Computer Science (arXiv)

Recent agent benchmarks increasingly ground evaluation in executable environments, from code repair to web navigation, app APIs, and function calling. Yet completing consequential work beyond code requires more than producing a plausible response or valid tool call: agents must gather missing information over multiple turns, follow domain policies, coordinate dependent tools, and realize the correct persistent state transition without collateral effects. In this paper, we introduce Thinkingbox, a sandbox for tool-agent-user interaction that provides isolated MCP-compatible tool sessions, complete execution traces, and outcome evaluation over terminal backend state. Built on this sandbox, Thinkingbox-bench contains 507 policy-conditioned workflows across numerous scenarios, including retail, hospitality, auto insurance, neobank internal IT, and consulting IT/HR support. Each attempt is evaluated by task-specific executable checks that accept valid trajectories while rejecting wrong, missing, or extra effects; designated tasks additionally check required properties of the final response. Across proprietary and open-weight models, the strongest achieves 65.36% pass@1, but only 25.25% pass^20. Moreover, many failed trials show clean termination and valid state-changing actions, showing that response or tool-call-level signals are not clear proxies for end-to-end task completion. Thinkingbox-bench reveals a large gap between occasionally finding a successful trajectory and reliably completing stateful business tasks. We release both Thinkingbox and Thinkingbox-Bench: https://github.com/microsoft/thinkingbox

This paper presents an open-source software stack that restores ROS 2 support for the Franka Emika Panda robot while resolving the long-standing unreliability of its external position control interface. We first analyze the root causes of unstable position control and show that the observed vibrations and protective stops arise from the timing of the external control loop and sampling jitter, rather than from limitations of the robot itself. Building on this analysis, we introduce an asynchronous hardware interface that decouples real-time communication from the ROS 2 control loop, a rate-matching mechanism for slower command sources, and a position-domain reference generation strategy that produces reliable, smooth position commands. Experimental validation shows that the proposed architecture reliably tracks velocity references by reducing motion artifacts introduced by the official implementation, and the stack is validated across motion planning, compliance control, position-controlled manipulation, and haptic teleoperation on two independent Panda platforms. By restoring a modern, reliable, and open ROS 2 ecosystem for the Panda, this work lowers the barrier to developing safe, responsive, and reproducible human-robot collaboration applications that integrate planning, perception, interaction, and shared autonomy. Code and videos are available on our website at https://sites.google.com/view/fer-ros2/.

Multimodal LLMs can see a document, but they often can't read it reliably. Small text, tables, visual cues, and topological elements still trip them up under direct visual inference, even when the page is already sitting in the model's context. Most document-VQA systems treat perception as fixed: they encode the page once, ask the question, and answer from whatever the model happened to extract in that single fast pass. We think document VQA needs slower, more deliberate perception: rather than answering from one fixed encoding, the model should spend a bit of extra compute at inference time working out what to look at next, and only then answer. We build this into \textbf{Q-Guide}, a small agent that reads a question, works out what evidence it is still missing, and calls targeted tool(s) to recover it---reading text where text is needed, zooming in where detail is needed, or grounding a region where position matters. On DocVQA2026 and Manga109, Q-Guide outperforms both direct prompting and recent multi-agent document systems ($65.0\%$ vs.\ $40.0\%$ on DocVQA2026, $32.4\%$ vs.\ $24.4\%$ on Manga109), and the improvement holds across three Claude backbones (Opus 4.6, Sonnet 4.6, and Opus 4.5). We find that accuracy scales with the perception budget---most of the gain appears within two to three deliberate rounds---and that the gain comes from directing perception to the right place, not from complex control logic: adding planners, routers, or multiple collaborating agents does not help.

Full-cycle biventricular geometry is essential for characterizing cardiac function. However, dense and temporally consistent 3D+t biventricular meshes are not routinely available, whereas end-diastolic (ED) anatomy can often be obtained reliably. We therefore investigate full-cycle biventricular motion synthesis from a single ED mesh. This task is challenging because cardiac deformation is spatially heterogeneous and phenotype dependent, while conventional global generative models often obscure localized motion patterns. In this study, we propose a region-specific and phenotype-adaptive framework that integrates motion-informed functional parcellation with conditional latent flow. A functional partition learned from reconstructed motion organizes the ventricular surface into regions with coherent dynamics and enables topology-aware regional feature exchange. A phenotype-conditioned rectified-flow model subsequently maps the ED anatomy to full-cycle motion latents through fine-grained conditioning and prototype-routed motion adapters. An optional control branch further incorporates available motion descriptors for controllable synthesis. Experiments on ACDC, M\&Ms, and M\&Ms-2 demonstrate consistent improvements in geometric accuracy and functional fidelity. Under ED-only synthesis, our method achieves biventricular ASSD, HD95, and vRMSE of \(1.49\pm0.34\)~mm, \(3.77\pm1.06\)~mm, and \(3.31\pm1.03\)~mm, respectively, outperforming all competing methods. Complementary functional and robustness evaluations further demonstrate that the synthesized sequences preserve physiologically plausible ventricular dynamics and generalize across cohorts and disease phenotypes. The code will be released publicly upon acceptance of the manuscript for publication.

Large vision-language models (LVLMs) have achieved remarkable progress in video understanding and reasoning. Despite extensive studies on text- and image-based jailbreaks, video jailbreaks against LVLMs remain largely unexplored. Existing video jailbreak methods mainly manipulate textual content embedded in videos, while overlooking how such information is organized over time. Our analysis reveals that jailbreak effectiveness depends not only on the semantics of textual information but also on its temporal presentation, including duration and timing-slot allocation. Motivated by this finding, we use subtitles, which are common in real-world videos and allow semantic content to be presented under precise temporal control without appearing visually intrusive, as a natural attack medium. Based on this insight, we propose TempJail, a black-box video-based jailbreak framework that constructs query-aligned dialogue-style subtitle sequences and optimizes their temporal scheduling to exploit temporal vulnerabilities in LVLMs and elicit responses that satisfy the harmful intent of the source query. Extensive experiments on four representative LVLMs and two datasets demonstrate that TempJail achieves the highest attack success rate across all evaluated model--dataset settings, outperforming the strongest baseline by 53 and 18 percentage points in dataset-averaged ASR on GPT-5 and Gemini 3.5-Flash, respectively.

Transmissive reconfigurable intelligent surfaces (T-RISs) integrated into the transmitter provide a viable realization of tri-hybrid multiple-input multiple-output (MIMO), where spatial processing is distributed across the digital, analog radio-frequency (RF), and electromagnetic (EM) domains. However, unlike conventional RIS-assisted links, a transmitter-native T-RIS directly participates in radiation formation, making T-RIS front-end modeling and weighted sum-rate (WSR)-oriented joint precoding challenging under practical hardware constraints. This paper develops a unified modeling and precoding framework for transmitter-native T-RIS tri-hybrid multi-user (MU) downlink transmission. Specifically, starting from a continuous-field description, the received field is characterized by the interaction among the feed array, the programmable T-RIS aperture, and the user-side propagation, leading to a cascaded baseband input-output model for MU precoding. Furthermore, the same representation is instantiated in the Fresnel, Fraunhofer, and mixed-field regimes, so that near-field focusing and far-field angular steering can be handled within one front-end model. Additionally, a WSR maximization problem is formulated over the digital precoder, analog network, and T-RIS coefficients under power and quantization constraints. A two-level solver is then developed by coupling outer weighted minimum mean-square error (WMMSE) updates with WMMSE-induced aperture-field shaping and hardware projection. Simulations validate the modeling accuracy and convergence, and show that the proposed full tri-hybrid design improves WSR over baselines while suppressing mixed-field cross-regime leakage.

We introduce RecPFN, a prior-fitted network that brings in-context learning to sequential recommendation. RecPFN is pretrained entirely on synthetic clickstream environments sampled from a broad structural causal prior, enabling it to amortize Bayesian-style inference from a small support set. At inference, a lightweight decoder-only transformer conditions on a handful of domain sequences and produces next-item predictions for queries in a single forward pass, without any weight updates. Across eight public benchmarks, RecPFN achives state-of-the-art zero-shot performance while remaining strongly competitive with supervised methods in low-compute and low-data regimes. It is deployment-efficient and robust to domain shift, outperforming strong zero-shot baselines that rely on large real-interaction corpora. RecPFN provides a practical path toward generalizable, data-efficient recommenders and opens avenues for richer priors, longer-context ICL, and multimodal extensions. Code for training and evaluation is publicly available at https://github.com/SAP-samples/tabular-ai-recpfn/.

Accurate performance analysis for near-shore maritime wireless communication is essential for ensuring robust and reliable operations. However, existing analytical models often rely on oversimplified propagation assumptions, such as a perfectly smooth sea surface, which fail to capture the full dynamics of the maritime channel. In this paper, we develop a physically grounded analytical framework using stochastic geometry that bridges this gap. The spatial distribution of vessels is modeled as a non-homogeneous Poisson point process to reflect realistic near-port densities. We replace the idealized smooth-sea assumption by deriving a novel reflection coefficient from the classical Rayleigh criterion, which explicitly links the path loss to the significant wave height. Integrating this roughness-aware channel model into the stochastic geometry framework, we derive new analytical expressions for the uplink coverage probability and average ergodic rate, providing the first tractable characterization of aggregate interference under such dynamic conditions. The analysis reveals a sea-state-dependent reliability--capacity trade-off: roughness-induced attenuation of the coherent specular reflection can suppress destructive-interference nulls and improve reliability-oriented coverage, while reducing high-SINR and average-rate performance. Available measurements support the underlying roughness-sensitive reflection mechanism, but direct VHF validation under rough sea conditions remains unavailable; the corresponding rough-sea results are therefore interpreted as model-based predictions. A cross-frequency ablation further confirms the wavelength dependence of the roughness effect and shows that the reflection coefficient must be evaluated for the operating frequency.

Vision-language-model-based embodied agents can complete instructed tasks but often violate safety constraints in the process, a problem recently framed as interactive safety. Training such agents to act safely is difficult, since safety and task success are distinct objectives, and safety arises only at a small number of safety-critical steps within a trajectory. Standard supervision is insufficient: imitating safe trajectories teaches behavior without explaining why it is safe, and contrasting arbitrary safe and unsafe trajectories mixes the safety signal with unrelated differences. We propose SafeBranch, a framework that aligns an embodied actor on safety through branch pairs constructed from the actor's own unsafe rollouts via environment rollback. SafeBranch rolls each unsafe rollout back to the safety-critical step that caused the violation, queries the actor for a safe alternative, and pairs the original action with the alternative so that the two branches differ only at that step. The trained actor acts safely at deployment with no critic in the loop. On IS-Bench, SafetyALFRED, and out-of-distribution variants with unseen tasks and objects, it handles safety reliably without sacrificing task success, achieving roughly ten times more safe successes than the untrained baseline on the unseen-object variant.

In this paper we introduce Hippogriff, a language with a module system that unifies syntax between the core level and the module level. Hippogriff's type theory is dependent, with modularity features enabled via a universe of small types, but Hippogriff still supports general recursion without making typechecking nonterminating. This paper contains two halves. In the first half, we describe Hippogriff and its implementation. In the second half, we build categorical semantics for our use of dependent types that justify the use of general recursion at the value level. Specifically, we use an extension of the second-order generalized algebraic theory framework to include a synthetic phase distinction, and this allows us to make a mathematical connection between dependent type theories and split-context type theories (like System F). Included as supplements are a Haskell implementation of Hippogriff and an appendix describing the full type theory of Hippogriff using a second-order generalized algebraic theory with phase distinction.

Watermarking is a central tool for provenance in generative models, yet its application to multivariate time series remains hindered by reliability failures under post-editing attacks. We show that existing detectors, which rely on globally coupled re-encoding, suffer from bidirectional drift of the null distribution: post-editing attacks can shift the z-score of non-watermarked samples in either direction, invalidating clean-calibrated thresholds. We argue that this instability is a property of the re-encoding, and that reliable detection requires each recovered unit to depend only on a bounded temporal neighborhood. Guided by this principle, we propose L-VQVAE, a generative model in which each discrete token is produced from a short contiguous window, and LVQMark, a watermarking method over this token space that combines logit-bias injection with robust re-encoding for attack-time detection. Experiments on four benchmarks spanning finance, energy, and neuroimaging show that our approach preserves generation quality while stabilizing both detection power and false-positive behavior under post-editing attacks.

The typical training process of a multimodal large language model (MLLM) involves adapting both the language model backbone and the projector between the backbone and a modality-specific encoder. We ask whether fine-tuning the backbone of an MLLM is necessary to adapt it to a new modality. Through experiments on 3D MLLMs, we find that training only the projector is sufficient to achieve strong multimodal performance relative to existing baseline models and our jointly trained MLLMs with the same encoder and backbone. We also show that joint training leads to undesirable drift in existing capabilities of the language model, which projector-only training avoids by definition. Furthermore, projector-only training has approximately twice the training sample throughput of joint training. We validate our findings across different language model backbones via 3D classification and captioning benchmarks as well as standard benchmarks evaluating language, vision, and spatial reasoning capabilities.

Streaming video understanding requires models to causally update state as video arrives and organize growing history into semantic units that can evolve, persist, and be recalled under bounded computation and memory. This challenge is pronounced in live soccer commentary, where a system must describe completed events, summarize recent play, recall earlier events, or remain silent using only information available before each utterance. We present StreamSoccer, an event-driven system that uses event memory as its intermediate representation. A fixed-budget active memory integrates the stream; completed event states are retained locally and consolidated into retrievable historical records. A unified generator uses current, recent, and historical context to produce three commentary modes, while a rule-assisted scheduler selects a mode or silence. Unlike streaming video-language models organized around frames, visual tokens, or caches, and soccer-commentary methods based on predefined clips or output timestamps, StreamSoccer explicitly models event lifecycles. We construct a three-track streaming soccer commentary dataset and a layered evaluation protocol. At common reference anchors, StreamSoccer obtains CIDEr scores of 38.62, 23.96, and 17.39 for current-event, recent-window, and historical-memory commentary, ranking first on the current-event and historical-memory tracks and second on recent-window. Controlled ablations show that local completed events improve all tracks and that the full system performs best on all three. Across 174 raw-video runs on 58 matches, per-minute RTF p95 ranges from 0.10 to 0.22 without sustained growth with match history. These results indicate that event memory supports streaming soccer commentary across temporal scopes while controlling long-history computation.

Reference-based diffusion stylization requires separating target geometry from transferable appearance. Existing tuning-based methods often rely on aligned content-style-target triplets or auxiliary visual encoders, which increases data cost and can transfer unintended scene structure from the style reference. We propose SEFS (Style-Encoder-Free Stylization), a style-encoder-free conditioning framework for diffusion transformers. SEFS forms style tokens from stochastic low-resolution crops of single training images. This crop bottleneck preserves local appearance statistics such as palette, stroke, texture, and material, while reducing access to global layout cues. Target content is encoded by edge and segmentation cues and fused with the noisy latent through parameter-efficient trainable projections. We add style-to-denoising re-normalization for token-statistic alignment and cross-block skip fusion for spatial detail. SEFS trains on unpaired single images; the frozen diffusion VAE is used only to place image conditions in the latent space. On artistic stylization benchmarks, SEFS improves content consistency and leakage diagnostics while retaining reference-style affinity, and ablations support the crop-resolution, re-normalization, and skip-fusion choices. The code of SEFS will be made publicly available.

Reliable underwater robotic perception remains difficult because optical imagery degrades under turbidity, wavelength-dependent attenuation, low illumination, scattering, and blur. Although sonar provides complementary information that is less affected by optical visibility, prior visual-sonar research has largely focused on feature alignment and nominal detection performance. We investigate cross-modal robustness as visual reliability deteriorates and assess whether pretrained visual foundation-model representations can be complemented by sonar under severe degradation. We use frozen DINOv2 as the visual encoder and construct a controlled five-level benchmark ranging from clean to extreme visual conditions. We compare conventional visual detection, frozen foundation-model representations, sonar context, fixed multimodal fusion, clean-trained adaptive gating, and degradation-aware gated fusion. Our method trains the fusion mechanism across the full range of degradation while keeping the visual and sonar encoders frozen, allowing modality contributions to adapt without fine-tuning the pretrained backbone. Under extreme combined degradation, the DINOv2 baseline achieves 0.4610 balanced accuracy, while degradation-aware visual-sonar fusion reaches 0.6152, a 33.5% relative improvement. The learned sonar contribution increases from 14.2% under clean conditions to 41.3% under extreme degradation, demonstrating adaptive redistribution of cross-modal reliance. Fusion provides the largest gains under severe turbidity and blur, whereas color attenuation alone yields little additional benefit. These results show that foundation-model representations remain valuable but insufficient under severe information loss, and that explicitly adapting fusion to modality reliability can improve robust underwater multimodal perception.

Radio-frequency (RF) radiance-field modeling is essential for wireless network optimization and sensing, yet remains challenging in dynamic and unseen environments. Existing learning-based methods synthesize RF fields from sparse measurements, but most struggle to generalize to dynamic and unseen environments. To address this limitation, we propose RFWM, a physics-guided RF world model that maps multimodal physical conditions like visual dynamics and AP configurations to spatiotemporal RF fields. RFWM adopts a two-stage training strategy with physics-guided priors and constraints. In the first stage, RFWM adapts a pretrained visual diffusion backbone to RF trajectories to predict RF sequences from a few past RF inputs, while conditioning the backbone on a Friis-guided prior for coarse attenuation guidance. In the second stage, RFWM learns the physical-to-RF mapping by training a ControlNet from scratch and fine-tuning the RF-adapted backbone, while six physics-guided regularizers enforce fine-grained propagation consistency. Cross-height heads then jointly generate RF trajectories at queried receiver heights in one forward pass. We construct a new benchmark of 7,715 sequences averaging 33 frames across 115 environments for dynamic RF-field generation. Experimental results show that RFWM improves MSE by approximately 7 dB and 3 dB over the state of the art under in-distribution and out-of-distribution settings, respectively.

Generative AI creates an assessment challenge in quantum software education: a student can provide a homework notebook to ChatGPT and request a completed submission. This study examined whether introductory Qiskit homework could remain autogradable while requiring students to run, review, and discuss results rather than banning AI. Three packages were tested: seeded basis-state circuits with bit flips and customized measurement mappings; Quantum Fourier Transform followed by inverse-transform recovery; and seeded Deutsch-Jozsa with customized oracle masks. The designs used personalization, simulator execution, JSON submissions, hidden references, circuit metrics, reflections, and optional IBM Quantum execution. For each package, one student-visible instance was tested in 50 separate ChatGPT sessions, yielding 150 sessions overall. Every final artifact was executed and passed its grader. Nine sessions were fully archived; none required operator code changes or correction of quantum logic. Under the study's operational definition, each tested instance had zero observed ChatGPT-resiliency. Seeds changed parameters rather than task structure, expected results remained derivable from visible assignment logic, scaffolding exposed key solution steps, and hidden grading verified output consistency without establishing independent authorship or understanding. Because one instance was repeated for each package, the results do not establish solvability for every seed or possible Qiskit assessment. The tested personalized, execution-oriented take-home designs therefore did not prevent successful completion under a minimally engaged-student workflow. Correct artifacts should be complemented by direct assessment through supervised modification, oral defense, prediction, and transfer tasks.

Embedded memories occupy a large portion of modern System-on-Chip (SoC) designs, especially in high-performance applications such as artificial intelligence and edge computing. Memory Built-In Self-Test (MBIST) is commonly used to ensure memory reliability, but it introduces additional area and test time overhead. Accurate early estimation of these overheads is important during design planning, yet conventional methods rely on full Register Transfer Level (RTL) synthesis and test pattern generation, which are slow and resource-intensive. This study proposes a supervised learning framework that predicts MBIST area and test time directly from RTL-level design parameters without synthesis. A dataset of 4,470 samples for area and 624 for test time was generated using Synopsys Design Compiler and MINT, an Intel-enhanced MBIST tool. Input features include memory count, word width, address depth, port configuration, and clock domains. For area prediction, the features are processed through polynomial expansion, log transformation, and scaling, followed by a stacked ensemble model using XGBoost, LightGBM, and a Neural Network with Gradient Boosting as the meta-learner. For test time, XGBoost and LightGBM are combined using Ridge Regression, with hyperparameters tuned through a 100-trial Optuna search. The models achieved 90.68% accuracy for area and 96.80% for test time within a +/-10% margin, improving over baseline methods by 8.53% and 48.80% respectively. The results show that this approach enables faster estimation of MBIST costs and supports more efficient design decisions in memory IP development.

Sequential agent search accumulates changes from its current champion but discards alternative branches; independent proposals preserve breadth but restart from the root. Loreley instead retains complete repository states in a Quality-Diversity (QD) archive and samples them as parents or supplies them as context for later edits. Candidates are Git commits produced in isolated worktrees and judged by a project-supplied evaluator. We compare configured Loreley QD, sequential champion editing, and independent root proposals in a matched Zstandard experiment: seven paired blocks and 48 physical candidate jobs per policy and block (1,008 total), with root-only initialization and each policy's native concurrency. Validation selected a winner at each budget checkpoint; an agent-hidden holdout measured the fixed candidate. At 48 jobs, QD was 0.135% below Sequential Champion (95% BCa interval for the paired effect: -0.556% to +0.161%) and 0.320% above Independent Root (-0.082% to +0.686%). Neither contrast established a QD advantage; Sequential had the highest observed 48-job mean and median. Archive retention and later sampling did occur. Four of seven final QD winners had a non-incumbent state in their primary-parent ancestry under a retrospective one-incumbent rule applied only to the observed QD stream. Including inspiration edges raised the count to six, without showing that supplied context caused an edit. Three earlier capability campaigns produced generation-4, multi-file improvements in two Python libraries and a separate Zstandard revision. Loreley engaged the intended stepping-stone mechanism, but the controlled experiment did not show an endpoint benefit at 48 jobs.

Long-term multi-agent systems continuously accumulate the memories produced by different agents. Existing memory methods typically treat retrieved memories as independent evidence and combine them through voting or weighting. However, this independence assumption often fails in multi-agent settings: memories written by different agents may inherit the same upstream source or shared bias, causing correlated evidence to be repeatedly counted and creating a false majority. We term this failure mode \textit{Memory Correlation Bias}. To address the issue, we propose the \textbf{C}orrelation-\textbf{A}ware \textbf{M}emory \textbf{A}rbitration (CAMA) framework that jointly decouples retrieved memories and recovers missing independent evidence. We model the retrieved memories as query-conditioned evidence groups and combine neural dependency inference with provenance-based symbolic priors to estimate the effective number of independent evidence sources, thereby preventing correlated memories from forming a false majority. Since critical independent evidence may be absent from the initial retrieval set, \textsc{CAMA} further learns a sequential recovery policy that actively retrieves alternative evidence or traces upstream sources before making the final decision, aiming to recover sufficient independent evidence for reliable arbitration while minimizing retrieval cost. Experiments on multiple benchmarks demonstrate the superiority of our method over the state-of-the-art baseline methods, suppressing false majorities induced by correlated memories.

In this paper, a bio-robot is an engineered living or biohybrid system in which living cells perform one or more core functions, such as sensing, information processing, actuation or output. We focus on systems whose cell-based functions are programmed by genetic circuits; physical movement is optional. Designing such a system requires translating application requirements into sensing, logic or memory, output, assembly, host and containment modules, while grounding each choice in traceable parts and evidence. We present micro_biorobot_agent, an offline multi-agent system built on Qwen3.5-27B. The system combines requirement analysis, module-specific retrieval, candidate assembly, conflict checking, local repair, independent review and validation over an integrated library of 23,762 records covering biological parts, measured combinations, literature-supported relationships and actuation evidence. Deterministic output checks align the final report with the retrieved part set and correct false gaps, unsupported part mentions and source-tracking errors. On two author-developed evaluation sets of 50 queries each, the system obtains mean overall scores of 7.35 and 8.04, the highest among the seven evaluated systems; on Scenario Design it exceeds the runner-up by 2.23 points. A 50-query paired ablation shows that the source-tracking check reduces false-gap incidents from 15 to 3, an 80% reduction, and increases source accuracy by 0.75 points. This paper reports the Qwen3.5-based v1 system and evaluates high-level design reports rather than experimentally validated circuits.

Sparse keypoint extraction and matching underpin core tasks in geometric computer vision, including structure-from-motion, visual SLAM, augmented reality, and medical image registration. Learning robust local feature representations, however, typically requires accurate camera poses or depth supervision, which are often unavailable in real-world settings. Reinforcement learning (RL) has recently emerged as a promising alternative, requiring only the information if two images show the same scene or not. However, existing RL formulations such as RIPE rely on coarse binary rewards and carefully constructed negative training pairs, limiting training stability and descriptor discriminability. In this paper, we revisit RL-based keypoint learning and propose a reward that fully exploits the geometric consistency signal, deriving both reward and penalty from a single positive pair without contrasting against negatives. This richer signal provides sufficient supervisory contrast to learn discriminative detectors and descriptors from positive image pairs alone, enabling representation learning under extremely limited supervision. Furthermore, we show that the same RL objective can be extended to the matching stage by adapting LightGlue, raising AUC@5 on MegaDepth1500 from 56.58 to 59.65 and enabling weakly-supervised training of the full sparse matching pipeline from image pairs with partial visual overlap. We validate our approach on established benchmarks, demonstrating competitive results compared to fully-supervised methods. We further show that the method can be even trained on low texture medical video sequences, where camera poses are usually unavailable and standard SfM pipelines often fail. Code and data are available at https://github.com/fraunhoferhhi/RIPEpp .

LLM-based social simulation is a promising complement to traditional methods such as surveys and behavioral experiments. A core question is how to evaluate the fidelity of LLM-simulated human behavior and optimize LLMs toward it. Prevailing practice evaluates by accuracy, checking whether the model selects the single response observed from a human, and trains the LLM to reproduce this hard label. However, human behavior is inherently subjective: the same person in the same situation may reasonably act differently, so an observed response is only one draw from an underlying response distribution, rendering accuracy-based evaluation unreliable and hard-label training misleading. To address these problems, we first introduce the subjectivity coefficient, an entropy-based quantity distinguishing objective tasks such as coding from subjective ones such as social simulation, and use it to systematically analyze how accuracy-based evaluation and hard-label training fail as subjectivity grows. Based on the subjectivity coefficient, we propose Subjectivity-Adaptive soft-Label Training (SALT): it pools observed outputs from semantically nearby inputs into soft distributional labels, with an aggregation radius adapted to the estimated subjectivity of each input; in the near-objective limit the neighborhood shrinks, so SALT naturally falls back to standard single-label training. Moreover, since existing datasets record only single observed responses and cannot support distributional evaluation, we construct SUBJSIM, a benchmark of 19,300 contexts covering 193 annotators and 100 subjective questions. Since real-world data typically provide only a single observation per input, our experiments train models from single observed outputs while evaluating them against the full response distributions, verifying feasibility in realistic settings. Results on SUBJSIM demonstrate the advantages of our method.

We develop a geometric framework for learning deterministic and stochastic forced Hamiltonian systems with neural networks. Motivated by the Lagrange-d'Alembert principle and the theory of variational integrators, we introduce the notion of a Lagrange-d'Alembert map and establish a $C^r$ convergence theorem for first-order one-step methods. Building on these results, we propose Generalized Forced Hamiltonian Neural Networks (GFHNNs), a class of structure-preserving neural networks obtained by concatenating Lagrange-d'Alembert-Euler maps, and prove a universal approximation theorem for this architecture. We further extend the framework to parameter-dependent systems, leading to Parametric Generalized Forced Hamiltonian Neural Networks (PGFHNNs). By interpreting the multiple Stratonovich integrals appearing in the Stratonovich-Taylor expansion as parameters, the same framework can be applied to stochastic forced Hamiltonian systems whenever information about the underlying Wiener process is available. Our numerical experiments demonstrate that the proposed geometric architectures provide significantly improved long-time stability and accuracy compared to non-geometric residual neural networks, while requiring substantially less training data to achieve a comparable level of performance.

Recent studies investigate how to leverage pre-collected datasets to improve the policy performance and sample efficiency of RL. One promising approach to achieve this goal is to employ a two-stage strategy: In the first stage, diverse skills are extracted as a low-level policy from a given dataset, and a high-level policy is trained to solve a specific task in the second stage. Typically, extraction of the low-level policy is performed based on unsupervised learning such as trajectory VAE. However, a limitation of this approach is that the quality of the low-level policy highly depends on the quality of the dataset. To address this issue, we introduce QDOS (Quality-Diversity Offline Skill learning), a unified pipeline for robust offline-to-online learning. Our approach incorporates an Advantage-Weighted Quality-Diversity pretraining objective, which weights the skill extraction and diversity objectives by the estimated advantage of each trajectory segment. This approach allows the model to extract diverse and high-value skills. By providing robust and task-relevant skill representations, QDOS significantly improves the quality of the embedded skill space used by the low-level policy. We further integrate this with a dual dataset reuse strategy, where offline data is used both for skill pretraining and for populating the online replay buffer via pseudo-labeling. Experiments demonstrate that QDOS significantly outperforms strong baselines in structured manipulation tasks and unstructured locomotion tasks, confirming its ability to accelerate exploration and improve final returns in challenging sparse-reward domains.