Papers

General-purpose agent memory summarizes conversations: it extracts salient snippets, embeds them, and retrieves the top-k into the prompt. A health agent cannot run on summaries: a dose becomes a sentence, "since last week" is resolved at the model's discretion, and a three-month glucose trend cannot be answered by text similarity. We present PIA, a personal intelligence agent deployed alongside a consumer health agent. PIA receives the agent's natural-language requests, decides for itself whether and how to write or read, and turns conversations into typed clinical records and records into a synthesized understanding of the user. Its memory harness consists of four controls – extraction, memory, retrieval, and understanding – each a domain-agnostic mechanism with a pluggable health module: schema, medical alias dictionary, knowledge graph, and temporal rules. We show how the same query receives a different answer as the memory injected into the response context deepens from one-dimensional recall, to a two-dimensional health snapshot, to a three-dimensional trajectory with causality, and report lessons from operation: self-reported health data are missing not at random, question phrasing governs the quality of synthesized understanding, and nearly a third of candidate causal links are structural noise that rules alone remove.

Sequential recommenders typically rely on learnable positional embeddings to encode the order of user interactions. In this work, we ask whether this ordinal signal can be replaced by a structural one derived from the item space. We propose to use Laplacian positional embeddings in SASRec: we build an item co-occurrence graph from training interactions, compute eigenvectors of its symmetric normalized Laplacian, and use them as frozen graph-derived positional embeddings. The backbone architecture and training objective remain unchanged. Experiments on four public sequential-recommendation benchmarks show that this simple replacement improves SASRec performance on most ranking metrics and remains competitive with strong positional and temporal encoding baselines. These findings indicate that item-item graph structure can be an effective substitute for standard ordinal positional embeddings in sequential recommendation.

The arrival of quantum computers threatens the security guarantees of classical cryptography, since quantum algorithms can break schemes that remain secure against conventional attacks. The National Institute of Standards and Technology (NIST) has therefore standardized a set of post-quantum cryptographic algorithms, among them Falcon, a lattice-based digital signature scheme with the most compact signature and public-key sizes of the standardized candidates. Falcon's reliance on floating-point arithmetic makes it hard to implement in hardware, and prior work offers only partial accelerators for specific operations such as signature generation or verification, or a single full implementation generated through high-level synthesis (HLS). This work presents Peregrino, the first hardware accelerator of the complete Falcon digital signature scheme designed from scratch in HDL, targeting resource-constrained edge devices without a native floating-point unit through an emulated floating-point datapath. Implemented on a single Artix 7 XC7A200T FPGA, Peregrino uses 85261 LUTs, 41382 FFs, 44 BRAMs, and 142 DSPs for the Falcon-1024 variant. Against the only prior full implementation, the HLS-based FalconTakesOff, it uses 1,9$\times$ fewer LUTs, 3,4$\times$ fewer FFs, 2,7$\times$ fewer BRAMs, and 9,9$\times$ fewer DSPs, fitting the entire scheme on one FPGA where the HLS design requires at least two. Operating as a peripheral of an on-chip MicroBlaze soft-core, the accelerator additionally reduces key-pair generation, signature generation, and signature verification clock cycles by 92%, 96%, and 85% over the emulated floating-point reference software.

We report fine-grained, deterministic instability in Dynamic Tensor Rematerialization (DTR), an online eviction policy for memory-constrained DNN training, measured on the reference DTR simulator (simrd) using public execution traces. On an LSTM trace, memory budgets differing by 0.10% of unconstrained peak memory select fast and slow execution regimes whose overheads differ by as much as 7.3x; the slow regime is driven by broadly repeated re-eviction of the same storages (evictions per storage rise from 1.33 to 8.27 while the set of distinct evicted storages is essentially unchanged: 5,233 vs 5,236, with the two sets overlapping at Jaccard 0.999). On a ResNet-32 trace, a fine budget sweep reveals a deterministic feasibility inversion: the run is feasible at ratio 0.101, infeasible (OOM) across 0.102-0.106, and feasible again from 0.107. We trace the immediate cause of the OOM to a fully pinned recursive rematerialization frontier that exceeds the budget after every evictable tensor has been evicted. Ablations using the DTR authors' own variants implicate the joint size-staleness scoring term in the observed LSTM instability. We argue these are at least two distinct budget-sensitive pathologies rather than one mechanism, and we separate what is demonstrated from what remains hypothesised. All results concern the reference simulator; reproduction in a production runtime is future work. Code, instrumentation, and raw results accompany this preprint.

3D Gaussian Splatting reconstructs a scene as a collection of Gaussian primitives from a set of posed photographs called the capture. The number of primitives used to represent the scene affects reconstruction quality, storage, and rendering cost. How to select this number automatically across capture scales remains unresolved: configurations effective on standard benchmarks can leave larger captures with too few Gaussians to reconstruct fine details. We observe that the surface to represent, given by the capture's extent and resolution, is known before training, whereas its content complexity becomes apparent during training, through the reconstruction quality on the training views. We introduce TangoGS, which combines capture-derived model sizing with training-based adaptation: the capture determines the scale of the model, and training feedback determines its final size within that scale. Before training, TangoGS derives a learning allowance for model growth from the capture's total pixels after discounting views that re-observe the same scene points. During training, reconstruction quality guides how many Gaussians to add and remove. On 13 standard benchmark scenes, TangoGS matches the mean PSNR of the best-performing evaluated baseline, LeGS, with $48\%$ fewer Gaussians. On eight large captures, the same configuration automatically scales to larger models when necessary, achieving the highest mean PSNR among evaluated methods: $0.54$ dB above the runner-up with $2.3\times$ as many Gaussians. Together, capture-derived learning allowances and training-quality guided density control enable a state-of-the-art quality–size compromise across scene scales without retuning.

Novel view synthesis (NVS) models can produce realistic new views of the same scene from different viewpoints. However, these generated views are not always geometrically consistent with one another. Multi-view (MV) consistency has shown promise as a tool for evaluating these NVS models. Its potential for multimedia forensics, however, remains largely unexplored, particularly for localizing geometric inconsistencies across wide-baseline image pairs. To enable research in this direction, we introduce DeformView, a wide-baseline MV dataset with pixel-level annotations of geometric inconsistencies. Using DeformView, we evaluate state-of-the-art MV consistency-scoring methods and show that approaches developed for NVS evaluation transfer poorly to the forensic task of geometric inconsistency localization. To address this limitation, we propose DEFECt3R, a lightweight learning-based classifier that uses cross-view feature relationships to localize geometric inconsistencies at the pixel level. By learning from explicit supervision, including hard negatives from geometrically consistent yet deformed views, DEFECt3R improves localization performance and substantially reduces false positives compared to existing consistency-scoring methods. Ablation experiments further show that both feature representations and correspondence quality contribute to localization performance. Overall, our findings demonstrate that MV geometric consistency is a promising yet underexplored signal for multimedia forensics and establish a benchmark and baseline for geometric inconsistency localization in wide-baseline MV image pairs. Code and dataset are available at https://github.com/IDLabMedia/DeformView-DEFECt3R

The semiring framework and its extensions form the basis of a rich collection of theoretical results and implementations for provenance tracking of database queries. Many real-world queries use aggregation and conditions on the aggregate values. Support for such queries has been proposed by introducing semimodule elements as aggregate values and formal comparisons between aggregate values as tuple annotations, which takes the approach outside the standard semiring framework. In this work, we show how to introduce a semantics for the provenance of such queries in arbitrary commutative semirings with monus (or m-semirings), without the need for additional operators. This semantics is shown to agree with the standard provenance of the aggregation-free self-join rewriting of HAVING COUNT(*) queries in semirings that are absorptive and where times distributes over monus. We derive algorithms for this semantics and implement them within the ProvSQL system, with viable performance on a real-world dataset for probabilistic query evaluation.

Individuals turn to large language models (LLMs) for guidance across a wide range of economic tasks, from comparing loan options and planning savings to deciding what raise to ask for or how much to charge for their services. LLMs are known to reproduce social biases, and biased economic guidance may influence what users believe they are worth, what they ask for, and what they ultimately accept. This risk is especially salient in India, where economic outcomes are shaped by demographic categories such as caste and urban-rural location. Existing LLM bias benchmarks, however, are largely designed around Western demographic categories and therefore miss key axes of economic disparity in the Indian context. We introduce RupeeBias, a benchmark for auditing demographic bias in LLM-generated economic guidance across Indian economic settings. RupeeBias consists of 39,150 prompts spanning four use cases: salary estimation, salary increment estimation, counter-offer recommendation, and service pricing recommendation. The benchmark follows a single-attribute counterfactual design, holding the description of the user's qualifications, experience, or service offering fixed while varying one demographic identifier at a time. RupeeBias covers 87 India-specific demographic identifiers across six axes: caste, religion, regional identity, gender, disability, and urban-rural location, with all prompts constructed in both English and Hinglish. We evaluate nine LLMs on RupeeBias and find systematic demographic disparities across all six axes. For otherwise identical prompts that differ only in demographic identifier, LLM-generated economic outputs differ by 20.2% on average. We publicly release RupeeBias to support future research on demographic bias in LLM-generated economic guidance across India-specific demographic and economic contexts.

The return migration of skilled workers provides origin countries many important benefits. The likelihood of return migration is thought to depend on relative economic opportunities, but quantitative evidence of its relative importance remains limited. Here we measure return migration rates using the public profiles of software developers on GitHub, geocoded across ten waves between 2012 and 2026, in a panel containing more than 270,000 international movers. Within five years of departure, 8.8% of movers have returned. Origin-country rates range from under 1% to 16%. Hazard models with corridor fixed effects show that migrants whose home countries experience greater catch-up growth return at significantly higher rates. Immutable features of country pairs such as cultural or geographic distance do not explain the estimate. We forecast future return rates using IMF growth projections, predicting, for example, that India's return rate will rise 12% by 2035 and China's 7%. The difference is driven mostly by India's faster projected growth at home. As developing economies close income gaps with the destinations their emigrants have settled in, they recover more of the workers they lost.

Artificial neural networks (ANNs) usually represent neural transmission with fixed trainable weights during a training batch, which omits short-term changes in synaptic efficacy. In addition, the discrete time-step simulation requires additional temporal processing that many conventional ANN architectures do not use. To overcome these challenges, we propose Purin, a biology-inspired and ANN-compatible mechanism, that introduces synaptic efficacy modulation into conventional convolutional neural networks. Purin uses a time-interval-based abstraction for neural activities, which allows Purin to introduce short- and long-term synaptic efficacy changes without using discrete time-steps. Purin introduces a bounded factor to represent temporary synaptic efficacy changes, together with two weight matrices that represent input-side and output-side efficacy. The weight matrices are updated by backpropagation and interpreted as the long-term synaptic efficacy changes. Experimental results show that after removing the confounding factors in the AlexNet, VGG11, and GoogLeNet architectures, Purin improves the classification accuracies in all three models across the evaluated datasets.

Problem. Ultra-high-resolution (UHR) remote sensing with vague user intents has two bottlenecks: visual tokens are expensive, and tool calling must be format-reliable (pretrained models emit zero tool calls zero-shot). Method. WeaveAgent, a two-stage tool-routing agent, decouples routing from visual perception. Stage A is routing-first: emission is trained, not elicited. Stage B executes conditionally: intrinsic queries enter visual answering (full-scene thumbnail; a WeaveEarth-style evidence board as an optional fixed-budget, approx. 5k-token compression interface); extrinsic queries execute tool call on original full-resolution imagery, answering from tool observations in a second, observation-masked round. Training: alignment SFT, then GRPO under reward R_WA2. Results. Alignment SFT lifts extrinsic routing from 0% to 80.75% (323/400); GRPO suppresses 9 intrinsic mis-emissions while tool selection is unchanged. The trained 2B system does not beat the zero-shot 8B baseline overall (0.263 vs. 0.250), a diagnostic contribution. Oracle attribution separates two repair ingredients: loading the observation into context lifts extrinsic answer accuracy from 0.025 to 0.425 under marker-free cross-mode returns, and the two-turn SFT stage adds a further +9.3 points to 0.518 at a small routing cost. A +/- image ablation shows emission suppression is visually grounded, and a query-register matrix shows LLM-rewritten queries cost trained checkpoints 2-11 points. Scope. All training and evaluation use the 5,000 / 3,273 / 1,000-record VagueUHR corpus (600 intrinsic + 400 tool-requiring; the base seeds synthesis and is not used for optimization). Single-pass evidence construction runs at 7.31 s per image on an RTX 4090. Code, data, and evaluation protocols will be released.

Bringing an aerial robotics method from simulation to flight should not require rebuilding a mature autopilot or adding a companion computer. Cybflight is an open-source embedded Rust research autopilot whose typed, replaceable interfaces connect hardware access, perception, state estimation, trajectory planning, and control. This modular development and compile-time optimization workflow is demonstrated with replaceable Rust implementations of model predictive contour-tracking control (MPCTC) and incremental nonlinear dynamic inversion (INDI) running on one STM32H743 without a companion computer. Using this configuration, the vehicle reaches 12.38 m/s during indoor flight, while an outdoor flight using global navigation satellite system (GNSS) position updates reaches 31.4 m/s. These flights show that running demanding estimation and nonlinear control entirely on a flight-controller microcontroller need not come at the expense of a modular autopilot structure.

Build pipelines are integral to modern software development, yet their energy footprint remains largely invisible to practitioners. Existing CI energy tools either rely on model-based estimation (due to hardware access restrictions in cloud runners) or report only total pipeline energy without decomposing it into meaningful phases. Joule-Profiler is an open-source command-line tool for Linux that measures hardware energy consumption via Intel RAPL (CPU), NVML (NVIDIA GPU), and attributes it to user-defined program phases by monitoring standard output. In this tool paper, we demonstrate Joule-Profiler in the context of Maven build pipelines using Google Gson as a case study. By applying a token pattern-matching Maven plugin invocations, we analyze builds across the 5 most recent Gson releases into per-phase energy profiles, comparing cold builds (empty local repository) and warm builds (cached dependencies).

Contact-rich manipulation requires robots to balance accurate motion tracking with compliant interaction, yet most visual-action policies leave compliance fixed at the controller level. We present Imp-ACT, a methodologically grounded and practical approach to incorporating direction-dependent Cartesian stiffness modulation directly into demonstration collection, without manual stiffness selection or offline target reconstruction. During teleoperation, a self-tuning impedance controller adapts stiffness along the instantaneous direction of motion while maintaining compliance in orthogonal directions. The adapted stiffness is applied and recorded alongside visual observations and motion commands, capturing motion and compliance under the same dynamics. We implement this pipeline using Action Chunking with Transformer (ACT) to predict end-effector pose, gripper action, and motion-direction stiffness from visual, proprioceptive, and wrench observations. The performance of Imp-ACT is evaluated on wiping and plug insertion using both success rate and quantitative measures of contact behavior. Compared with fixed low- and high-stiffness baselines, Imp-ACT achieves comparable or higher success while maintaining low interaction forces. In wiping, it reduces contact-force vibration by approximately $29\times$ relative to the compliant baseline and $180\times$ relative to the stiff baseline. In plug insertion, it reduces forces orthogonal to the insertion direction by $43\%$ relative to the better fixed-stiffness baseline. These results highlight the benefit of maintaining sufficient stiffness along the direction needed for task execution while preserving compliance in other directions to limit contact forces and accommodate environmental constraints.

Full-duplex speech language models continuously accumulate acoustic key-value (KV) states, making long-running interactions memory-intensive. During listening, the model can finish processing an audio unit before the next arrives; we term the remaining interval listening-time slack. We propose acoustic-to-text KV compression, which introduces a transcription side channel to convert incoming speech into compact textual memory within this interval. When the cache exceeds a target budget during inference, older acoustic states are evicted while transcripts and recent acoustic context remain. We train the side channel with LoRA using cross-entropy on transcription segments. To preserve listening and speaking behavior, we apply knowledge distillation to the original model's token-level output distributions at native prediction positions. On ten-minute LongSpeech sessions, our MiniCPM-o 4.5 implementation reduces peak streaming KV-cache size by 64.6% compared with the same model without eviction. The proposed method also improves transcription, temporal question answering, and summarization over the baseline. Full-Duplex-Bench evaluations further show comparable pause-handling, turn-taking, and interruption performance.

We study interactive multiprover games and many-round protocols in which the communication between the verifier and the provers is quantum. Parallel repetition is known to suppress soundness error of classical two-prover games arbitrarily close to zero, as shown by Raz (STOC '95), even with quantum strategies as shown by Yuen (ICALP '16), and Bavarian, Vidick and Yuen (STOC '17). Yet, we show that quantum communication can have the opposite, unexpected effect. Specifically, we exhibit a one-round quantum game and two-round $\mathsf{QMIP}$ protocols whose local (unentangled) value is strictly less than one for a single instance, yet equals one under $n$-fold parallel repetition for every $n \geq 2$. The key is that parallel repetition not only imposes additional winning conditions but also supplies additional quantum resources as entanglement in the exchanged states can be exploited jointly across copies. Our multiround examples build on superactivation of zero-error capacities of quantum channels. To establish this connection, we introduce quantum games and interactive protocols which capture the one-shot zero-error classical and quantum capacities, both with and without entanglement assistance. Our constructions use quantum-state verification and a teleportation-based reduction from two rounds to one. In contrast, when shared entanglement is allowed, we show that parallel repetition cannot increase the entangled value when at most $3$ messages are exchanged, consistent with classical results of Bellare, Impagliazzo and Naor (FOCS '97), and that of single-prover systems by Kitaev and Watrous (STOC '00). Combining this monotonicity with our quantum game-channel correspondence, we show that entanglement-assisted zero-error classical and quantum capacities cannot be superactivated.

Pocket-conditioned molecular diffusion updates ambient atom coordinates, but many lead-optimization objectives are expressed on quotient features such as distances, contacts, and anchored substructures. We introduce budgeted quotient-residual guidance (QRG), an inference-time correction that makes these quotient objectives active without retraining the molecular generator. QRG lifts quotient covectors to metric-horizontal ambient directions and delivers them through a trust budget set by the frozen sampler's own step norm: quotient geometry chooses the direction, while sampler motion bounds the scale. We derive the horizontal lift, closed-form sampler-budget update, KL/kinetic interpretation around a frozen reverse step, equivariance conditions, and a product-budget split for budget-capped section and residual controls. Controlled quotient tasks confirm that sampler-relative delivery activates signals that raw local quotient gradients leave dormant. On frozen TargetDiff backbones, official seed-0 CBGBench ligand-generation/editing sweeps show practical quality-runtime gains: Local-QRG improves validity from 0.815 to 0.864 on fragment growing, 0.664 to 0.707 on scaffold hopping, and 0.681 to 0.712 on linker design, while PredNext-QRG improves fragment/scaffold and remains near-neutral on linker. Novelty remains 1.000 and diversity is preserved in the matched multi-seed molecular slice, giving task-dependent improvements without sampler retraining or backbone modification. Overall, QRG provides a lightweight route to quotient-aware inference for frozen molecular samplers with explicit runtime accounting.

Background. Agentic AI systems independently decompose tasks such as literature search, data analysis, and programming into subtasks, search the web, access databases, and execute code. This allows them to perform digital research tasks at high speed. Objectives. Under what conditions does the use of agentic systems produce reliable efficiency gains, and which tasks remain with researchers? Materials and methods. Summary of current studies on literature searches, data analysis, software development, and clinical decision support. Results. For digital activities, work shifts from execution to steering and review. Efficiency gains are greatest when expected behavior can be formalized in advance and tested automatically. In complex agentic systems, recorded sequences of reasoning steps and tool calls can quickly become too extensive for human review. Furthermore, explanations generated by the model do not reliably reflect how an output was produced. One possible step toward more reliable systems is the validation of individual components. The limited reviewability extends beyond research itself; the review of scientific articles and grant proposals is also reaching capacity limits. In pathology, curated and annotated data, researchers' own analytical skills, and institutional exchange of experience are becoming increasingly important. Conclusions. Researchers remain responsible for their results. They must determine what to delegate and how to review the results. The importance of a research question to patients, the field, and society cannot be fully assessed using formalized criteria and remains a matter of expert judgment. Agentic systems can free up time for this.

Misbehavior detection in vehicle-to-everything (V2X) systems is essential for ensuring the semantic correctness of exchanged messages and preventing the dissemination of falsified information. Existing data-centric misbehavior detection approaches largely rely on statistical validation or supervised machine learning models under the implicit assumption of fully observable V2X streams. In practice, however, vehicular environments are inherently partially observable due to hardware failures, intermittent connectivity, and environmental occlusions. Moreover, missingness itself can be strategically exploited by adversaries to evade detection. In this paper, we study misbehavior detection under incomplete V2X observations and propose a deep reinforcement learning (DRL)-based detection framework that learns adaptive policies with incomplete data. We further introduce an adversarial threat model in which attackers exploit or deliberately induce missingness to evade detection, including evasion via natural occlusions and adversarial feature suppression. Extensive experiments conducted on the VeReMi dataset under various missingness patterns demonstrate that DRL significantly outperforms a powerful XGBoost baseline under natural partial observability. However, results also reveal a critical vulnerability: DRL policies can be highly susceptible to evasion attacks that strategically exploit natural missingness. In contrast, DRL exhibits more gradual degradation under direct feature suppression compared to static tree-based models.

Large language models (LLMs) as a judge enable scalable evaluation, but their judgments can be sensitive to response order and, even after removing such position effects, can still diverge systematically from human preferences.We introduce DIAL, a unified framework that combines abundant LLM comparisons with limited human comparisons to separate judge-specific position effects, learn shared structure in position-debiased LLM preferences, and adaptively calibrate that structure toward the human preference target. Theoretically, we study three aspects of DIAL: (i) identification of latent LLM preferences, position effects, and human calibration; (ii) adaptive estimation that balances LLM anchoring against limited human evidence; and (iii) fixed-weight uncertainty quantification for the calibrated human preference. Empirically, we evaluate position debiasing and human alignment separately in controlled simulations and on three human-preference benchmarks, showing that DIAL remains robust to unbalanced response order, achieves strong human-aligned rankings with limited labels, and adapts toward human evidence when LLM information is imperfect. Our real-data study collects over 410K judgments from 21 LLM judges in both display orders, providing a resource for future studies of LLM-judge bias, heterogeneity, and human alignment.

Estimating the influence of training examples on model behavior is essential for data debugging, valuation, and attribution. Existing influence estimators often produce incompatible rankings, which are commonly ascribed to approximation error. We argue that a more fundamental source of disagreement is specification mismatch: influence depends on the behavior being attributed, the intervention applied to each training example, and the counterfactual training process that maps the intervention to a model response. These choices are especially important when the target behavior requires a tractable surrogate, such as query loss, a logit, or a margin. We formalize influence as a counterfactual estimand, distinguish specification mismatch across estimands from approximation error in estimating a fixed estimand, and organize representative estimators by their implied specifications. We further derive a local decomposition that exposes how behavior signals, training signals, and counterfactual parameter responses interact. Controlled experiments show that exact estimands under different specifications can induce different rankings, whereas approximation error grows as perturbations move farther from their linearization points. Experiments on noisy label detection and LLM attribution show that specification choices significantly affect attribution quality, especially for the choice of behavior surrogate. Behavior-aligned specifications can identify target-specific training examples obscured by default loss-based or similarity-based specifications. These results establish specification analysis as a necessary first step for interpreting and comparing data influence estimators.

Underwater robotic inspection depends on acquiring views that reveal task-relevant structure. For a structurally complex coral colony, recognising the target is only the starting point: the robot must select and execute a viewing motion suited to the inspection task. We present CoralPlan, a vision-language system that selects an observation skill from a current camera image and task text supplied by an episode manifest. A shared motion interface executes orbit, patch, or survey as target-relative trajectories; the remaining plan fields provide operator guidance. Observation completion requires target keeping and primitive-specific coverage, while joint success also requires selection to match the recorded reference. We evaluate this interface in 144 simulated episodes and 36 matched simulation-hardware pairs. In a clear-water pool with external target-reference poses, hardware observation completion reaches 77.8% and joint success reaches 63.9%. The experiments identify both reference-mismatched completions and incomplete observations after a matching skill selection. These results connect observation-skill choice to measurable underwater execution outcomes and identify where task-directed acquisition succeeds or fails.

Achieving robust cross-embodiment generalization in imitation learning demands overcoming a critical representation flaw that inextricably entangles task semantics with hardware-specific visual geometry. We propose an interaction-centric framework that leverages the shared structure of two-finger grippers via a parameterized universal gripper abstraction, yielding a canonical gripper-frame representation. Given language and RGB-D observations, a VLM infers the subtask and grounds an interaction triplet (gripper, held, target), while SAM~2.1 tracks masks to reduce VLM queries. We design concise hybrid features that combine target/collision artificial potential fields for global guidance with segmented gripper-frame point clouds for local geometry, and use a Flow-Matching Transformer to predict smooth 7-DoF action chunks. Experiments in simulation and real-world tasks demonstrate that ours is the first imitation learning approach to simultaneously achieve competitive benchmark scores and extreme cross-embodiment/cross-viewpoint zero-shot sim-to-real transfer to completely distinct, heterogeneous robot platforms.

Auroral spectrographs such as the Auroral Spectrograph In Skibotn (ASIS) record hundreds of thousands of emission spectra, but only a few hundred can be labelled by an expert. To exploit the rest, we pretrain a 1D Vision Transformer with a masked autoencoder on 223,000 unlabelled spectra. Without labels, its representation recovers the emission-line intensity ratios that physicists use to diagnose the precipitating particles (R^2 0.91 vs. 0.77 for an untrained control) and, under one linear probe, classifies as well as 13 features designed by experts. Fine-tuned, the model outperforms the previous supervised auroral classifier on its own benchmark (macro-AP 88.5 vs. 77.8), reaches 0.870 mAP, and exceeds the same architecture trained from scratch by +0.159 with 10% of the labels; attribution shows that it uses both N2+ bands. Could an existing pretrained model replace it? Two astronomical spectral foundation models and a time-series model transfer according to their spectral window: SpectraFM, trained in the infrared, falls below the untrained control, whereas SpecFormer, trained in the optical, approaches in-domain pretraining without reaching it.

For distinct integers $a$ and $b$, an $\{a,b\}$-edge-weighting assigns $a$ or $b$ to each edge and labels each vertex by the sum of its incident weights. Such a weighting is proper if adjacent vertices receive distinct labels. We prove that, for every fixed pair of distinct integers, deciding whether a proper weighting exists is NP-complete even for simple cubic planar graphs. On planar multigraphs with $m$ edges, we give an exact $2^{O(\sqrt m)}$-time algorithm and, assuming the Exponential Time Hypothesis (ETH), exclude $2^{o(\sqrt m)}$-time algorithms even for simple cubic planar graphs. As a consequence, locally irregular $2$-edge-coloring is NP-complete on simple cubic planar graphs, admits a deterministic $2^{O(\sqrt n)}$-time algorithm on $n$-vertex graphs in this class, and admits no $2^{o(\sqrt n)}$-time algorithm under ETH. For maximizing the number of edges joining vertices with distinct labels, we give a deterministic efficient polynomial-time approximation scheme (EPTAS) on planar multigraphs, a polynomial-time $1/2$-approximation on multigraphs, and APX-completeness even on simple cubic graphs. Extending a partial $\{a,b\}$-edge-weighting to a proper one is NP-complete for every fixed pair even on simple cubic planar bipartite graphs, while it is polynomial-time solvable on trees. The hardness persists even when the prescribed edges form disjoint paths of length $6$ and all edges of each path have the same prescribed weight.

Can the spatial organization of cortical activity help transfer frozen image representations to fMRI prediction? We introduce FlatClip, which renders cortical activity as anatomy-linked flatmaps and extracts features with a frozen SigLIP2 encoder. Only a lightweight downstream probe is trained, without fMRI-specific encoder pretraining or fine-tuning. Across four resting-state benchmarks, FlatClip outperforms the evaluated ROI-level baselines on HCP and ADNI, but is weaker on PPMI and below the strongest voxel-level models overall. On NSD visual-fMRI decoding, restricting inputs to visual or NSD-provided task-active cortex improves recognition over whole-cortex inputs. Spatial perturbations reduce performance with both retrained and fixed readouts, and anatomy-linked arrangements outperform cortical-pixel permutations across three colormaps. Together, these results demonstrate a practical route for reusing image-pretrained features in fMRI and highlight cortical arrangement and input-region selection as important design choices. Code is available at https://github.com/OneMore1/FlatClip.

Composed image retrieval (CIR) aims to retrieve images that satisfy a user-specified modification while preserving relevant visual content from a reference image. Collecting target images for this purpose is costly, motivating zero-shot CIR methods that instead use target captions as supervision. However, target captions may omit source details that should be preserved. We therefore propose, Preserve-and-Compose Training, which complements target-caption supervision with visual evidence from the source image. PACT learns from image–text–text (ITT) triplets without target images or gallery updates, aligning composed queries with target captions while preserving source evidence through visual supervision. We further introduce Chord scoring, which combines target similarity with source-relative directional agreement in the frozen image space. Results across four ZS-CIR benchmarks show that combining target-caption supervision with source-image evidence leads to strong retrieval performance across datasets, backbone scales, and external galleries. The code is available on https://github.com/sehyunkwon/PACT.

Scaling studies typically represent training data by a single count of samples. For hierarchically and spatially structured data, however, the same number of samples can be drawn from few or many sources and distributed differently across the underlying domain. We therefore study data scaling as an allocation problem, separating unique sample count, source diversity, and spatial coverage. We study this decomposition in microscopic whole-brain histology, where a source is an individual brain, and a sample is an image patch at a specific spatial location. Across 93 controlled pretraining runs of a contrastive model that uses spatial proximity for supervision, we vary data allocation, compute, and model capacity over 11.6 million spatially anchored image patches from 21 human brains. Performance improves with more unique samples, broader spatial coverage, additional compute, and larger model capacity. At fixed sample count, distributing samples across one to 18 subjects produces no detectable improvement, even though representations generalize substantially better to subjects encountered during pretraining. Inter-subject variation therefore strongly affects generalization, but additional subjects provide no benefit when a fixed sample budget is distributed across more sources. These results establish sample count, source diversity, and spatial coverage as distinct axes of data scaling in spatially structured representation learning.

Large-scale Digital Surface Models (DSMs) can be produced cost-effectively from satellite images via stereo-photogrammetry. However, the resulting 3D maps are often contaminated by noise, outliers, and voids. On the other hand, aerial LiDAR provides high-accuracy elevation measurements at a substantially higher cost. In this work, we study diffusion models conditioned both on photogrammetric DSMs and Pléiades imagery to refine vertically co-registered DSMs. We introduce a modified Stable Diffusion 3 architecture with a pruned text stream and a patch-wise normalization strategy, enabling stable training on LiDAR data and transfer from natural images to elevation maps. Experiments in French cities demonstrate that multimodal conditioning improves elevation accuracy, reducing Dense Urban RMSE from 6.00 to 3.45 m in the in-context cities and from 4.16 to 2.77 m in the held-out city of Bordeaux.

We present Light Field Primitives (LFP), a formulation for novel view synthesis that replaces the dense ray database with a compact set of differentiable primitives in the classical two-plane parameterization. Each primitive condenses a group of rays into one learned record, and its response to a query is governed by how closely that query belongs to the group. Rendering a camera ray then reduces to compositing all responses it elicits, and a scene can be optimized directly from posed images and rendered in real time with rays. Beyond its competitive performance on standard benchmarks, the main advantage of LFP is structural: its primitives reside directly in the 4D ray space, so optical and appearance effects that are already operations on the light field become behaviors of a single shared renderer. With minimal changes to that renderer, LFP supports multi-scale anti-aliasing, defocus deblurring with refocusing, rendering for fisheye cameras, and even transparent object reconstruction with ray refraction, matching specialized frameworks that devote substantial machinery to these effects.