensitivity to flare-driven changes in D-region ionization. The observations reveal clear, frequency-dependent responses and measurable time delays between X-ray and VLF peaks. These delays, including cases of near-zero or negative lag for stronger events, highlight the role of flare spectral characteristics and D-region recombination processes. Our results demonstrate the scientific value of long-baseline Antarctic VLF observations for detecting and timing GOES-class flares, while also highlighting key limitations, such as background variability and path-dependent propagation effects, which must be quantified for reliable VLF-based flare monitoring.