Papers

Reliable localisation of shallow subsurface vessels is important for safe robot-assisted venous access and vessel-aware manipulation, but collecting diverse tactile data on physical hardware is costly, time-consuming, and can degrade soft vision-based tactile sensors. We present a robot-assisted sliding-palpation framework in which a calibrated digital twin generates labelled tactile sequences, reducing reliance on real-world data. The twin models sensor-vessel contact, is calibrated against real palpation trajectories using Bayesian-optimisation-based domain adaptation, and is randomised over sliding direction and contact conditions. A spatio-temporal graph neural network trained on simulated marker trajectories performs per-node vessel classification and produces a human-verifiable top-view localisation map through 2D-to-3D-to-2D geometric projection. We evaluate three datasets: Sim, Silicone, and Meat, the latter a raw-meat phantom with vessel models at nominal depths of 0 to 30 mm, using four train-to-test configurations: Sim to Sim, Sim to Silicone, Sim to Meat, and Meat to Silicone. The calibrated twin achieves a simulated-to-real marker-alignment mean absolute error of 0.50 mm at deepest contact across four canonical interactions. After reprojection onto a 1 mm top-view grid, predicted vessel pixels lie on average 1.05 to 5.49 mm from the nearest true vessel pixel across the four models, with 1.05 to 1.31 mm for all except Sim to Meat. The larger error for Sim to Meat reflects the greater domain shift and current limit of simulation transfer. These results demonstrate progress toward trustworthy tactile palpation through calibrated simulation, interpretable localisation, and transparent cross-domain evaluation. Code, model weights, and data are publicly available on GitHub and Zenodo.

Vision-language models such as CLIP and SigLIP provide strong zero-shot recognition, but their predictions can degrade when deployed on target data that differ from the pretraining distribution. Test-time adaptation offers a practical way to improve robustness without source data or target labels, yet existing methods often rely on either prompt-side adaptation or image-side target evidence alone. In this work, we introduce GATE, a training-free two-pass transductive test-time adaptation framework that uses the unlabeled target set while keeping the image encoder, text encoder, and prompt parameters fully frozen. Instead of representing each class with a single prototype, GATE builds two complementary Gaussian sources of evidence in the shared vision-language feature space: a text Gaussian estimated from multiple language descriptions and an image Gaussian estimated from reliable unlabeled target samples. A class-wise reliability gate controls the influence of image-derived pseudo-evidence, and a score-level generalized Product-of-Experts fusion produces a normalized residual correction to the original zero-shot logits. Across fine-grained recognition datasets, ImageNet-family distribution shifts, multiple CLIP backbones, and SigLIP-B/16, GATE achieves the best average accuracy in every benchmark/backbone group. It improves zero-shot performance by an average of 5.41 points and outperforms the strongest non-GATE baseline by 1.94 points, demonstrating the benefit of reliability-gated distributional evidence for frozen VLM adaptation.

Floating things invite touch. We present Feelium, a blimp-based telepresence platform that enables visual embodiment and touch interaction through its inflatable skin. Through a VR headset, a remote person inhabits the blimp, looking out of it first-person, appearing on its skin as a face or avatar, and steering it through the room. Partners in the room pat it, press a palm against it, draw on it, or lean into it; the skin senses each contact, renders it into the wearer's view in VR spaces. Touch thus provides a physical interaction channel for remote presence, turning the skin into a shared surface between remote and co-located partners.

The standard formulation of function-correcting codes uses systematic encodings. We study the redundancy cost of this constraint when both the prescribed function and the encoding are linear. We introduce the function-separation distance and show that several classical unequal error protection parameters are special cases. For linear functions and encodings, this distance is the first relative generalized Hamming weight of the code relative to the encoded kernel. We formulate free and systematic linear separation problems. We prove that the optimal free redundancy depends only on the rank of the function, and determine the optimal systematic redundancy for prescribed separations $d\leq3$.

Recent machine learning research has increasingly focused on equilibrium analysis in non-cooperative games rather than solely on optimal solutions. Many such problems involve shared constraints and can be formulated as Generalized Nash Equilibrium Problems (GNEPs). For strongly monotone games, existing methods compute consensus-based variational GNEs (v-GNEs) by exchanging Lagrange multipliers. We propose a fully distributed continuous-time algorithm for shared linear equality constraints that converges without multiplier exchange and reaches any GNE, reducing communication overhead and improving privacy. Discrete-time schemes are also provided, and the method is validated on a multi-robot placement task.

Large language models can generate interactive web interfaces, but reliable generative UI requires maintaining an executable artifact as user requests evolve. We introduce EvoGenUI-Bench, a benchmark for multi-turn interface maintenance comprising 150 five-turn tasks and 750 turns across three scenarios: information presentation, executable interaction, and tool-grounded external state. We execute generated artifacts in a browser and evaluate them using screenshots, source and DOM evidence, actor traces, and runtime logs. Beyond turn-level and episode-level success, we measure cross-turn retention with Adjacent Pass Retention. Across eight models, even the strongest achieves 74.9% Turn Pass while completing only 37.3% of five-turn episodes; APR further falls to 52.4% on tool-grounded tasks. Diagnostic analysis shows that presentation failures center on information architecture, interaction failures on derived-state propagation and affordance binding, and tool-grounded failures additionally involve external-state grounding and requirement decomposition. These results reframe generative UI evaluation from judging isolated outputs to testing whether interface behavior, derived state, external state, and assistant claims remain synchronized as the artifact evolves.

The growing availability of dense commercial Synthetic Aperture Radar (SAR) time series enables temporal Interferometric SAR (InSAR) analysis, but fixed classical filters fail under heterogeneous acquisition geometries, degrading phase quality and temporal consistency. We propose FiLM-GPNet, a geometry-conditioned network for wrapped-phase restoration that explicitly adapts to acquisition differences using Feature-wise Linear Modulation (FiLM) and a 7D per-pair geometry descriptor. The model is trained with pseudo-supervision from Goldstein-filtered interferograms and regularized by interferometric physics via triplet-closure consistency, while also estimating per-pixel aleatoric uncertainty. Experiments on three Capella Spotlight stacks from the IEEE GRSS 2026 Data Fusion Contest show that FiLM-GPNet reduces temporal residual by 68% (Hawaii) and 66% (Western Australia) relative to the Goldstein baseline, alongside closure error reductions of 10% and 13%, respectively. In Western Australia, it further improves unwrapping success rate by 7.7 percentage points and Digital Elevation Model (DEM) Normalized Median Absolute Deviation (NMAD) by 31%. The model also shows strong zero-shot generalization to a geographically and geometrically distinct third stack (Los Angeles) without retraining, supporting geometry-conditioned restoration as an effective alternative to fixed classical filtering across heterogeneous stacks.

AI agents are moving toward persistent, stateful execution across various applications, accumulating execution state and external effects that are costly to reconstruct after failures. Checkpoint and rollback (C/R) are becoming essential for recovery, yet their security implications remain largely unexplored. Correct rollback does not imply secure recovery: a faithfully restored checkpoint may resume an execution whose states, assumptions, and external effects never coexisted in any valid history. In this paper, we present the first systematic security study of checkpoint and rollback in existing agent systems. By examining representative agent C/R systems, we characterize the design space of existing C/R mechanisms and develop a general execution model that captures their recovery boundaries and state dependencies. From this model, we identify five fundamental failure modes spanning incomplete or inconsistent internal state, stale external dependencies, nondeterministic replay, and unrecorded external effects. We further demonstrate their security impact through three end-to-end attacks on Hermes, Cline, and LangGraph, enabling malware-verification bypass, unauthorized mail forwarding, and double payment. To systematically study these failures in practice, we develop a multi-agent analysis pipeline that reconstructs execution semantics, identifies violations of the five failure conditions, and validates them through actual rollback. Across five representative frameworks, our evaluation shows that these failures recur across heterogeneous C/R designs and stem from a common gap between the state restored by a checkpoint and the dependencies required for secure continuation.

Since the seminal work of Selinger and Valiron, the standard design for quantum lambda calculi has kept the quantum state outside the program: terms manipulate pointers to an external register. This is largely due to the difficulty of eliminating tensor products. For example, the calculus $\lambda_\rho^\circ$ embeds density matrices directly within terms, where terms carry the entire computation state, a feature particularly appealing for program verification. However, lacking a tensor elimination construct, it can neither access the individual qubits of a compound state nor discard them. Borgna showed that this inability to discard qubits makes the calculus strictly less expressive than the quantum lambda calculus of Selinger and Valiron. In this paper we show that tensor elimination is possible in this setting. The key observation is that the Pauli decomposition, combined with the spectral decomposition of the Pauli matrices, allows any $n$-qubit density matrix to be expressed as a real linear combination of tensor products of single-qubit density matrices. Exploiting this fact, we extend $\lambda_\rho^\circ$ with a construct $\mathrm{let}\ x^{\otimes n} = \rho\ \mathrm{in}\ t$, which binds each $x_i$ to a single-qubit density matrix arising from the decomposition of $\rho$. We equip the extended calculus with a rewrite system, a type system, and a denotational semantics, and prove Subject Reduction, Progress, Strong Normalisation, Soundness, and Adequacy. The new construct also recovers the missing ability to discard qubits, thereby restoring expressiveness. Moreover, we show that this is achieved in a physically principled way: a variable unused in $t$ is interpreted exactly as being partial-traced out, as dictated by the no-deleting theorem. We illustrate the resulting compositionality through quantum teleportation and the three-qubit bit-flip code.

A language-guided robot operating in a real kitchen must do more than produce a plan that appears correct. It must also execute that plan safely in cluttered environments under imperfect perception. Large language models (LLM) can decompose instructions into action sequences, yet a language-action gap remains: a plan may appear valid linguistically while being physically infeasible under kinematic and collision constraints. We bridge this gap by formalizing the reasoning-execution boundary as a typed contract. From RGB-D observations, the system grounds perceived objects in an explicit, collision-aware scene model and constrains language-level decisions through schema-validated tool calls defined by the Model Context Protocol (MCP), rejecting malformed commands before they reach the robot. Each validated call is deterministically grounded in a MoveIt Task Constructor pipeline, where candidate motions are evaluated against the reconstructed planning scene in a verify-then-act step. Only trajectories that pass both kinematic and collision checks are sent to the robot. On a physical UFactory 850, the method achieves up to 80% success across ten trials per task on pouring tasks involving liquids, granular media, and discrete solids. It achieves 90% success on a grasp-and-place task using the same planning, protocol, and verification stack. Although a scripted policy slightly outperforms our method on the easiest task, its success rate falls to 10% on the hardest, compared with 60% for our method.

Arabic large language models must refuse harmful prompts without over-refusing benign or sensitive prompts, yet a single refusal rate hides this trade-off. We evaluate it using benign refusal B and harmful-prompt refusal H, where H measures refusal rather than harmful compliance. Across five Arabic-capable models and 130 runs on the full human-written AraSafe set, refusal-only supervised fine-tuning (SFT) collapses toward blanket refusal, whereas selected mixed-SFT configurations reach H = 90% to 93% at B = 14% to 23%; four selected configurations exceed the H = 90% target in all three runs, while Fanar does so in two of three. Direct Preference Optimization (DPO) and inference guards change B and H differently across models rather than acting as uniform upgrades. In a blinded 300-response audit, annotator binary-refusal agreement is 89.0% (kappa = 0.78); Qwen3Guard and Aya Expanse 32B reach 88.7% and 91.0% accuracy, respectively, with no conclusive paired difference. Selected SFT raises H on Arabizi for all five models, but none reaches 90%, showing only partial transfer from Modern Standard Arabic. Overall, the results support model-specific operating-point selection: set a deployment target and retain only interventions that improve it.

Code generation increasingly relies on large transformer models, whose capability advances with scale. Yet such a scale is costly, creating demand for small models, especially where data is limited. Recursive models address this by reusing a single block to add depth rather than stacking independent layers. Such models are typically evaluated by teacher-forced fit (next-token loss on ground-truth prefixes) or task accuracy, at a single checkpoint, whereas code is produced by free-running generation, where the model extends its own output. Whether a teacher-forced advantage survives free-running generation, and whether it holds across training, remains open. To study both, we compare a ~28M-parameter autoregressive Tiny Recursive Model (TRM-AR) on natural-language-to-Python code generation against parameter-matched and depth-matched controls, tracking fit and generation across 40 epochs and three seeds. The fit ranking between the recursive model and the depth-matched control reverses twice. Selecting each checkpoint by validation loss and examining the trajectory yields a consistent comparison. At equal parameters, TRM-AR fits, generates, and generalizes better than the parameter-matched control while recovering approximately 45% of the validation-loss gap and 57% of the generation-quality gap between the two controls, at roughly 175 times the per-step cost of the parameter-matched control. However, at equal effective depth, the larger transformer fits and generates better at its validation optimum, suggesting TRM-AR's advantage lies in resistance to overfitting, not greater capability. These findings suggest that recursive code generation models should be evaluated jointly on fit and generation across the training trajectory rather than at a single checkpoint.

Tool-augmented multimodal reasoning integrates external tools (e.g., object detection, depth estimation) into multimodal large language models (MLLMs) to address perceptual bottlenecks in complex visual tasks. However, existing approaches rarely verify tool outputs, limiting their ability to detect and recover from tool failures. We propose ReVISE, a framework that equips MLLMs with verification and dynamic error recovery for tool-augmented reasoning. ReVISE introduces (1) a curated training dataset that supervises reflective behaviors, enabling models to validate tool-derived evidence, reformulate queries when visual mismatches arise, and fall back to intrinsic grounding when external tools are unreliable; and (2) a reinforcement learning based targeted rewards that encourage internal reflection and penalize spatial misalignment. Experiments on several benchmarks demonstrate consistent improvements over existing methods, highlighting the importance of error detection and correction in tool-augmented multimodal reasoning.

Retrospective backtests provide a limited test of adaptive trading agents: they cannot rule out historical contamination, expose sensitivity to a single realized market path, or reveal internal degeneration during long-horizon adaptation. We introduce FORESIGHT-9, a prospective and process-aware benchmark built from nine auditable counterfactual stress worldlines branching from a common July 2026 information boundary. Each worldline specifies staged macro-financial events and joint multi-asset anchors; a deterministic generator realizes the trajectories, while observations are disclosed according to in-world time. A common contract standardizes observations and execution while preserving each agent's native adaptation loop. We evaluate two adaptive trading-agent frameworks with two foundation-model backbones across 36 long-horizon runs. Agent rankings vary substantially across worldlines and backbones, and a fixed equal-weight policy outperforms 31 of 36 runs. Process telemetry exposes failures that terminal returns conceal: in one high-return run, the live factor library collapsed while executed holdings converged to the equal-weight fallback, even though decision records continued to report an active factor ensemble. FORESIGHT-9 therefore evaluates not only portfolio outcomes, but whether adaptive agent state and execution remain coherent across alternative futures. We release the worldlines, trajectories, audit traces, and regeneration scripts.

DNS resolution has no notion of a verifiable resolver path. When an ISP outsources resolution to a third-party provider, a user's queries can cross organizational and national boundaries without their awareness—and the client that issued them has no protocol mechanism to learn which resolvers handled the query, where they sat, or who operated them. We argue that this opacity is an architectural gap rather than a deployment accident, and that a reported, verifiable resolver path should be a first-class goal of the resolution protocol. We motivate this with measurement and then show the gap is cheap to close. Using RIPE Atlas across 190 countries, we find that resolution routinely leaves the client's organization and country: under a conservative in-AS treatment of unattributable observations, 39.8% of observed resolver chains (6,622 of 16,636) use a frontend in a different AS than the client, one in four geolocatable anycast frontend pairs resolves outside the client's country, and a single operator—Google Public DNS—accounts for roughly two-thirds of those out-of-country cases. We then present Resolver-Path, an approach in which participating resolvers report their identity as they forward the query. Its base layer is cooperative in-band disclosure, it carries resolver-path metadata at near-neutral throughput, latency, and CPU cost. Because a non-participating resolver can ignore or strip the option, disclosure alone establishes the verifiable presence of reported hops, not the absence of hidden ones. Attestation authenticates the integrity, ordering, and freshness of the cooperative assertions carried by the selected response. Together, disclosure and attestation provide bounded evidence about the selected response's reported resolver chain—the substrate DNS currently lacks for jurisdictional accountability.

Spatio-temporal graphs are widely used in modeling complex dynamic processes such as temporal forecasting, molecular dynamics, and healthcare monitoring. Recently, stringent privacy regulations such as GDPR and CCPA have introduced significant new challenges for existing spatio-temporal graph models, requiring complete unlearning of unauthorized data. Since each node in a spatio-temporal graph diffuses information globally across both spatial and temporal dimensions, existing unlearning methods primarily designed for static graphs and localized data removal cannot efficiently erase a single node without incurring costs nearly equivalent to full model retraining. To address this, we propose CallosumNet, a spatio-temporal graph unlearning framework biologically inspired by the corpus callosum structure. CallosumNet makes two key technical contributions: (1) it reconstructs subgraphs using biologically-inspired virtual edges; and (2) it restores interlinked spatio-temporal dependencies among subgraphs via a lightweight meta-graph integration layer. Empirical results on four diverse real-world datasets show that CallosumNet achieves complete unlearning while maintaining accuracy very close to the gold model. The code is publicly available at https://github.com/wenlu-lab/STGraphUnlearning.

The most disruptive step that enterprises of small-medium size and maturity can take to make the most of the latest technological advances in AI is to step back from the hype and focus on establishing or reviving a good knowledge foundation layer. It is a hard message to present to the executive team; therefore, it needs to be backed by evidence, and its implementation needs to be of minimal impact on the existing processes. In this vision statement, we discuss how we need to rethink what evidence speaks to the decision-makers and propose a low-impact data strategy that adapts to the existing and ever-changing data flows and processes across the company. We firmly believe that knowledge graph techniques will increasingly become non-negotiable in the data strategy of an AI-powered enterprise, provided that we approach their design in a modular, dynamic and cross-functional way.

Early-stage product design retrieval often requires more than category recognition: designers may need reference examples that match both a short semantic intent and a rough structural cue. Existing product-image resources and generic image–text retrieval benchmarks rarely separate category retrieval from within-category structural fit. We present Sketch2Inspire, built from a curated subset of Amazon Berkeley Objects with aligned text queries, edge-based sketch-proxy queries, and fused text–sketch queries. The resource separates broad category-level retrieval from structure-sensitive within-category retrieval and includes a human-graded reference protocol for calibration. We evaluate a lightweight reference system based on pretrained CLIP-family encoders, comparing text-only retrieval, sketch-only retrieval, weighted late fusion, and text-first reranking without updating model weights. Under broad relevance, late fusion obtains the highest score (nDCG = 0.9962). Under automatic structure-sensitive relevance, late fusion again obtains the highest score (nDCG = 0.7015), exceeding text-only retrieval (nDCG = 0.5912). In the human-graded results, late fusion obtains the highest nDCG@10 (0.9133), while text-only retrieval ranks second (0.9030). These results show that the retrieval gain from multimodal input depends on how relevance is defined. Sketch2Inspire therefore provides a diagnostic resource for evaluating modality contribution and supports the development of structure-aware product-retrieval protocols with independent human annotation.

Enterprise data agents answer business queries by chaining many tool calls over multiple reasoning steps, routinely accumulating hundreds of thousands of context tokens per session. Existing compression strategies typically allocate retention budgets without accounting for the downstream consequences of removing individual tool outputs. Aggressive compression may therefore trigger costly tool re-invocations that offset the initial savings. We call this the compression–consequence gap. To close it, we propose TRACER, which formulates compression as a sequential per-tool decision problem. A lightweight REINFORCE policy assigns query-conditioned retention ratios using only information available at each compression event. Its consequence-aware objective jointly accounts for task success, total token consumption, and post-compression tool re-invocations. To improve credit assignment, TRACER uses a learned outcome model to compare the predicted consequences of the selected retention ratio with those of fully retaining each tool output. On held-out production queries across three compressor backends, TRACER reduces total token consumption by 29–46% relative to keeping all context while maintaining comparable or higher task success. Compared with a tool-type-conditional static policy, TRACER provides an additional 15–18% of token savings. Interventional rollouts show that the learned per-tool credit scores correlate with measured single-tool consequences. The learned policy also yields positive savings when transferred across agent backbones and compressor architectures, and reduces token consumption by 18–25% on five held-out LOCA-bench environments. These results demonstrate the value of consequence-aware, per-tool context retention for improving the efficiency of long-horizon language agents.

Sparse computations are fundamental to scientific computing, graph analytics, and machine learning, yet their performance is highly sensitive to the diverse sparsity and patterns. This is because cache reuse, memory coalescing, and load balancing depend critically on the sparsity patterns. This work gives the first known exploration of the connections between sparse matrix computation and spectral analysis by treating sparse matrices as two-dimensional signals and analyzing their frequency-domain representations through Fast Fourier Transform. We show that spectral signatures uncover global structural characteristics that are not sufficiently captured by conventional spatial statistics and provide complementary information for understanding sparse computation performance. Experiments on incorporating spectral features into machine-learning-based SpMV format selection demonstrate the usefulness of such spectral analysis over a state-of-the-art spatial-only model. By uncovering the principled connections between spectral characteristics and sparse matrix computations, this work introduces a novel analytical perspective into sparse computation, and provides a new approach to enhancing the current sparse structure characterization and optimization. On pruned LLM decoding, adding spectral features improves kernel selection and yields 1.035–1.245$\times$ kernel speedups.

Spatiotemporal graphs underpin applications such as traffic forecasting, weather forecasting, and healthcare monitoring. Privacy regulations such as the GDPR and the CCPA require the complete removal of unauthorized data from trained models, but achieving this on a spatiotemporal graph is difficult: because information propagates globally through both spatial and temporal message passing, fully erasing a node's influence forces costly full-graph retraining. ST-graph unlearning requires both exactness and efficiency. We propose IsleNet, which uses spatial-entropy-guided partitioning to create balanced, locally coherent subgraphs and reconnects them with lightweight virtual edges. Upon an unlearning request, only the affected subgraph encoder and virtual-edge layer are retrained, ensuring exact removal with low cost. Experiments on four real-world benchmarks show that IsleNet attains up to 94% of full-graph accuracy while reducing unlearning time by up to an order of magnitude. Our code is publicly available at https://github.com/wenlu-lab/STGraphUnlearning.

Securing grid interconnection capacity has become a bottleneck for AI data center projects and can take longer than constructing the facilities themselves. This mismatch can delay deployment for years, making early interconnection planning essential. This paper develops ICP-AI, an interconnection capacity planning framework from a data center developer's perspective. The framework minimizes grid import capacity under a prescribed onsite investment budget while jointly sizing photovoltaic (PV) and battery energy storage system (BESS) resources and scheduling deadline constrained workload flexibility. A secondary refinement fixes the minimum grid capacity and selects the minimum-investment PV-BESS portfolio among solutions that achieve that capacity. The framework is evaluated using monthly composite stress profiles across varying temporal assumptions, load shapes, flexible load fractions, and deferral windows. Results show that interconnection capacity reduction depends strongly on the planning environment: at a $100M budget, it is about 6% for the high load factor baseline, exceeds 10% under monthly average solar availability, and reaches 13.3% for a more diurnal load. At a $10M budget, 5% flexible load with a 1 h workload deferral window reduces BESS capacity from 15.30 to 4.87 MWh while increasing capacity reduction from 4.43% to 4.84%. To test sensitivity to temporal compression, the model is also solved over the full 8,760 h chronology, which preserves the main capacity and flexibility trends. Overall, ICP-AI quantifies the interconnection capacity and infrastructure substitution value of workload flexibility, providing an investment-interconnection frontier to support capital allocation and early project planning in constrained grid environments.

Multimodal search agents answer visual questions by interleaving image understanding, web retrieval, tool use, and evidence synthesis. Strong systems exist, but in two expensive regimes: proprietary frontier models such as GPT-5 and Gemini, or large open vision-language backbones trained with substantial agentic data and reinforcement learning. We ask a different question: when released agent trajectories are distilled into much smaller backbones under a single-node budget, what is actually transferred? We study this with LiteSearch-VL, a low-compute recipe for Qwen3-VL-2B and Qwen3-VL-4B that uses only released OpenSearch-VL trajectories, parameter-efficient LoRA adapters, and synthetic step-level preferences: DPO on GPT-5-generated hard negatives targeting five local failure modes (premature answer, wrong tool, weak query, repeated query, ignored image). Across 12,400 GPT-5-judged rollouts on SimpleVQA, FVQA, LiveVQA, and VDR-Bench-testmini, the dominant effect is behavioral rather than a uniform accuracy lift: full-trajectory supervised fine-tuning transfers the agent contract, taking the 2B model from almost never emitting a usable answer (1,237/1,240 no_answer rollouts) to 28.4% macro Pass@1, matching or slightly exceeding the off-the-shelf 4B base (25.6%). Synthetic preference learning and compact tool distillation act as refinements rather than phase transitions (best 4B configuration: 30.8% macro Pass@1). Finally, a controlled VDR step-budget ablation shows that extra search turns convert abstentions into wrong_entity errors rather than correct answers, identifying answer verification, not search depth, as the next bottleneck for small multimodal agents.

Artificial Intelligence (AI) has long been inspired by studies of biological intelligence. Reinforcement learning, for instance, drew inspiration from studies involving animal learning and is now a powerful paradigm for solving many real-world problems. Recently, plant biologists have uncovered a wide range of complex behaviors in plants that enable them to flexibly adapt to variable environments. Here, we argue that such behavior can motivate new AI frameworks encompassing a range of problems overlooked by existing problem-solving frameworks such as supervised learning, tree search, and constraint satisfaction. We illustrate this idea with two examples of intelligent problem-solving in plants: (1) leaf mimicry in Boquila trifoliolata, a vine capable of altering its leaves' morphology to resemble those of multiple host trees simultaneously; and (2) coordinated root-shoot growth, wherein plants allocate resources across organ systems exploring distinct environments. While leaf mimicry is highly specific to Boquila, coordination of root-shoot growth is shared across most plants. For both examples, we capture underlying computational principles and identify problems fitting these frameworks that are currently unaddressed by AI. Finally, we outline preliminary task formulations and discuss how these formulations may be applied to non-plant problems.

Large non-uniform point sets make direct attention-based surrogate modeling costly for ship hydrodynamics. We introduce APPSolver, a point-wise flow-prediction framework built around Adaptive Patch Partitioning (APP), a deterministic quadtree representation for fixed two-dimensional horizontal slices extracted from ship CFD simulations. APP assigns finer patches near the hull and coarser patches farther away, downsamples patch contents, and recovers predictions to the full reference point set. Under a corrected protocol that constructs natural $(t,t+1)$ pairs before splitting, reuses training-set normalization statistics, and reports three model seeds, learned tokenizers are more accurate than APP-Transformer, and a persistence baseline has lower one-step MAE on all three ShipBench hulls. The supported benefit of APP is therefore computational rather than universal predictive superiority: on a representative DTC input, APP-Transformer requires 1.815 GFLOPs and 1.309 ms per model forward, while a matched ablation shows that adaptive partitioning reduces MAE by 16.4-24.9% relative to a uniform partition augmented with learned slicing. Condition encoders provide setting-dependent gains in leave-one-hull-out evaluation, but the current absolute next-state objective does not establish accurate long-horizon dynamics. These results characterize APP as a compact spatial representation with an explicit accuracy–efficiency trade-off. Code is available at https://github.com/wenhuahuo/APPSolver .

Large Language Models (LLMs) still exhibit limited capability in following complex instructions. While existing approaches often rely on preference learning to enhance this ability, they typically overlook the relationships between the permissible response spaces of different instructions, which restricts a model to align with subtle and diverse constraint variations. To address this, we propose Cross-Relational Preference Learning (CRPL), a novel framework for constructing preference data that explicitly models inter-instruction relationships through two key techniques: Cross-Relationship Perturbation and Cross-Region Pair Sampling. This enables the generation of more diverse preference data that captures a wide spectrum of constraint variations. Additionally, we introduce an atomic constraint-based verification mechanism to rigorously assess response satisfaction, ensuring high-quality preference pair construction. Extensive experiments across multiple preference learning methods (e.g., DPO, KTO), LLM backbones and four instruction-following benchmarks demonstrate that our approach achieves substantial improvements over prior baselines and exhibits strong generalization.

Gaussian process (GP) regression with a single global GP (GP-glo) incurs cubic computational cost, limiting scalability to large datasets. Product-of-experts GP models (GP-pro), which combine local GP models to capture global correlations, alleviate this computational burden. However, training local experts on disjoint data subsets can lead to overestimated posterior variances. We propose GP-pro-c, a product-of-experts GP model that calibrates these variances using an information-based method. The method exploits the monotonicity and submodularity of information gain in GPs to define a calibration ratio that reduces the posterior variance of individual local GP models. We evaluate GP-pro-c using negative log-likelihood (NLL), root mean squared error (RMSE), and expected normalised calibration error (ENCE). Experiments on four synthetic functions and six regression datasets show that GP-pro-c achieves average reductions of 2.3% in NLL and 12.0% in ENCE compared with the uncalibrated GP-pro model. The proposed method mitigates posterior variance overestimation while maintaining predictive accuracy and reducing computational complexity. GP-pro-c provides a promising approach for uncertainty estimation in scalable GP models and may serve as a useful surrogate model for Bayesian optimisation with high-dimensional and large-scale data.

Background: Patient details and acquisition metadata are important for clinical decisions, image quality control, and automated research pipelines, but may be missing or unreliable in imaging archives. Purpose: To develop and evaluate a fast open-source model that predicts patient and acquisition characteristics directly from CT and MR images. Materials and Methods: Separate 3D ResNet-10 ensembles for CT and MR were trained on 57,291 and 43,200 clinical examinations acquired from 2011 to 2025. Both predicted weight, height, age, sex, contrast presence, vertebral coverage, and image noise. The CT model additionally predicted scanner manufacturer, tube voltage, tube current, convolution kernel, and post-injection time; the MR model predicted sequence class. Performance was evaluated on internal CT (n=501) and MR (n=636) test sets and an external CT dataset (n=54). Results: Internal CT MAEs were 3.90 kg, 3.68 cm, and 4.42 years for weight, height, and age, with sex F1=0.990; corresponding MR results were 4.34 kg, 4.62 cm, 7.13 years, and F1=0.970. The CNN outperformed a segmentation-derived XGBoost baseline for all four core targets in both modalities (adjusted P<=.042). F1 scores were 0.963 for CT contrast, 0.953 for MR sequence, and 0.823 for MR contrast. External CT MAEs were 4.45 kg, 4.05 cm, and 5.17 years, with sex F1=0.971. CPU inference required 20 seconds for CT and 12 seconds for MR. Conclusion: One 3D multitask model per modality can rapidly recover patient and acquisition characteristics from heterogeneous CT and MR examinations. Models are available in TotalSegmentator: https://github.com/wasserth/TotalSegmentator

Operators remain essential to Remotely Operated Vehicle (ROV) operation, yet often suffer from low situational awareness and high workload, both of which negatively affect safety. This paper presents a cognitive architecture consisting of an ontology and multiple Large Language Models (LLMs) to assist the operator at all stages of the mission. Each LLM is grounded with domain-specific information from the ontology and given a simple role to create a system that can support the operator at all stages of an operation. We are aiming to prove that using the two together will allow decisions to be grounded in the relevant domain knowledge, but also benefit from the reasoning capabilities of the LLM. Our framework determines if a mission is possible for a given Unmanned Underwater Vehicle (UUV), performs mission planning, and executes a given mission in simulation. The operator can be involved in planning and execution, ensuring the resulting plan is valid and that the vehicle behaves safely during execution. We compare different LLMs, Llama3, GPT-OSS, and Qwen2.5, to determine which are best suited to the different roles within our framework. We find that GPT-OSS performs best for feasibility assessment, planning, and execution, while Qwen2.5 is best suited to identifying mission types from natural language input.

3D Gaussian Splatting has achieved remarkable success in photorealistic rendering, yet it suffers from severe overfitting and geometric artifacts in sparse-view scenarios due to the inherent deficiency of photometric supervision. Recent advances have attempted to regularize optimization by incorporating external priors, such as depth, point clouds, or diffusion models. However, these methods typically overlook the non-uniform distribution of supervision across the viewing space, resulting in limited specificity in prior use and primitive control. In this paper, we propose GSPotential, a framework that quantifies view-space supervision imbalance using a Camera Potential Field. Our key insight is to identify supervision valleys where photometric constraints are most deficient, and use the potential field to guide reconstruction from two complementary aspects. First, we devise a probabilistic spherical sampling strategy that places informative virtual cameras in low-potential regions. Point-cloud renderings from these views then provide targeted geometric guidance. Second, the same field provides a directional coverage cue for conservative Gaussian updates in weakly covered spatial sectors. Extensive experiments demonstrate that GSPotential achieves high reconstruction fidelity while maintaining competitive training efficiency.