Computer Science (arXiv)

Even well-aligned large language models confidently generate factually incorrect text, making hallucination a persistent reliability risk in high-stakes deployments. These models nonetheless carry linearly separable truthfulness signals in their internal representations. Existing white-box detectors, however, collapse this evidence to isolated components or a single depth, discarding discriminative information distributed across the full forward pass. We introduce HalluTracer, a detection framework that reads and aggregates truthfulness evidence across every layer of the forward pass before the model emits any answer token. A geometric analysis reveals that the per-layer signals are weakly correlated, so that simple depth averaging suppresses layer-specific noise and captures nearly all linearly accessible information. Across six open-source language models and five hallucination benchmarks, HalluTracer consistently outperforms matched white-box baselines, with gains ranging from one to fourteen points. Collectively, our work recasts hallucination detection from a layer-selection problem into a depth-aggregation problem governed by the geometric sparsity of the truthfulness signal.

Dexterous grasp generation that considers arm-related constraints is crucial in real-world scenarios involving arm environment collision avoidance, workspace boundary grasps, and consecutive grasping. Existing hand-centric grasp models, which primarily focus on the floating hand's pose, are insufficient for such cases. Conventional arm-aware methods either rely on rejection sampling to discard infeasible samples or require retraining on arm-specific data, leading to low sample efficiency under adverse conditions or limited generalization across different robots and environments. To overcome these limitations, this letter presents an arm-aware dexterous grasp generation framework that leverages pretrained arm-agnostic grasp models while integrating arm and environmental information only at inference time. Specifically, we formulate arm-aware constrained grasp generation as a joint optimization of hand pose and arm configuration, and derive closed-form gradients for arm-related constraints. Assuming the hand pose distribution is represented by a diffusion model, we prove that gradient-based optimization is equivalent to guided diffusion sampling, steering near-feasible samples toward the feasible region. Through comprehensive evaluation involving 10k objects across 6 scenarios, we demonstrate that the proposed framework generates feasible grasps in highly constrained settings with significantly higher probability, highlighting its advantages in real-world applications. Supplementary materials and appendix are available at https://arm-aware-dexgrasp.github.io/.

Large language model (LLM) agents may assist flight crews with complex decisions and task execution, but existing aviation evaluations centered on static knowledge do not support systematic testing of procedural execution and safety compliance in interactive environments. This paper presents the AeroCopilot Operational Environment (ACOE), a reproducible interactive virtual-cockpit test environment, and AeroCopilotBench, a two-tier aviation agent evaluation benchmark. Tier-1 evaluates aviation knowledge using 1,200 multiple-choice questions, while Tier-2 comprises 73 emergency and abnormal tasks derived from the manufacturers' Pilot's Operating Handbooks (POHs) and instantiated in ACOE. ACOE converts natural-language procedures into executable state transitions, final-state goal conditions, and hard safety constraints, enabling models to interpret cockpit state, diagnose faults, and operate aircraft systems through standardized tool interfaces. We establish a safety-gated evaluation framework in which a trajectory succeeds only when all task goals are achieved without violating any hard safety constraint, while safe goal progress and trajectory safety are measured separately. Across 12 models, the highest Tier-2 success rate is 72.6%, while static knowledge performance does not consistently translate into procedural execution. Analysis of 451 failed episodes from 3 representative models identifies recurring failures in procedural completeness, use of state feedback, and long-horizon execution management. These findings motivate state-aware agent orchestration, joint assessment of task completion and trajectory safety, and repeated regression testing. ACOE and AeroCopilotBench provide a reproducible foundation for testing knowledge application, interactive execution, and operational safety in aviation agents.

Direct communication between AI systems relies on natural language as an intermediate layer, incurring encoding/decoding overhead, token cost, and latency. We ask whether internal activation states can instead be transferred causally between different large language model (LLM) architectures via a learned projection, evaluated at three levels: representational similarity, cross-model retrieval from projected states, and end-to-end causal transfer via activation injection during generation. Using four architecturally diverse open-weight models (Qwen2-0.5B, Phi-3-mini, Mistral-7B, FLAN-T5-base), we find that representational alignment in trained models exceeds a random-initialization null baseline and is best captured by a rank-based metric (mutual k-nearest-neighbour alignment), more robust to activation-magnitude outliers than centered kernel alignment (CKA) or Procrustes analysis. A learned projection network retrieves the correct target-model representation from a held-out set well above chance for the three causal decoder-only model pairs (45-50% top-1 accuracy vs. 5% chance) but at chance level for the encoder-based FLAN-T5. Injecting projected activations into a target model during generation produces a statistically significant, pre-registered causal effect on retrieval-based output similarity for only one of the three decoder-only pairs (Qwen2-0.5B to Phi-3-mini: 23.3% vs. 0.0% under negative control, p=0.047, FDR-corrected); the two pairs targeting Mistral-7B show no such effect despite comparable representational alignment at the hidden-state level. We interpret these results as evidence for causal transfer of the representational vehicle, not of meaning, and conclude that end-to-end activation-state transfer between LLMs, as currently implemented, is architecture-dependent rather than universal.

Choreographic programming is a programming model for developing distributed applications where an entire communication protocol is written as a single program, which a compiler then projects to one process per participant. Choreographic programming abstracts over low-level network communication primitives such as sockets, and provides a high degree of safety guarantees with deadlock freedom ensured by construction. Mechanizing choreographies necessarily deals with both operations specific to distributed programming and standard (local) operations that also occur in non-distributed programs, as well as the typical issues of binding and substitution. We aim to sidestep the latter issues, thereby obtaining a more concise mechanization that focuses on the essential distributed aspects of choreographies. To this end, we use a method recently proposed by Thiemann to elegantly model deadlock-free processes in a dependently typed language: Using state transformers to represent the computations performed by each process. We bring the state transformer model to choreographies, allowing us to reduce the usual mechanization effort around binding and substitution, and to abstract over the details of the "local" aspects of the language. We mechanize in Lean a choreographic language that supports point-to-point communication, broadcasting, recursive procedures, and local stateful methods, allowing each participant to be assigned a different set of methods. We prove soundness and completeness of endpoint projection, establish deadlock freedom for the projected processes, prove confluence, and verify a Hoare logic for choreographies.

Pre-trained models (PTMs) provide a strong foundation for continual learning by offering stable representations that facilitate lightweight adaptation to new tasks. However, adapting well to each task does not ensure reliable inference over all learned tasks. Since task boundaries are often artificial and semantically entangled, an input from an unknown task can remain ambiguous even with strong PTM features, making cross-task prediction a key bottleneck. We propose Task-Anchored Inference Latent Shaping (TAILS), a lightweight post-PTM module that can be integrated into diverse continual learners and optimized through a decoupled step. TAILS uses fixed task anchors as persistent references to accumulated knowledge. It interprets each sample's feature representation relative to these references, then composes relevant evidence across tasks into latent recall. Rather than selecting a task-specific path or adjusting classifier outputs, TAILS uses latent recall to directly correct the feature representation before prediction. It therefore resolves cross-task ambiguity at the representation level, while leaving the original PTM, method-specific modules, and classifier unchanged. Extensive experiments across multiple PTM-based continual learning paradigms show that TAILS can improve classification and task-inference performance with modest parameter overhead and negligible inference cost.

Indic quality estimation (QE) and automatic post-editing (APE) data is spread across separate releases, so no single resource supports training and evaluation across tasks and language pairs on one footing. We consolidate the WMT 2020--2024 shared-task lineage with an extended English--Malayalam resource into \indicqe: $126{,}754$ instances over nine directional pairs, with up to four label types aligned on the same segment, a direct assessment, a human post-edit, word-level OK/BAD tags and an error explanation, and a test set stratified over four difficulty axes. On it, we benchmark six prompted LLMs and three COMET metrics on segment-level QE, and three systems on APE. Two of the axes are defined partly on the direct assessment and select a compressed slice of it, so each axis is compared against a control drawn from the same language pair with the same score distribution. Only one survives that control: segments whose holistic and token-level quality signals conflict are ranked worse than equally-scored segments of the same language, for all nine systems and all seven pairs that carry the axis. Annotator disagreement, which looks second-hardest without the control, has no effect with it. Few-shot prompting costs every model $\leq$ $3.4$B both correlation and output-format compliance. Within-language accuracy does not make scores comparable across pairs: of the three trained metrics, the one with the best within-language correlation loses most when the pairs are pooled. The benchmark and code will be released.

The combination of traditional statistical models and neural network (NN) components into semi-structured hybrid models is an intriguing approach to construct models that, ideally, combine traditional interpretability with the unprecedented flexibility of NNs. In order to preserve interpretability, it is usually necessary to restrict the NN components to prevent them from dominating the model. However, existing methods that enforce structural constraints on their NN components severely limit their models' flexibility; in contrast, methods that only enforce weak, indirect constraints lose meaningful interpretability. The method we propose therefore leverages invertible residual neural networks (i-ResNets) to equip generalized linear models with both nonlinear parameter estimation and a flexible correction of their distributional assumptions while always retaining stochastic monotonicity of the modeled distribution in the (formerly linear) predictor. The i-ResNets correspond to a controlled deviation from identity and by constraining their Lipschitz constant one can rigorously limit and quantify how far the hybrid model deviates from its traditional counterpart. This enables a user-specifiable compromise between flexibility and interpretability without limiting the structure of nonlinear and interaction effects that can be learned. Furthermore, we develop specific inherent interpretation techniques for our model and enforce model identifiability through an adapted post-hoc orthogonalization.

Local graph clustering aims to find a well-connected cluster near a given seed node without exploring the entire graph. A key step in the classic local clustering algorithm of Andersen, Chung, and Lang (ACL; Internet Math. 2007) is to approximate the PageRank vector from the seed node. Their local push method computes an ACL $\varepsilon$-approximate PageRank vector with teleportation parameter $α$ in $O\bigl(1/(α\varepsilon)\bigr)$ time. We give an algorithm that computes an ACL $\varepsilon$-approximate PageRank vector in $\widetilde{O}\bigl(1 / \varepsilon^2\bigr)$ time with high probability. This bound is independent of the graph size and has only a polylogarithmic dependence on $1 / α$, albeit with a quadratic dependence on $1 / \varepsilon$. As a direct consequence, we obtain a new running-time tradeoff between the target conductance and target volume in local graph clustering. Our method also applies to the optimization problem of $\ell_1$-regularized PageRank and computes an additive approximate minimizer with a polylogarithmic dependence on $1/α$, improving the $1/\sqrtα$ dependence in the previous bound of Martínez-Rubio, Wirth, and Pokutta (COLT 2023). Our algorithm is based on an intuitive process that maintains a growing active set of nodes: it performs push operations on the current set until convergence and then expands the set and repeats the process if necessary. We show that for each active set, the corresponding limiting state is the solution to a symmetric diagonally dominant (SDD) linear system on the set. We apply nearly-linear-time SDD solvers to these systems and prove that the approximation preserves the properties of the push process.

Video lane detection requires predictions that remain stable across frames, yet severe vehicle occlusions can break temporal cues. In streaming recurrent models, corrupted observations may enter the hidden state and produce errors that persist into later frames. Existing occlusion-aware refinements usually provide obstacle masks as auxiliary inputs, so the state-update path is only indirectly protected. We propose SIGMA-Lane, which treats this failure mode as state contamination in State Space Model (SSM)-based temporal modeling. SIGMA-Lane places occlusion-aware gates on the SSM write and residual-fusion paths, controlling how current observations enter temporal memory and are fused back after temporal propagation. After coordinate-consistent affine alignment, the model combines two complementary paths: SSM-consistent dual-gating for temporal filtering and Structural Spatial Retrieval (SSR) for recovering missing lane structure from aligned historical priors. Experiments on VIL-100 and OpenLane-V show improved temporal stability under heavy occlusion, with competitive F1 and mIoU scores.

Modern LLM serving deployments must simultaneously satisfy heterogeneous service-level objectives (SLOs) across a diverse population of user tiers, ranging from latency-critical API calls to background batch processing. Llumnix introduced a dynamic, migration-capable multi-instance scheduler for LLM inference that achieves load balancing, defragmentation, prioritization, and auto-scaling through a unified "freeness" metric. However, Llumnix's priority model is restricted to two levels (high and normal), an abstraction too coarse to express the richer SLA classes common in production deployments. In this work, we extend Llumnix's priority model to support an arbitrary number of tiers and evaluate the effects of this extension under three realistic priority distributions (uniform, Gaussian, enterprise) using Vidur, a high-fidelity LLM inference simulator. We implement per-tier headroom with exponential decay, tier-aware dispatch ordering, and the full Llumnix migration pipeline inside Vidur's hierarchical scheduling framework. We compare our extended scheduler against INFaaS (global routing baseline), vLLM, Orca, and Sarathi-Serve (per-replica baselines), sweeping priority levels from 1 to 10. Our experiments demonstrate that four priority tiers yields the best cost-effectiveness tradeoff, achieving prefill mean speedups of up to 8.3x and end-to-end P99 speedups of up to 3.1x over INFaaS with cost-per-latency improvements of 46 to 68%, while preserving strong SLO differentiation across tiers. We further show that the system sustains these gains at 10 priority levels without tail latency collapse, with overhead concentrated in the prefill phase.

Allocation schemes that greedily maximize a readiness metric over the actuator fiber bundle of an overactuated multirotor produce commands that jump between disconnected optimal strata, demanding actuator rates no motor can deliver; effort-minimizing schemes are continuous but cannot guarantee that wrench-rate authority stays above any certified level. We reconcile the two by treating authority as a forward-invariant quantity: a control barrier function on the log-determinant of the drag-aware actuator-authority co-metric, enforced at torque level by a quadratic program in the allocation null space. A single design inequality renders the certified set compact and strictly interior to the actuator box, with the readiness cost of any rotor deactivation given in closed form as $\ln(n/(n{-}m))$ for symmetric designs. Tracking is sacrificed only through an explicit alignment ratio, with wrench error bounded by $\mathcal{O}(ρ^{-1/2})$ and a robust variant handles motor-parameter uncertainty with a closed-form floor shift independent of the airframe matrix. On a hexarotor and a fully-actuated octorotor the closed-form gap matches simulation to machine precision; in the authority-scarce regime greedy maximization violates the certified floor and commits wrench errors up to eighty times larger than the proposed filter, which holds invariance of the certified set at negligible tracking cost.

Keystroke dynamics (typing patterns) can be used as a behavioural biometric modality for user authentication, with applications such as fraud prevention. While the modality has been shown to work well for single device authentication, its application to cross-device scenarios is more challenging. Dynamics learned on one device (eg., phone) may not be directly applicable to authentication on a secondary device with a different form factor (eg., tablet) due to changes in typing patterns that can lead to distribution drifts. To address this, we propose a cross-device user authentication system based on inductive transfer learning, where keystroke dynamics learned on one device are adapted to a secondary device. The adapted data is then combined with necessarily limited training data for the secondary device, which is used to robustly train a binary classifier. Furthermore, an extended set of keystroke features is used to better capture discriminative dynamics. Experiments on the BBMAS dataset show that proposed system achieves an equal error rate of 14.2% for the cross-device scenario, surpassing state-of-the-art methods.

On-policy distillation (OPD) aligns a student model with a teacher's logit distribution on student-generated trajectories. This approach has achieved strong empirical gains and can often surpass conventional off-policy distillation with substantially less data. However, standard token-level OPD can provide only fragmented corrections along an erroneous student trajectory and cannot unfold a complete and correct repair path. Motivated by this limitation, we propose \emph{Step-Level On-Policy Distillation} (SOPD), which combines the long-horizon correction of supervised fine-tuning (SFT) with the on-policy advantage of OPD to provide step-level supervision over complete student-generated trajectories. We show that, at different limits of step length, SOPD reduces to SFT or approximates OPD. Compared with SFT, the teacher responses in SOPD are conditioned on student trajectories and therefore align more closely with student-visited states; compared with OPD, SOPD provides longer-horizon corrections rather than fragmented token-level guidance. Across both reasoning and agent tasks, SOPD substantially outperforms conventional SFT and OPD. For example, on ALFWorld, SOPD improves the average success rate by 13.4 points over Vanilla OPD. We hope this work offers a new perspective for future research on distillation methods.

Referring Video Object Segmentation (RVOS) aims to segment referred objects at the pixel level in video sequences based on natural language descriptions. Existing methods typically introduce motion information within a unified cross-modal temporal modeling framework, where language cues are used for target localization and segmentation. However, the dependency of expressions on motion semantics is not explicitly modeled, making it difficult to adaptively adjust the use of motion information according to different semantic requirements. To address these issues, we propose an Expression-driven Motion Calibration (EMC) framework for RVOS that explicitly unlocks and leverages the motion semantics within expressions. The proposed method extracts interpretable motion control signals from expressions via a Motion Signal Processing (MSP) module, and employs a Motion Influence Calibration (MIC) module to adjust the contribution of motion cues during temporal decision making. In addition, a Semantic Temporal Stage Construction (STSC) module is introduced to build expression-relevant temporal stages, providing a compact temporal candidate space for motion calibration. Through extensive evaluation on six standard benchmarks, including Ref-YouTubeVOS, Ref-DAVIS17, MeViS (valid/valid$^u$), A2D-Sentences, and JHMDB-Sentences, the superiority of our method is validated. We will release the code on https://github.com/Jeven7/EMC.

Public administrations and private companies have announced plans to deploy networking infrastructure to support future robotic and human presence on or near space targets, such as the Moon and Mars. While using an IP protocol stack for deepspace communication had been neglected, recent events have motivated the reconsideration of IP to enable the Interplanetary Internet. This new paradigm facilitates the integration of IP-based Internet of Things (IoT) protocols for deep-space environments. This paper illustrates the similarities between deep-space and IoT scenarios, presents related IETF standardization work, and discusses opportunities and future directions for IP-based IoT protocols in the Interplanetary Internet.

Instruction-based general video editing seeks to unify diverse editing operations within a single, intuitive interface. Existing approaches often rely on resource-intensive conditioning, using either heavyweight branches or costly source concatenation. Is there any efficient way to model editing intent? Thus, we introduce GRNEdit, a lightweight two-stage framework. GRN inspires our approach by encoding visual semantics through combinations of bits. Through task-specific fine-tuning, we take this representation further and recast editing semantics as local retain-or-flip decisions over individual bits. Source information is consequently modeled as coordinate-wise evidence supporting the observed binary states, while the GRN backbone remains responsible for resolving their global composition into coherent generative semantics. In Stage I, a compact encoder translates discrete source codes into continuous evidence signals, which GRN assimilates throughout binary refinement. Inspired by null-prompt training for classifier-free guidance, we further assign the null condition an editing-specific meaning: an empty instruction denotes no edit and is supervised through source reconstruction. This identity pathway not only implicitly strengthens evidence utilization and content preservation in Stage I, but also produces a source-preserving state in the same representation space as the edited state. Stage II can therefore directly compare each edited state with its source-preserving counterpart and use their discrepancy to revise unresolved target-bit decisions. Trained on only 0.6M pairs with less than 3\% conditioning parameters, GRNEdit-2B and GRNEdit-8B achieve scores of 4.03 and 4.18 on OpenVE-Bench. The 2B model outperforms multiple 14B open-source editors, while the 8B model performs on par with leading open-source editors.

Multi-unmanned aerial vehicle (UAV) systems provide scalable service platforms for large-scale environmental tasks, such as grassland ecosystem restoration. However, coordinating fleet operations requires solving the restoration area maximization problem (RAMP). This non-linear combinatorial optimization challenge is complicated by payload-dependent energy dynamics and heterogeneous ecological degradation. We propose a novel knowledge-guided collaborative bilevel formerpointer reinforcement learning framework (KC-BFPRL) to address this complexity. Using a hierarchical paradigm, KC-BFPRL decomposes RAMP into global task allocation and local restoration planning, with the latter further divided into upper-level trajectory planning and lower-level restoration area allocation. Our specialized architecture pairs featuring a Transformer-based encoder that fuses static environmental features with dynamic UAV states, and a Pointer Network decoder trained via a robust actor-critic framework. By embedding ecological priority rules and heuristic logic, KC-BFPRL achieves a structured warm-start, solving the RL cold-start problem while ensuring strict constraint satisfaction. Extensive experiments demonstrate that KC-BFPRL consistently outperforms state-of-the-art baselines, achieving superior objective values and efficiency. It maintains a $0.00\%$ optimality gap in the most complex scenarios U8-R160 and operates nearly three times faster than MAPDP, validating its robustness, scalability, and real-time applicability for large-scale automated ecological restoration.

We present LaGSplat (Latent Lagrangian Gaussian Splatting), a framework that infers interactive, physics-governed dynamics from one or a few monocular videos. At inference it lets a user push on the filmed object, rigid or deformable, with an external force that was never measured, annotated, or seen during training. This is possible because a low-dimensional latent state $\mathbf{q} \in \mathbb{R}^d$ plays two roles at once: it is the generalised coordinate of a learned dissipative Lagrangian and the conditioning variable of a Gaussian Splatting decoder. The inductive bias of this decoder, whose primitives are explicit points $μ_i(\mathbf{q})$ that move with the object, is what lets a force $f$ applied in the image pull back into a latent generalised force $J(\mathbf{q})^\top f$ and enter the equations of motion, which pixel-space (CNN) or neural-field (NeRF) decoders cannot do. We validate LaGSplat on test cases of increasing difficulty, from rigid to deformable and from autonomous to forced real systems, combining monocular video and sensor measurements. We further demonstrate interactive use: forces of arbitrary magnitude and direction can be applied to the reconstructed object at any time, its response rendered in real time, in 2D or 3D. Assuming a dissipative Euler-Lagrange equation over a few generalised coordinates trades generality for a bounded, plausible response to unseen forces, where an unconstrained predictor diverges.

The spread of misinformation on social networks poses a significant challenge to online communities and society at large. Not all users contribute equally to this phenomenon: a small number of highly effective individuals can exert outsized influence, amplifying false narratives and contributing to significant societal harm. This paper seeks to mitigate the spread of misinformation by enabling proactive interventions, identifying and ranking users according to key behavioral indicators associated with harmful content dissemination. We examine three user archetypes -- amplifiers, super-spreaders, and coordinated accounts -- each characterized by distinct behavioral patterns in the dissemination of misinformation. These are not mutually exclusive, and individual users may exhibit characteristics of multiple archetypes. We develop and evaluate several user ranking models, each aligned with a specific archetype, and find that super-spreader traits consistently dominate the top ranks among the most influential misinformation spreaders. As we move down the ranking, however, the interplay of multiple archetypes becomes more prominent. Additionally, we demonstrate the critical role of temporal dynamics in predictive performance, and introduce methods that reduce data requirements by minimizing the observation window needed for accurate forecasting. Finally, we demonstrate the utility and benefits of explainable AI (XAI) techniques, integrating multiple archetypal traits into a unified model to enhance interpretability and offer deeper insight into the key factors driving misinformation propagation. Our findings provide actionable tools for identifying potentially harmful users and guiding content moderation strategies, enabling platforms to monitor accounts of concern more effectively.

This study presents estimates of the global expenditure on article processing charges (APCs) paid to 14 publishers for open access (OA) between 2019 and 2025. APCs are charged for publishing in fully OA journals (gold) and making individual articles OA in subscription journals (hybrid), but how much is paid, and for which articles, is not publicly known. We therefore curated an open dataset of publicly listed APC prices from 14 academic publishers (ACS, CUP, De Gruyter, EDP, Elsevier, Frontiers, IEEE, IOP, MDPI, OUP, PLOS, Sage, Springer Nature, and Wiley) and combined it with counts of OA articles from OpenAlex. We estimate that \$15.08 billion (in 2025 USD) was spent globally on APCs between 2019 and 2025. Adjusted for inflation, annual spending quadrupled from \$0.9 billion in 2019 to \$3.7 billion in 2025, with >85% concentrated among a few large publishers. Hybrid OA fees exceed gold fees, and the median fee paid is higher than the median price listed for both. Our approach addresses major limitations in previous efforts to estimate APC spending, offering much-needed insight into an opaque aspect of scholarly publishing, especially as transformative agreements make it more challenging to understand the costs of publishing OA.

Streaming video understanding demands direct responses from the causally observed prefix of an unfolding video. Existing systems add inference-time memory, retrieval, and compression, yet a training-free sliding-window baseline already matches them. We therefore fix a memory-free recent-window protocol and ask how far post-training alone can go. Reinforcement learning with verifiable rewards fits this regime poorly, encouraging long ``think-then-answer'' generations, while on-policy distillation (OPD) supplies dense token-level teacher supervision on student trajectories but is stable only when both models train in thinking mode. These observations lead to \textsc{StreamOPD}, a recipe combining verifiable streaming-video data, thinking-mode OPD, and instruct-mode deployment. It raises StreamingBench from $77.9\%$ to $83.9\%$---within $0.3$ points of the 9B teacher---and improves OVO-Bench excluding its hallucination-detection subtask (HLD) by $9.1$ points under unchanged inference. As a teacher-privilege extension, \emph{Spatio-Temporal CueGate (ST-CueGate)} aggregates cue-versus-no-cue teacher likelihood ratios into a group-relative response score that reweights OPD. It reaches $71.9\%$ on OVO-Bench (excluding HLD) and $64.9\%$ on Video-MME, and is the only variant that stays above the base model on all four benchmarks. Replacing the teacher with a frozen copy of the student's initial policy---on-policy self-distillation---retains most of these gains and lifts HLD to $57.0\%$, above both the untrained student and the 9B teacher, so abstention loss is not intrinsic to the recipe. We provide a transparent and reproducible reference for open-source streaming-video research.

This first release of Prior Labs in relational learning shows our continued commitment to open science. We open-source three pieces of software that we expect to accelerate research in the field towards meaningful real-world impact. We aim to steer further development based on feedback from, and in collaboration with, the community. Given the early stage of development, our $α$-release targets researchers and early-adopting practitioners. Over the past years, a variety of datasets and tasks for relational learning have emerged, but the community has not converged on a reliable, reproducible way to compare different methods on these tasks. Our $α$-release, RelArena-$α$, provides a unified framework for running and comparing baselines on RelBench v1 by standardizing data loading, evaluation protocols, tuning regimes, and support for systems with custom tuning, inspired by established tabular benchmarks such as TabArena. We plan to work with the research community to further develop RelArena-$α$ into a catalyst for progress in the relational learning community. We release the initial version of TabPFN-Rel, a purpose-built relational harness for TabPFN-3. Currently ranked first among models on RelArena-$α$, TabPFN-Rel makes key improvements upon RDBLearn. Beyond its ranking, TabPFN-Rel serves as a strong baseline, adding to the growing evidence that flattening a relational database into a single table remains competitive with specialized relational architectures on real-world tasks. To facilitate adoption of relational learning methods in research and industry, we release an initial $α$-version of our Relational Predictive Interface, RPI, an open-source, model-agnostic interface that enables early adopters to easily define problems on new databases and apply any model implemented in RelArena-$α$, including TabPFN-Rel, to these problems.

Recent advances in Generative Artificial Intelligence (GenAI) have substantially improved the ability of large language models (LLMs) to generate and explain source code. However, their performance on authentic object-oriented programming (OOP) assessments remains insufficiently understood. This study evaluates five widely used GenAI systems, ChatGPT-5.2, DeepSeek-V3, Gemini 2.5 Flash, Claude Sonnet 4.5, and M365 Copilot, using programming tests and examination tasks from an introductory university OOP course. The generated solutions were assessed using the same grading criteria applied to students and compared with historical student results from the same course, as well as findings from the previous year. Common errors were also analyzed to identify recurring limitations across models. All evaluated GenAI systems achieved higher scores than the average student cohort and frequently obtained full marks on longer programming tasks. Nevertheless, they occasionally produced non-compiling code and continued to struggle with advanced OOP concepts, particularly interfaces, abstract classes, and certain inheritance-related tasks. Performance was also limited on graphics-related questions involving image interpretation. Compared with the previous year, the evaluated systems demonstrated noticeable improvements across most assessments while exhibiting several recurring error patterns. The findings provide an updated evaluation of the capabilities and limitations of contemporary GenAI systems on authentic introductory OOP assessments. They also offer evidence that can inform the design of programming assessments, the responsible integration of GenAI tools into software engineering education, and future studies evaluating the evolution of AI-assisted programming.

Large Multimodal Models (LMMs) for video reasoning have long been hindered by the high computational cost of processing vast amounts of visual information. This dilemma motivates the transfer of the reasoning capabilities of large models to smaller, more efficient ones. On-Policy Distillation (OPD) offers a promising solution by matching output-token distributions along student-generated trajectories. However, video reasoning often depends on evidence accumulated across multiple frames. In this context, output-level supervision only captures information expressed through token predictions and does not directly constrain the latent representations formed during reasoning. To address this limitation, we propose Latent-OPD, which augments OPD with trajectory-level latent distillation. Specifically, our method focuses on the position at the end of each trajectory, where hidden states effectively summarize the accumulated visual evidence and reasoning context. Furthermore, we introduce a progressive teacher-lookahead strategy, which aligns middle-to-late student layers with increasingly deeper teacher layers. Experiments on six video reasoning benchmarks show that Latent-OPD consistently outperforms output-only OPD. Notably, the improvements are particularly pronounced in scenarios with limited frames, long videos, or tasks requiring complex evidence aggregation. These results establish Latent-OPD as a highly effective approach to frame-efficient video reasoning.