arXiv:2606.00096v2 Announce Type: replace
Abstract: Visual agents employ external visual tools within visual chains of thought to incorporate fine-grained evidence. While prior work has mainly studied these tools in visual search tasks, their role in more complex visual reasoning remains underexplored. In this paper, we move beyond simple visual search tasks to investigate more challenging tasks, including 3D spatial reasoning and medical visual question answering, where agents must integrate tool-acquired local evidence with the global context. We identify a {tool-use collapse phenomenon: models progressively stop using tools while still achieving higher task accuracy. Moreover, we observe a clear asymmetry: (i) completely eliminating tool use degrades performance, whereas (ii) incentivizing tool use yields only marginal gains despite substantially increasing usage. We find that vanilla training and tool-use encouragement both reduce rollout diversity, explaining why higher tool use does not yield stronger reasoning performance. Motivated by these findings, we add an entropy regularization term to encourage diverse rollout exploration, achieving the best performance despite gradually declining tool usage. Overall, our findings suggest a training-time view of tools as scaffolding, where broader exploration over language generation and visual tool invocation improves reasoning despite tool-use collapse. Project page: https://scaffolded-exploration.github.io
Science Journals
arXiv:2606.03097v1 Announce Type: new
Abstract: Incorporating news into time series forecasting is appealing because news can reveal abrupt exogenous events that historical values alone cannot recover. However, existing LLM-based news-forecasting pipelines face two practical limitations: relevant news articles often exceed the model's context window, and iterative retrieval of supplementary news is typically unguided, leading to redundant updates and slow convergence. We address these issues with a novel framework that combines importance-aware news compression and process-level retrieval supervision. First, we train an importance reward model that estimates the forecasting utility of each article and uses this signal to allocate compression budgets during sequential pairwise fusion, preserving informative content within a fixed context limit. Second, we introduce a process reward model (PRM) that ranks multiple supplementary-news candidates conditioned on the current error profile and the history of previously selected articles, replacing one-shot blind retrieval with quality-controlled selection. Both components are trained offline using historical data with ground truth; inference uses the frozen filtering logic and compression modules without any reflection loop. Experiments on finance, energy, traffic, and bitcoin forecasting benchmarks show that our method improves prediction accuracy over strong baselines, significantly reduces the number of refinement iterations compared to the iterative baseline, and remains effective when relevant articles span thousands of tokens.
arXiv:2606.03059v1 Announce Type: new
Abstract: Microchannel flow boiling is a promising technique for micro-device thermal management, and understanding the bubble dynamics in microchannel flow boiling is important for the applications. Previous studies only focused on single, isolated bubbles, but the bubbles in microchannel flow boiling applications often exist as bubble trains, in which the bubbles interact with each other. Here, we investigate numerically vapor bubble trains in microchannel flow boiling by adopting the flow-focusing technique to form monodispersed bubbles in the upstream of the microchannel. With increasing the initial vapor-liquid volume ratio, the bubble frequency increases while the growth rate of the bubbles decreases because of the reduced bubble size. With increasing the heat flux on the wall or reducing the latent heat of the working fluid, the bubble train growth rate increases because of the increased vaporization rate. The vaporization of the fluid in the upstream causes the bubble expansion and accelerates the bubble movement in the downstream. The wall temperature and the Nusselt number fluctuate because of the periodic pass-through of bubbles.
arXiv:2606.02996v1 Announce Type: new
Abstract: Inertial odometry (IO) using only Inertial Measurement Units (IMUs) provides a lightweight solution for human motion tracking in augmented reality (AR) and wearable devices. Recent learning-based IO methods have improved the generalizability of inertial localization through large-scale pretraining on human motion datasets. However, these approaches remain prone to drift and noise because they do not explicitly capture human motion dynamics, especially on daily activity datasets such as Nymeria. In this work, we propose to ground inertial odometry in human kinematics through a learned IMU-inferred pose prior, which promotes physically consistent motion constraints. We integrate this pose prior into existing IO architectures and reduce positional drift by up to 36% on the challenging Nymeria dataset, which is 5x larger than datasets used in prior work. We further improve long-term performance with a sensor-fusion framework that incorporates auxiliary signals from lightweight sensors already available on commercial AR glasses, including magnetometers, barometers, and secondary IMUs. With this fusion strategy, positional drift is reduced by up to 42%, improving robustness and generalization across diverse motion conditions. Together, our results introduce a new paradigm for inertial and lightweight odometry by unifying human motion kinematics with multimodal sensing, setting a new benchmark for accurate and robust camera-less human tracking. Our website is available at https://spice-lab.org/projects/MARIO/.
arXiv:2606.03962v1 Announce Type: new
Abstract: Classical reinforcement learning (RL) typically seeks a deterministic policy that maximizes the expected sum of a scalar reward. Yet, modern applications such as language model fine-tuning or scientific discovery demand diversity. Existing remedies such as entropy regularization or diversity bonuses often require fragile trade-offs that sacrifice performance for stochasticity or rely on heuristic metrics that can misalign policy rankings. We argue that diversity is more naturally understood as the rational response to uncertainty in the reward. When the reward function is not perfectly known--as is the case with ambiguous preferences or imperfect reward models--committing to a single action can be sub-optimal. Building on this, we propose a fundamental reformulation of the RL objective by replacing the scalar reward with a distribution over reward functions, and applying a non-linear objective over sets of actions. The result is a framework in which calibrated behavioural diversity emerges naturally, remains controllable through the reward function distribution, and is obtained without sacrificing expected reward. Focusing on the contextual bandit setting, we derive a principled gradient estimator for this objective and prove that our formulation naturally generalizes both vanilla policy gradient and more recently developed action-set approaches. Our empirical results demonstrate that this framework offers a robust and theoretically grounded alternative for complex RL tasks where the traditional formulation of the problem fails to induce the desired breadth of agent behaviour.
arXiv:2605.22018v2 Announce Type: replace
Abstract: The Flooded Road Environments Dataset (FRED) is, to our knowledge, the first multi-modal autonomous driving dataset specifically targeting the collection of data from scenarios involving water hazards on the road. The dataset contains images from a 2.3 MP FLIR Blackfly USB3 camera, 64-beam 360 degree point clouds from an Ouster OS1-64 LiDAR, and data from an iXblue ATLANS-C IMU corrected by a Geoflex RTK GNSS, from five separate locations captured both during and after flooding events. The data has been released in two formats: a KITTI-style format for easy integration with existing data tools, and the RTMaps format for direct replay of the vehicle's data capture. We provide semantic labels to enable the training and evaluation of both single-sensor and sensor-fusion methods for water hazard detection. Position and velocity, as well as data captured under dry conditions, are provided to enable the development of location-based detection methods that may incorporate maps, and to evaluate other tasks such as localisation and SLAM.
arXiv:2606.03362v1 Announce Type: new
Abstract: This study examined emerging and established topics in drone research, focusing on citation impact and knowledge flows across China, the United States, the EU, Ukraine, and Russia between 2020 and 2025 using OpenAlex bibliographic data. The findings revealed that drone-related science is characterised by growing geopolitical asymmetries in scientific production, citation concentration, and international knowledge exchange. In particular, China increasingly dominated scientific production, fractional authorship contribution, and domestic citation circulation. In contrast, the United States and EU countries maintained comparatively more internationally distributed citation structures. However, China-affiliated publications became increasingly integrated into global citation networks, particularly through growing citation exchange with the United States and European countries.
Notably, the interpretation of authorship and citation patterns was complicated by the high proportion of publications with unidentified affiliations, which reached 50% in 2025 within weak-signal topics. These findings underscore the importance of developing comprehensive national Research Organisation Registries (RORs).
Although China demonstrated a citation advantage, this was partly driven by high internal domestic citation concentration rather than exclusively by global integration. Moreover, China still imported proportionally more knowledge from the EU-14 and the United States than it exported, with this asymmetry increasing over time. EU-14 countries maintained the strongest citation impact in weak-signal topics, suggesting a more prominent role in shaping emerging research directions. At the same time, China-affiliated publications cited the United States more frequently than the EU-14 in both strong- and weak-signal topics, with this pattern being particularly pronounced in weak-signal areas.
arXiv:2606.03948v1 Announce Type: new
Abstract: We implement simultaneous translation capability with the offline direct speech-to-text translation model Canary, using the state-of-the-art policy AlignAtt, and submit it to IWSLT 2026 Simultaneous Speech Translation Shared task for Czech to English and English to German and Italian.
The strengths of our system are: (1) high translation quality, outperforming similarly sized baselines both in low- and high-latency regimes in computationally unaware simulations; (2) low computational requirements, as the model has only 1B parameters; (3) multilinguality -- support of 25 source and 25 target languages.
arXiv:2606.03947v1 Announce Type: new
Abstract: It is shown that the ranked query enumeration problem for a fixed MSO-query on strings can be solved with linear preprocessing and constant delay in the grammar-compressed setting, where the input string is given by a so-called straight-line program, i.e., a context-free grammar that produces exactly one string. Moreover, `ranked' means that the output tuples of the MSO-query are printed in a specific order that has to be MSO-definable. This is the first result for ranked query enumeration on compressed data. A corollary of this result is that for a fixed polyregular function $f$ and a word $w$ that is given by a straight-line program of size $n$, one can list after preprocessing time $\mathcal{O}(n)$ the symbols in $f(w)$ from left to right with constant delay, which generalizes a result of Bojanczyk for the case where $w$ is uncompressed. The proofs for these results are based on factorization trees, which are made accessible to the grammar-compressed setting (a contribution of independent interest).
arXiv:2606.03946v1 Announce Type: new
Abstract: Database vendors recently released AI functions that can be used in filter predicates. As such functions often rely on costly, black-box ML models, they unveil new data management challenges. Concretely, traditional data skipping techniques for integer and string data fail to be applicable to the new filter type. Indeed, there is no known mechanism for pruning non-qualifying row groups, e.g., when reading files from blob storage.
In this work, we initiate the study of data skipping techniques for ML filters. We make the case that Parquet's default min-max metadata is enough to enable pruning. To this end, we draw connections to two lines of research: (i) the recently proposed query language for ML models and (ii) neural network verification.
Our preliminary results on ReLU architectures show that on tables from TPC-H and TPC-DS, the average pruning effectiveness for filters of selectivity below 0.1% amounts to 27.4%. Finally, inspired by research on spatial joins, we propose an enhanced metadata structure: a size-bounded 2D convex hull that verification tools can make better use of, increasing the pruning effectiveness to 38.31%, while occupying at most 45 bytes per row group and column pair. We observe an end-to-end speedup of 1.07$\times$ over PyTorch in DuckDB.
DAD4TS: Data-Augmentation-Oriented Diffusion Model for Time-Series Forecasting with Small-Scale Data
arXiv:2605.17866v2 Announce Type: replace
Abstract: Small-scale data is a critical problem in time-series forecasting tasks. Data augmentation is an effective strategy for this task, but it has a limitation in generating meaningful data. To address this limitation, we propose DAD4TS, a diffusion-model-based data augmentation method with reinforcement learning, designed for time-series forecasting with small-scale data. In DAD4TS, a data generator is simultaneously trained with a time-series model and controlled by a reinforcement learning model to efficiently generate samples that improve the forecast accuracy of the time-series model. To support small-scale data, we use mathematical methods instead of conventional VAE methods to train the diffusion model by projecting the time-series data into the geometric space. We validated the effectiveness of DAD4TS with seven comparative methods through qualitative and quantitative experiments on six real-world datasets and eight time-series models. As a result, DAD4TS was validated on five datasets.
arXiv:2606.03912v1 Announce Type: new
Abstract: Motivated by the vision of making sixth-generation (6G) networks sustainable, we study the sparse antenna/array activation problems in uplink cell-free massive multiple-input multiple-output (CF mMIMO) networks. We first develop an antenna-level optimal bilinear equalizer (OBE) weighting framework, in which each access point-user equipment (AP-UE) pair is assigned a matrix-valued long-term weight to shape the contribution of individual antenna elements, thereby generalizing the conventional large-scale fading decoding (LSFD) strategy from scalar coefficients to antenna-element-aware weighting. Building on this structure, we formulate sparse antenna activation as structured sparsity-inducing mean square error (MSE) minimization problems, and design four activation schemes at two granularities: antenna-level and array-level, each with UE-specific and network-wide (all-UEs) variants. The resulting convex problems are solved efficiently via the proximal method with closed-form group-wise updates, while the network-wide schemes are modeled through hierarchical sparsity and handled by a tree-structured proximal operator. Numerical results under correlated Rician channels and a detailed power consumption model demonstrate that the OBE weighting scheme consistently improves spectral efficiency over the LSFD, with gains increasing with the number of antennas. Meanwhile, the studied sparse activation schemes can achieve substantial energy efficiency improvement and power reduction with controllable spectral efficiency loss.
arXiv:2606.03315v1 Announce Type: new
Abstract: Graph foundation models aim to learn transferable knowledge from diverse graphs for generalization to unseen graphs and tasks. Unlike text and images, graphs lack a shared vocabulary or regular spatial grid, making cross-graph transfer challenging. This challenge comes from both feature discrepancies and, more critically, diverse graph structures. Existing GFMs mainly improve transferability by unifying feature spaces or incorporating structural tokens and vocabularies. However, existing topology-aware designs still have limitations. Structural tokens are usually discrete, while structural vocabularies often rely on predefined substructures such as trees and cycles, whose limited coverage may miss richer relational patterns across graphs. Moreover, graph signals contain both high-frequency local patterns and smoother low-frequency patterns, which require different propagation behaviors. These components are often entangled in raw graph signals, while this spectral perspective is rarely explored in existing GFMs. To address these challenges, we propose SPG, a graph foundation model with spectral parsing and prototype-guided spatial propagation. SPG applies learnable Chebyshev filters to decompose node features into multiple spectral responses, reducing the mismatch between frequency-specific graph signals and propagation behaviors. It then constructs a Gromov-Wasserstein prototype geometry to distill transferable pairwise relations beyond predefined substructures into a shared structural space. The learned prototype geometry is further projected back as a prototype-guided propagation operator. Experiments demonstrate consistent improvements in cross-domain generalization.
arXiv:2605.11954v2 Announce Type: replace
Abstract: Large language models (LLMs) are increasingly used in social science as scalable measurement tools for converting unstructured text into variables that can enter standard empirical designs. Measurement validity demands more than high average accuracy, which requires well calibrated confidence that faithfully reflects the empirical probability of each measurement being correct. This paper studies the model miscalibration in LLM-based social science measurement. We begin with a case study on FOMC and show that confidence based filtering can change downstream regression estimates when LLM confidence is miscalibrated. We then audit calibration across 14 social science constructs covering both proprietary models, including GPT-5-mini, DeepSeek-V3.2, and open source models. Across tasks and model families, reported confidence is poorly aligned with tolerance-based correctness. As a simple mitigation, we propose a soft label distillation pipeline for calibrating Bert with LLM. The method converts an LLM score and its verbalized confidence into a soft target distribution, then trains a smaller discriminative classifier on encoder models for these targets. Averaged across datasets, this approach reduces ECE by 43.2\% and Brier by 34.0\%. These results suggest that LLM-based social science pipelines should treat calibration as part of measurement validity, rather than as an optional post-processing concern.
arXiv:2605.11170v2 Announce Type: replace
Abstract: Noise-based certified machine unlearning currently faces a hard ceiling: the noise magnitude required to certify unlearning typically destroys model utility, particularly for large-scale deletion requests.
While leveraging public data is a standard technique in differential privacy to relax this tension, its role in unlearning remains unexplored. We address this gap by introducing Asymmetric Langevin Unlearning (ALU), a framework that uses public data to mitigate privacy costs. We prove that public data injection suppresses the unlearning cost by a factor of $O(1/n_{\mathrm{pub}}^2)$, guaranteeing a strict computational advantage over retraining. This establishes a new control mechanism: practitioners can mitigate the need for high noise-and the associated utility loss-by increasing the volume of public data. Crucially, we analyze the realistic setting of distribution mismatch, explicitly characterizing how shifts between public and private sources impact utility.
We show that ALU enables mass unlearning of constant dataset fractions -- a regime where standard symmetric methods become impractical -- while maintaining high utility. Empirical evaluations using variational R\'enyi divergence and membership inference attacks confirm that ALU effectively thwarts privacy attacks while preserving utility under reasonable distribution shifts.
arXiv:2606.02813v1 Announce Type: new
Abstract: Laws and institutions shape individual outcomes through complex interactions with citizens' diverse circumstances, yet how different voting methods navigate this coupled landscape remains poorly understood. We model collective governance as optimization on NK fitness landscapes, where shared bits (laws) are updated by voting while individual bits (personal traits) remain fixed. A cross-dependency parameter $\alpha$ controls how legislation's effects depend on individual circumstances. We compare eight standard voting methods and a generalized scoring family across landscape ruggedness $K \in \{1,\ldots,20\}$ and $\alpha \in [0,1]$ with 1000 runs per configuration.
Under direct democracy, the optimal voting method undergoes sharp phase transitions as a function of landscape complexity: cardinal score voting dominates on smooth landscapes, ordinal scoring with $p=0.35$ at low-to-moderate ruggedness, Borda count across a wide middle range, and STAR voting at the highest complexity. A two-parameter empirical formula reduces the $(K, \alpha)$ plane to a single complexity axis for visualization. Borda count achieves the highest mean fitness and lowest variance across most of the parameter space.
We further introduce a representative democracy model parameterized by identity weight $\beta$ and candidate self-interest $p_{\mathrm{self}}$. Representation reshapes the complexity-dependent structure even under favorable conditions: cardinal score voting dominates across most regimes, with plurality emerging as the top method at high $\beta$ and low-to-moderate $p_{\mathrm{self}}$.
arXiv:2606.02812v1 Announce Type: new
Abstract: Modeling patient trajectories from longitudinal electronic health records (EHRs) requires reasoning over sparse, noisy, and long-context multimodal sequences. Existing LLM-based multi-agent systems address context length but process patients in isolation, failing to mirror how clinicians leverage accumulated experience from similar prior cases. We present Traj-Evolve, a self-evolving multi-agent system with two complementary evolving mechanisms. First, an Experience Pool (ExPool) acts as a non-parametric memory, indexing rejection-sampled reasoning traces to retrieve similar patients as few-shot contexts. Second, multi-agent reinforcement learning (MARL) via reward-ranked fine-tuning parametrically optimizes inter-agent and agent-memory collaboration. A leave-one-out cross-retrieval strategy unifies the two, aligning training- and inference-time behavior under retrieval augmentation. On a lung cancer prediction task utilizing up to five years of multimodal EHRs, Traj-Evolve outperforms 9 strong baselines on the overall population and a challenging never-smoker population. Analysis of the evolving dynamics highlights three key findings: (1) expanding the ExPool shifts optimal retrieval from diverse to specific samples; (2) under MARL, the manager agent's prediction loss converges quickly while the worker agents' temporal reasoning continues to benefit from more verified patients; and (3) the two mechanisms are complementary on the predicted risk, where ExPool improves specificity while MARL improves sensitivity.
arXiv:2606.02741v1 Announce Type: new
Abstract: Large language models (LLMs) are increasingly used in sustainability-related decision support, reporting, and public communication, yet little systematic evidence exists on the environmental attitudes embedded in their outputs. This paper develops a benchmark for evaluating environmental cognition, affect, and behavioural recommendations in LLMs and applies it to 31 widely used proprietary and open-weight models. Drawing on questions from established environmental awareness surveys and additional sustainability-related behavioural measures, we compare LLM responses 1) among models and 2) between models and human survey benchmarks from Germany. We assess their robustness across prompting conditions. We find that many LLMs align more closely with environmentally progressive attitudes than the average survey respondent, exhibiting higher levels of environmental affect and cognition and recommending behaviours associated with substantial potential CO2 reductions. At the same time, we observe no systematic relationship between sustainability-oriented responses and model origin, size, or release context. However, models exhibit contextual sensitivity, controlled by persona-based prompting and show sycophantic shifts mirroring user-specified ideological positions, which raises concerns about steerability and normative reliability in real-world deployments. Our findings provide a reusable evaluation framework for assessing sustainability-related value alignment in LLMs and highlight the importance of governance, transparency, and critical oversight as AI systems become increasingly embedded in sustainability transformations and public decision-making.
arXiv:2606.02737v1 Announce Type: new
Abstract: Dense retrieval models exhibit positional bias: retrieval effectiveness degrades when relevant information appears later in a passage (Zeng et al., 2025). We ask whether this bias can be reduced at inference time, without retraining and without sacrificing overall retrieval effectiveness. To this end, we adapt inference-time attention calibration (Schuhmacher et al., 2026) to downstream retrieval and extend it with a strength coefficient lambda that interpolates between the original and fully calibrated attention distributions. Across three embedding models on SQuAD-PosQ and FineWeb-PosQ, we examine how basket size, calibrated layer set, and strength affect the trade-off between positional fairness and retrieval effectiveness, finding that partial calibration frequently outperforms full calibration. A single configuration (B=128, lambda=0.5, 50% layer depth) improves the harmonic mean of nDCG@10 across positional groups on FineWeb-PosQ for all three models without per-model tuning, and applies to both <s>-pooled and last-token-pooled architectures. This default configuration transfers without modification to PosIR, which spans 10 languages and 31 domains, reducing the Position Sensitivity Index in all 16 length-quartile x model x retrieval-setting combinations, while preserving or improving aggregate nDCG@10. We release our extended codebase at https://github.com/impresso/fair-sentence-transformers
arXiv:2606.02648v1 Announce Type: new
Abstract: We present a physics-based framework that runs from fundamental interactions and constants to biospheres, using a sequence of quantitative nonequilibrium thresholds ("gates"). Each gate is an inequality in measurable variables-free-energy flux, reaction-transport rates, replication fidelity, coding capacity, ecological closure, and climate feedback gains. Crucially, the gate vector is anchored in fundamental physics: dimensionless constants, nuclear resonance placements (e.g., the $^{12}$C Hoyle state), statistical mechanics (Landauer's bound $k_BT \ln 2$) fix the energetic, kinetic, information-theoretic margins that propagate through the gates. This anchoring lets us propagate sensitivities of the constants into biosphere-level metrics (net primary productivity (NPP), cycle-closure ratios, and climate feedback gain), yielding an end-to-end map from constants to biospheres. The framework is predictive: it yields testable inequalities, margin rankings, and population-level correlations between stellar and planetary boundary conditions and biosphere feasibility. It does not claim point predictions of life prevalence; rather, it specifies which gate margins are observable-bounded versus prior-dominated under explicitly stated chemistry/solvent families and forward models. Darwinian dynamics (heritable variation under selection) appears mid-pipeline; the end of the pipeline is a planet-scale biosphere capable of sustaining positive NPP, closing elemental cycles over geologic time. Questions of prevalence are secondary; our primary objective is to establish a constructive physics->chemistry->biology->genetics->ecosystems pipeline with testable margins and observables. As a result, we recast abiogenesis and biosphere persistence as a gate vector of falsifiable inequalities and map their margins to exoplanet observables, turning the problem into a phase diagram with explicit, testable slack.
arXiv:2606.03786v1 Announce Type: new
Abstract: Online debate platforms offer a unique window into the mechanisms driving opinion formation: they capture both explicit political preferences and the peer environment in which those preferences are expressed. In this work, I develop a Bayesian logistic regression model, inspired by ideal point models from political science, to disentangle two competing mechanisms of voting behaviour in online debates: conviction, driven by prior ideological beliefs, and conformity, driven by peer influence. I apply this framework to the Debate.org dataset, comprising approximately 341k votes across 78k debates on 48 socio-political topics. As the debate platform does not provide predefined topic labels for each debate, I infer the topic and stance from the debate text using large language models, and, with a Bayesian approach, I quantify the relative contribution of each mechanism. I find substantial heterogeneity across topics: conviction dominates on issues tied to personal freedoms and lifestyle choices, such as drug legalisation and legalised prostitution, while conformity dominates on several topics widely regarded as paradigmatic cases of moral conviction, including abortion, gun rights, and global warming. These results have implications for the stability of online political discourse and the design of deliberative platforms.
arXiv:2603.07664v3 Announce Type: replace
Abstract: The reflective appearance, especially strong and typically near-field specular reflections, poses a fundamental challenge for accurate surface reconstruction and novel view synthesis. Existing Gaussian splatting methods either fail to model near-field specular reflections or rely on explicit ray tracing at substantial computational cost. We present \textbf{Ref-DGS}, a reflective dual Gaussian splatting framework that addresses this trade-off by decoupling surface reconstruction from specular reflection within an efficient rasterization-based pipeline. Ref-DGS introduces a dual Gaussian scene representation consisting of geometry Gaussians and complementary local reflection Gaussians that capture near-field specular interactions without explicit ray tracing, along with a global environment reflection field for modeling far-field specular reflections. To predict specular radiance, we further propose a lightweight, physically-aware specular adaptive mixing shader that fuses global and local specular features. Experiments demonstrate that Ref-DGS achieves state-of-the-art performance on reflective scenes while training substantially faster than ray-based Gaussian methods.
arXiv:2606.03668v1 Announce Type: new
Abstract: A forefront research question is how energy evolves in weakly collisional plasmas for which departures from local thermodynamic equilibrium (LTE) are significant. The standard approach is studying the terms in the non-LTE energy evolution equation derived by taking the second moment of the Boltzmann equation, but the resultant fluid metrics do not retain information about which particles at which velocities drive energy evolution. A widely studied channel for internal energy density evolution is the pressure-strain interaction. Here we employ the kinetic pressure-strain [S. A. Conley et al., ${\it Phys. Plasmas,} {\bf 31}$, 122117 (2024)], a phase space diagnostic whose velocity-space integral recovers the pressure-strain interaction to disambiguate the contributions to pressure-strain interaction from disparate particle populations in composite phase-space densities. We develop phase-space analogs of the pressure-strain interaction decompositions to provide the phase-space origins of normal vs. sheared flow. We introduce the "kinetic strain-rate" tensor, the phase-space analog of strain-rate tensor, which we argue is needed to interpret phase-space origins of pressure-strain interaction. To demonstrate the utility of these quantities, we investigate them for composite electron distributions near the electron diffusion region in two-dimensional particle-in-cell simulations of antiparallel symmetric magnetic reconnection. We find that the phase space-based diagnostics isolate the roles of distinct populations. These results contribute to a growing body of work providing new methods for quantifying phase space energy evolution for a broad array of processes, from magnetic reconnection to collisionless shocks and turbulence, opening new pathways for answering longstanding problems of particle energization in weakly collisional plasmas.
arXiv:2606.03026v1 Announce Type: new
Abstract: Spiking language models expose activation sparsity that dense Transformer runtimes do not directly exploit. This paper studies that property from a systems perspective. Building on the SymbolicLight V1 spike-gated language model family, we implement a C++ CPU inference runtime that treats sparse binary spike states as an execution primitive rather than only applying post-hoc weight compression. The runtime combines a manifest-driven weight loader, mixed row/column memory layout, AVX2/FMA kernels, per-channel symmetric INT8 quantization, and integer-domain accumulation for spike-conditioned sparse paths. On an AMD Ryzen 7 5800X, an early scalar FP32 baseline decodes at 9.5 tokens/s. Mixed-layout AVX2 FP32 raises this to 14.7 tokens/s, and AVX2 INT8 reaches 19.9 tokens/s on the same step-30k export while reducing the weight footprint from 3.49 GB to 1.06 GB. For the available 186k-step 874M-parameter INT8 export, the C++ runtime decodes at 22.63 tokens/s in a single-thread CPU benchmark, compared with 16.31 tokens/s for TinyLlama-1.1B Q8_0, 11.26 tokens/s for Falcon3-1B Q8_0, and 9.70 tokens/s for Qwen2.5-1.5B Q8_0 under llama.cpp. Thread scaling reaches 47.90 tokens/s at four CPU threads, and 512-token prefill improves from 29.86 to 94.68 tokens/s from one to eight threads. The throughput result comes with a quality cost: the SNN reports WikiText-2 perplexity 24.80, worse than the dense baselines in the same benchmark. We frame the result as an inference-systems study for sparse language runtimes, with longer-term motivation in embodied and edge agents that may benefit from local, low-core inference near sensors and actuators. Spike-aware execution can improve CPU throughput and memory behavior for sparse spiking language models, while model quality, controlled dense training baselines, embodied-task evaluation, and measured CPU energy remain open problems.
arXiv:2606.02998v1 Announce Type: new
Abstract: Automated cough analysis offers a path to low-cost respiratory screening, but most existing work stops at binary COVID-19 detection. A practical tool needs to tell apart several respiratory conditions from one cough recording on a consumer smartphone. We present CoughSense, a system that sorts cough recordings into five classes. These are healthy, COVID-19, asthma or respiratory condition, bronchitis, and pneumonia. We aggregated 18,301 recordings from four public datasets (Coswara, CoughVID, Virufy, and the West China Hospital Pediatric Cough Dataset) and used the OpenAI Whisper encoder as a pretrained backbone for cough disease classification. The main contribution is active-frame QKV attention pooling, which restricts attention to the first 200 of 1500 encoder tokens. This avoids the silence-dilution problem that arises because a 3-second cough fills only 150 tokens of Whisper's 30-second input window. Other training parts handle the 19 to 1 class imbalance and the four-dataset domain shift. These include WeightedRandomSampler, SpecAugment, Balanced Mixup with forced minority pairing, a supervised contrastive auxiliary loss, FiLM symptom conditioning, and gradient-reversal domain adaptation. A dual-encoder model fuses Whisper with the OPERA-CT respiratory foundation model through cross-attention. CoughSense (Whisper-tiny, 8.6M parameters) reached 82.3 percent balanced accuracy on five-fold cross-validation (macro-F1 of 0.817, AUC of 0.941). It beat an ImageNet-pretrained EfficientNet-B2 by 11.1 points and a ViT trained from scratch by 29.6 points. All five classes passed 74 percent recall and four of five passed 80 percent. The dual-encoder model reached 85.4 percent balanced accuracy. Active-frame pooling is the largest single contributor across all ablation components at 5.1 points, which should help any short-audio task using Whisper as a backbone.