arXiv:2606.05106v1 Announce Type: new
Abstract: We investigate whether methods of human mathematics pedagogy can guide the training of language models toward arithmetic reasoning. Building on the GASING method -- an Indonesian pedagogy that solves basic arithmetic through a left-to-right procedure aligned with the causal order of token generation -- we operationalize each operation as a computational procedure whose execution trace is serialized into natural-language Chain-of-Thought (CoT) supervision. A small GPT-2 decoder (86M parameters) with a syllabic-agglutinative TOBA tokenizer for Indonesian is trained from scratch on this data using only a next-token prediction objective, without reinforcement learning or reward-based optimization. Monitoring training reveals three distinct learning phases, and mechanistic analyses -- attention-masking interventions on the CoT information graph, residual-stream probing, and logit-lens inspection -- show that the model first internalizes a procedural pathway and subsequently develops an associative, ``mental-arithmetic'' capacity that retrieves intermediate results without explicit step-by-step computation. The trained model reaches over 80% accuracy on held-out problems and attains competitive performance against substantially larger language models, indicating that targeted, pedagogically grounded training can yield strong and economical arithmetic capability at small scale.
Science Journals
arXiv:2606.05029v1 Announce Type: new
Abstract: Controlled experiments are the backbone of machine learning research, but at the scale of modern foundation models, they have become prohibitively expensive. Instead, the community increasingly relies on research strategies that approximate the ideal experiment at a fraction of the cost: proxy experiments and scaling laws, observational studies with publicly available models, and single-run designs that leverage variation within individual training runs. In this work, we argue that there is no free lunch when approximating large-scale experiments on a compute budget. Specifically, savings in compute come at the cost of validity threats -- hidden and sometimes untestable assumptions that, when violated, can invalidate research claims. To help navigate such threats, we propose an evaluation framework that casts foundation model research as a causal inference problem. Within this framework, we evaluate different research strategies through four types of validity adapted from the empirical social sciences -- statistical, internal, external, and construct validity. We find that each strategy comes with a characteristic validity profile: proxy experiments trade external and construct validity for statistical and internal validity; observational studies face confounding and effect heterogeneity; and single-run designs are strained by interference between treated units. This analysis reveals several validity threats that have received insufficient attention in the literature. Overall, our evaluation framework provides researchers with a practical toolkit for scrutinizing validity threats in foundation model research~designs.
arXiv:2606.05027v1 Announce Type: new
Abstract: In this paper, we present a formula for calculating the complex amplitude of the output electric field for a given input wave that impinges on a Dove prism. We use Fourier optics to decompose the input wave into plane waves, then find the output plane waves of the Dove prism as functions of the spatial frequencies of the input components. The total output image is then obtained by integrating all the output plane waves, resulting in a final formula in integral form. Since we conduct a wave-optical analysis for beam propagation and the incidences at Dove prism surfaces, all the physical aspects of electromagnetic waves are involved, including polarization, Fresnel losses, wave interference, phase, and intensity. The formula also explains why a rotated Dove prism rotates its input image twice its rotation angle. In addition, the formula is not limited to paraxial beams, as we find the Dove prism output as a function of the input Fourier components in general, without limiting the input spatial frequencies to small values. However, since in most cases the paraxial approximation is valid and sufficient, a simplified formula is also extracted for paraxial beams. Two ray tracing simulations are conducted to demonstrate the correctness and accuracy of our final simplified formula. All the advantages mentioned make our derivation accurate, complete, comprehensive, and, to the best of our knowledge, the first to wave-optically prove the rotational feature of a Dove prism.
arXiv:2606.05012v1 Announce Type: new
Abstract: The evolution toward 6G networks envisions a heterogeneous Radio Access Network (RAN) comprising diverse access technologies, such as private 5G, public 4G/5G, and Wi-Fi, managed by multiple stakeholders. While considerable research effort has been devoted to O-RAN-based frameworks enabling rApp and xApp implementation and validation, few works provide integrated support for cross-technology RAN orchestration, end-to-end multi-tenancy, and a unified subset of SMO functionalities, including Non-RT RIC components. This paper introduces COSMO, a novel RAN Service Management and Orchestration platform designed to support heterogeneous 3GPP (5G NR, LTE) and non-3GPP (Wi-Fi) access networks. COSMO enables cross-technology multi-tenancy, defined as the capability to allow multiple tenants to dynamically share heterogeneous RAN resources with explicit resource allocation guarantees based on Service Level Agreements (SLAs). This is achieved through management primitives that support flexible and on-demand resource allocation. Additionally, the platform includes a cross-technology Non-Real-Time RAN Intelligent Controller (Non-RT RIC) that enables the development of intelligent rApps for closed-loop control and network orchestration. Beyond its architectural design, COSMO improves resource utilization and operational flexibility through unified orchestration of heterogeneous multi-tenant RAN resources. Through prototyping and benchmarking, we demonstrate the effectiveness of COSMO in resource allocation, SLA enforcement, and scalability. In our prototype, the SLA-based rApp reduces SLA violation from approximately 21% to below 10% under dynamic traffic conditions in a heterogeneous RAN deployment including 5G, 4G, and Wi-Fi access networks. Our results confirm that COSMO offers an efficient solution for managing and orchestrating future multi-tenant cross-technology RAN environments.
arXiv:2606.05004v1 Announce Type: new
Abstract: With the widespread deployment of public large language models (LLMs) such as ChatGPT, protecting user prompt privacy has become an increasingly critical issue. Existing privacy-preserving inference methods sacrifice either utility or efficiency, and often require model-specific modifications that limit their compatibility. In this paper, we propose SharedRequest, a model-agnostic framework for privacy-preserving LLM inference that reformulates privacy protection at the batch level rather than the individual-prompt level. The key idea is to obscure sensitive information by mixing original prompts with noisy variants, while grouping semantically equivalent instructions to amortize the inference cost over a large batch of queries with minimal impact on LLM response quality. This design is independent of the LLM architecture, requiring no access to model parameters or architectural modification. Empirical results demonstrate that SharedRequest achieves over $20\%$ higher utility compared to prior differential privacy baselines, and its shared-prompt mechanism reduces query cost by up to $5\times$ compared to non-batched inference.
arXiv:2606.05003v1 Announce Type: new
Abstract: Large language models (LLMs) increasingly assist in experimental design, yet fluent protocols often remain physically infeasible. We introduce PhysDox, a physical feasibility auditing benchmark for biomedical protocols comprising a 683-sample expert-curated Gold set and a 5,000-sample Silver set across six sensing domains. We formulate the task as a two-stage evaluation: severity detection classifying protocols as valid, minor, or fatal, followed by the constraint-level diagnosis of fatal violations. Evaluating 6 LLMs across 4 inference strategies yields a peak Stage-1 macro-F1 of only 53.0. Moreover, strong oracle diagnosis collapses during end-to-end evaluation due to correlated cascade errors. Error analysis reveals scaffold bias, where models conflate procedural completeness with physical validity. Consequently, implicit constraints exhibit a 2 times higher miss rate than explicit hardware violations, supported by strong statistical correlation at $\rho{=}0.81$ and $p{<}0.01$. Trace analysis of false negatives exposes a 54%--46% split between attention and judgment failures, ultimately demonstrating that protocol auditing demands calibrated feasibility reasoning rather than factual recall or longer rationales.
arXiv:2606.04992v1 Announce Type: new
Abstract: The handling and assembly of instruments during surgery imposes high cognitive demands on scrub nurses, particularly when instruments are unfamiliar. We present a supporting guidance system for surgical instrumentation that combines multi-camera 6D pose estimation with augmented reality in-situ visualization on a head-mounted display without the requirement for additional markers. Pose estimation and consecutive camera calibration are achieved through known objects. The 6D pose estimation network is trained purely on synthetic data, aiming for better generalizability and real-world applicability. The AR guidance displays tooltip localization cues and step-wise assembly animations. Via gaze-based selection and a foot pedal, users can switch between assembly steps in intraoperative use.
In a technical evaluation, our approach outperforms state-of-art 6D pose estimation. A user study with 29 scrub nurses was conducted in a surgical simulation of knee arthroplasty, comparing the system against a paper manual. AR guidance significantly reduced the perceived workload compared. Objectively, AR guidance reduced task completion time by 21.3\% (4.76 minutes). Specifically, scrub nurses less experienced with the instrument set benefited when using the system. Error frequencies were comparable between conditions. Qualitative feedback highlighted improved process clarity, reduced information overload, and perceived independence. To summarize, our marker-free multi-camera AR guidance approach for surgical instruments can, subjectively and objectively, improve intraoperative instrumentation performance, particularly for untrained scrub nurses.
arXiv:2606.04989v1 Announce Type: new
Abstract: Although much is known about the physical danger of cycling situations, less is understood about the perceived danger of cycling. Furthermore, perception of danger may be filtered at a subconscious level and therefore difficult for one to self-report. To this end, these subconscious perceptions can be revealed through physiological metrics such as eye gaze. This paper explores the perceived safety of cycling in Oxford, United Kingdom and explores the ability of wearable eye tracking glasses to produce insights about the differences in perception under different environments and events. This paper finds that eye gaze patterns change between using bike lanes, car lanes and shared bus lanes, representing different cognitive challenges of each lane type. This paper presents that different intersections have significantly different eye gaze patterns which may have implications for cyclist stress. Finally, eye gaze patterns differ in the presence of events such as passes and pedestrians in the road compared to when cycling with no events. This paper draws conclusions on the benefits and limitations of using wearable eye trackers to estimate stress and cyclist workload.
arXiv:2606.04982v1 Announce Type: new
Abstract: This paper analyzes the convergence of parallel overlapping domain-decomposition methods with impedance boundary conditions for the time-harmonic Maxwell equations in heterogeneous media. We prove that the parallel iterative method is well-posed in an appropriate function space, and characterize the error propagation operator through impedance-to-impedance maps that describe interactions between neighboring subdomains. For strip domain decompositions, we derive explicit convergence estimates in terms of the norms of the impedance-to-impedance maps. At the discrete level, we develop the finite-element counterpart of these results based on N\'{e}d\'{e}lec-element discretisations. Under the assumption that the discrete impedance-to-impedance maps approximate their continuous counterparts as the mesh is refined, we show that the discrete method inherits the convergence behavior of the continuous method. We illustrate this theory with numerical experiments for strip domain decompositions, and also present numerical experiments for checkerboard domain decompositions that go beyond our theory.
arXiv:2606.04981v1 Announce Type: new
Abstract: Energy-intensive data centers (DCs) have emerged as substantial and flexible loads in modern power systems, underscoring the critical need for computation-electricity coordination. Harnessing the spatio-temporal flexibility of DC workloads is a promising approach to facilitate this coordination. However, existing studies overlook the collaborative potential of computational resource sharing among geo-distributed DCs, thereby failing to fully unlock this flexibility. In this paper, a bi-level computation-electricity coordination framework is proposed to explicitly capture the bidirectional interactions between DCs and power grid. Firstly, a peer-to-peer cloud service market (P2P-CSM) for geo-distributed DCs is proposed, which enables bilateral cloud service transactions to leverage regional heterogeneities (e.g., electricity prices, cooling efficiency). Secondly, locational marginal prices are embedded into the framework to reflect network congestion and nodal price disparities. Thirdly, a dual consensus alternating direction method of multipliers (ADMM)-based decentralized algorithm is developed as the P2P market clearing algorithm, and a bisection-assisted iterative algorithm is proposed to ensure rigorous convergence of the framework. Case studies conducted on modified IEEE 30-bus system validate that the P2P-CSM achieves a win-win computation-electricity coordination: it not only increases total DC operational profit by 22.8\%, but also effectively alleviates grid congestion and yields a 3.2\% reduction in total energy consumption.
arXiv:2606.04980v1 Announce Type: new
Abstract: Mixture-of-Experts (MoE) architectures scale model capacity through sparse expert activation, but their deployment remains memory-bound because all expert weights must reside in memory. Mixed-precision quantization can substantially reduce this footprint by assigning different bit-widths to different experts. Existing approaches, however, typically rely on calibration data to estimate expert importance and determine bit allocation. For frontier MoE LLMs, the original training data, and hence the true training distribution, is proprietary and inaccessible. As a result, calibration sets are inevitably imperfect surrogates, and this can misestimate expert utilization and lead to suboptimal bit allocation. Motivated by the substantial cross-expert quality variability observed in modern MoE models, and by the success of Heavy-Tailed Self-Regularization (HT-SR) theory at predicting neural network model quality without access to training or testing data, we propose AlphaQ, a calibration-free bit-allocation method for MoE quantization. AlphaQ draws on HT-SR theory and follows a simple principle: experts with more heavy-tailed weight spectra are typically better trained and hence should receive higher bit-widths, while experts with weaker heavy-tailed structure can be quantized more aggressively. AlphaQ operationalizes this principle by measuring expert-wise spectral heavy-tailedness and solving a budget-constrained optimization problem that minimizes total quantization error under a global bit-budget constraint. Across several MoE models, AlphaQ consistently outperforms calibration-based baselines under matched bit budgets. Notably, on Qwen1.5-MoE, AlphaQ achieves near full-precision accuracy with an average expert precision of only 3.5 bits, while delivering more than 4$\times$ memory compression. Our code is available at https://github.com/Superone77/AlphaQ.
arXiv:2606.04978v1 Announce Type: new
Abstract: LLMs can appear cautious in risk decision-making tasks, yet cautious-looking outputs do not necessarily indicate alignment with human decision-making mechanisms. We investigate this distinction using the St. Petersburg game as a controlled testbed, a classical paradox in which the expected payoff is infinite, yet humans typically report low, finite willingness to pay. We evaluate 28 LLMs with a structured prompt suite that includes the original game; controlled decision variants that perturb truncation, repeated play, numeric endowment, and occupational identity; a human-perspective prompt that asks models to reason as human decision makers; and paired comparisons between base models and their instruction-tuned counterparts. In the original game, most models generate finite bids, creating the appearance of human-like risk behavior. However, this outcome-level resemblance masks substantial mechanism-level differences. The controlled variants reveal that rather than maintaining human-like behavior seen in the original game, models often shift to conditionally and computationally rational behavior. Human-cue prompting and instruction tuning often lower bids and reduce some visible pathologies, but most mechanism-level response patterns remain largely unchanged. These findings show that behavioral alignment in risk decision-making can be surface-level: LLMs may produce human-like risk decisions without exhibiting human-consistent mechanisms. High-stakes evaluations of LLM decision-making should therefore move beyond outcome similarity and examine whether the alignment is supported by mechanism-level consistency.
arXiv:2606.04928v1 Announce Type: new
Abstract: Large Language Models (LLMs) are increasingly deployed across diverse applications, raising critical questions for governance, accountability, and data provenance. Understanding which training data most influenced a model's output remains a fundamental open problem. We address this challenge through training data attribution (TDA) for auto-regressive LLMs by expanding upon the inverse formulation: How would training data be affected if the model had seen the generated output during training? Our method perturbs the base model using bidirectional gradient optimization (gradient ascent and descent) on a generated text sample and measures the resulting change in loss across training samples. Our framework supports attribution at arbitrary data granularity, enabling both factual and stylistic attribution. We evaluate our method against baselines on pretrained models with known datasets, and show that it outperforms previous work on influence metrics, thereby enhancing model interpretability, an essential requirement for accountable AI systems.
arXiv:2606.04925v1 Announce Type: new
Abstract: Video panoptic segmentation (VPS) aims to jointly detect, segment, and track all objects while partitioning the video into semantically consistent regions. We introduce the task setting of unsupervised VPS, omitting any human supervision. Existing unsupervised scene understanding works mainly focused on image segmentation tasks; the video domain remains underexplored. We propose VideoCUPS, the first unsupervised VPS approach. VideoCUPS generates temporally consistent panoptic video pseudo-labels from scene-centric videos by exploiting unsupervised depth, motion, and visual cues. Training on these pseudo-labels using a novel Video DropLoss yields an accurate, unsupervised VPS model. To benchmark progress, we introduce a comprehensive evaluation protocol and four competitive baselines, extending state-of-the-art unsupervised panoptic image and instance video segmentation models to VPS. VideoCUPS outperforms all baselines and demonstrates strong label-efficient learning. With VideoCUPS, our evaluation protocol, and baselines, we provide a strong foundation for future research on unsupervised VPS.
arXiv:2606.04924v1 Announce Type: new
Abstract: The widespread use of Large Language Models (LLMs) as writing tools challenges the validity of crowdsourced data, as crowdworkers may outsource tasks to models. To better understand how this is addressed, we surveyed 155 researchers in NLP and related disciplines about their experiences and opinions on collecting free-text responses via crowdsourcing. This paper provides an overview of practitioners' challenges, mitigation strategies, and the foreseen implications on data quality. 44% of respondents reported observing LLM usage in their crowdsourced data. While 93% of them had anticipated this, half were unsure what precautions to take. The most prevalent detection strategies are distinctive textual style patterns and unusually fast completion times. Overall, survey responses show that the research community is aware of the problem and taking measures, but existing efforts remain insufficient to fully address it. Finally, we derive a set of considerations to guide future crowdsourced free-text data collection in the era of LLMs.
arXiv:2606.04923v1 Announce Type: new
Abstract: Rubric-based reinforcement learning (RL) uses an LLM-as-a-Judge (LaaJ) to score model outputs according to rubrics as rewards. However, policy models may exploit latent biases in the judge, leading to reward hacking and ineffective or unsafe training outcomes. In real-world rubric-based RL, such hacking behaviors are often subtle and entangled with multiple judge biases, making them difficult to analyze, detect, and mitigate. In this paper, we introduce CHERRL, a controllable hacking environment for rubric-based RL. By injecting known biases into LaaJ, CHERRL enables stable reproduction of reward hacking, explicit observation of reward divergence, and precise identification of hacking onset. This provides a clean experimental testbed for studying the mechanisms and mitigations of reward hacking in rubric-based RL. To demonstrate its utility, we analyze different judge biases from the perspectives of discoverability and exploitability, and explore an agent-based system for automatically detecting reward hacking onset from training logs. The code and environment are publicly available at https://github.com/THUAIS-Lab/CHERRL.
arXiv:2606.04922v1 Announce Type: new
Abstract: Current prompt-based and adapter-based tuning of vision-language models (VLMs) is attractive for medical imaging, where clinical data sensitivity favors frozen backbones and annotations are limited. However, these methods typically optimize only the ground-truth class, treating all other classes as equally incorrect, ignoring clinically meaningful class relations and yielding unstable decision boundaries in limited-supervision settings. We propose Omni-Geometry Knowledge Distillation (OGKD), a new framework that injects class-relation structure into the teacher to produce directional targets that preserve the ground truth while respecting inter-class geometry. Using these targets, we develop two distillation losses: Global Geometry-Aware Distillation (GAD) operates on the global image token, and Label-Guided Geometry Distillation (LGD) applies the same geometry to attentive patch tokens to improve fine-grained alignment. Across comprehensive experiments and analyses on 11 widely-used medical datasets for base-to-novel and few-shot evaluations, our OGKD achieves substantially better performance, consistently improving accuracy by an average absolute gain of 1.7%-2.8% over all prior state-of-the-art VLM adaptation counterparts. It also robustly generalizes to unseen classes and yields more reliable predictions than other approaches. Our code is available at https://github.com/tientrandinh/OGKD.
arXiv:2606.04905v1 Announce Type: new
Abstract: Event Calculus (EC) implemented in answer set programming (ASP) has proven suitable for specifying requirements on safety-critical systems thanks to its elegant representation of both discrete and continuous changes and its semantic closeness to semi-formal natural language. However, continuous changes and the size of value domains of time and system properties (fluents) pose significant challenges. Grounding-based ASP solvers, e.g., clingo, which implement Discrete EC (DEC), lead to combinatorial explosion in program size and inaccurate representation. The grounding-free s(CASP) does not discretize but struggles with non-termination due to its top-down execution. This paper introduces Hybrid EC, an extended axiomatization of DEC, that tackles the challenges via functional fluents and a mapping of time to abstract steps. We implement it using clingcon and clingo-lpx (Hybrid ASP systems over integers and rationals, respectively) where the value (dense) domains of fluents and time are represented as linear constraints and evaluated by external solvers, while ensuring termination whenever solutions exist. We validate both implementations on a number of examples and observe that they are unaffected by the size of the domains and that handling rationals does not impact scalability. Most importantly, the ability of clingo-lpx to handle dense domains enables accurate modeling of continuous change.
arXiv:2606.04903v1 Announce Type: new
Abstract: We introduce the LLM agent architecture Agentic Redux, intended for use with nontrivial problem domains that require linear auditability. Using the typed lambda calculus, we prove that, run on appropriate domains, Agentic Redux executions are semantically guaranteed to be correct, with all decisions recorded in an append-only ledger. We present two production-grade appropriate domains, in healthcare billing compliance, and security vulnerability disclosure. Working code for Agentic Redux run on both domains is available in a supporting code repository. We also introduce Ontology-First Agent Design, a methodology for creation of agent frameworks on a problem domain, in which a human expert ontologizes the problem domain with Basic Formal Ontology, and then assigns an LLM to derive roles that agents and humans-in-the-loop can fill, in order to work the problems in the domain.
arXiv:2606.04898v1 Announce Type: new
Abstract: Anatomical landmark detection is a fundamental task in medical image analysis supporting a wide range of diagnostic and interventional workflows. Although recent methods have achieved sub-millimetric localisation, accuracy alone is not sufficient for clinical deployment, requiring reliability and robustness in prediction. Despite its clinical relevance, the impact of representation learning in this context is still underexplored. In this work, we introduce CDPM-align, a multi-scale guidance-aligned conditional diffusion pre-training for anatomical landmark detection. Our experimental setup focuses on a few images and a few annotation regimes. Specifically, we employ three popular heterogeneous small-scale benchmark datasets for representation learning via conditional generative pre-training. Furthermore, we consider low-annotation scenarios for the downstream task of landmark detection, with 10 and 25 annotated images, reflecting realistic trade-offs between clinical effort and resource constraints for annotations. Our results confirm that generative pre-training enables the model to learn a robust representation. This improves both accuracy and uncertainty on the downstream tasks, advancing towards safe and efficient clinical deployment.
arXiv:2606.04866v1 Announce Type: new
Abstract: Large-scale hyperparameter optimization (HPO) in automated machine learning (AutoML) consumes substantial computational resources, raising growing concerns about scalability and energy efficiency. Existing methods use prior information heuristically to accelerate both black-box and multi-fidelity settings, but they lack a characterization of how prior informativeness quantitatively reduces sample complexity. In this work, we provide the first distribution-dependent sample complexity bounds for multi-fidelity HPO with priors through the formal lens of fixed-budget best-arm identification. By modeling priors directly over arm means as configuration performance, we derive explicit, distribution-dependent error bounds that quantify the relationship between priors and evaluation budget. Our analysis shows that informative priors, which concentrate probability mass on near-optimal arms, yield reductions in the number of required evaluations, whereas baseline performance is recovered with uninformative or misleading priors. We conduct proof-of-concept experiments on a synthetic benchmark and on LCBench, a common multi-fidelity HPO benchmark for deep learning, to confirm our theoretical results, achieving up to 90% budget reduction while retaining solution quality. Together, our results provide a principled foundation for prior-guided and compute-efficient green AutoML.
arXiv:2606.03810v2 Announce Type: replace
Abstract: Consistency training encourages a model to produce similar outputs across related inputs or sampling procedures. Such methods are simple, scalable, and largely label-free, but their effects on model alignment remain poorly understood. Could the self-bootstrapping nature of these methods amplify undesired behavior in models? We test seven consistency training methods on 108 model organisms: open-source models (7B--70B) fine-tuned to exhibit various forms of controlled misaligned behavior. We find that outcomes vary significantly: consistency training generally suppresses reward hacking and emergent misalignment but amplifies sycophancy. We present evidence that distribution shifts induced by the consistency labeling process, rather than variation in the selection operators, may be the primary driver of systematic alignment effects. Finally, we present a unifying theoretical framework to derive conditions under which consistency training will amplify or suppress misalignment. In total, our study establishes that consistency training is not alignment-neutral, and that its use in critical systems should be carefully audited.
arXiv:2606.03441v2 Announce Type: replace
Abstract: Autonomous vision-based perching of quadrotors on moving inclined platforms is critical for air-ground collaboration but remains challenging due to the limited field of view (FOV). In this paper, we propose PerchRL, a reinforcement learning (RL) framework for vision-based agile perching on inclined platforms under rapid and irregular motion. Specifically, we employ a two-stage learning strategy consisting of state-based pre-training followed by vision-based fine-tuning. To improve generalization across diverse platform motions, we employ randomized platform trajectories to prevent overfitting and temporal augmentation methods to capture latent motion patterns from historical observations. During vision-based fine-tuning, a hybrid learning framework consisting of visibility-aware state augmentation and active perception rewards is presented to improve robustness under intermittent visual loss. Extensive simulation and real-world experiments demonstrate the feasibility, stability, and real-time performance of PerchRL, while successful deployment across distinct quadrotor platforms further validates its adaptability. The source code will be released to benefit the community.
arXiv:2606.04231v1 Announce Type: new
Abstract: Recent advances in multimodal retrieval-augmented generation (MM-RAG) have shifted toward minimal parsing, relying on page-level images for producing retriever embeddings and for answer generation. While efficient, this trend often neglects explicit handling of the rich, structured information in complex enterprise documents, instead depending on pre-trained embeddings or vision-language models to implicitly capture such structure. In this work, we take a more direct approach: MM-BizRAG proactively extracts and represents document structure via a document structure-aware split that dynamically routes documents through orientation-specific ingestion pipelines, applying explicit layout-aware parsing for vertically structured documents (e.g., reports) and holistic page-level representations for horizontally structured documents (e.g., slide decks). A unified LLM-driven artifact transformation pipeline with placeholder-based positional alignment preserves natural reading order, while inference-time multimodal assembly decouples retrieval representations from generation context, enabling richer, more grounded answers without any finetuning requirement. Through experiments on a large, heterogeneous enterprise dataset and two public benchmarks (SlideVQA and FinRAGBench-V), MM-BizRAG consistently outperforms state-of-the-art vision-centric baselines by up to 32% points, with especially strong gains on report-style layouts. Furthermore, we introduce FastRAGEval, a single-call LLM Judge metric for fine-grained generative recall that halves RAGChecker's cost while achieving stronger human alignment.
arXiv:2606.04806v1 Announce Type: new
Abstract: LLMs and agentic systems are increasingly deployed in social environments, making normative competence critical for safe and appropriate behavior. However, existing approaches either assess normative judgment in text alone or reduce it to choosing among a fixed set of candidate actions. We argue both are insufficient. In practice, agents are never handed a menu of options; they must identify a reasonable action from scratch, grounded in visible facts and supported by inspectable reasons. We introduce NoRA, a visual first-person video benchmark that requires models to generate candidate next actions and justify each through an explicit fact-reason-action support graph. The benchmark comprises 1,420 annotated video clips, including HumanGold-190 and LLMSilver-1230 splits. Each instance is evaluated through action alignment, factual grounding, and support binding, aggregated into a single grounded reasonableness score. We benchmark 12 multimodal systems under direct, deliberate, and structured prompting regimes, finding that current VLMs frequently recover plausible actions and relevant scene facts, but consistently struggle to construct the full reasonable action space and bind selected actions to the correct local support. NoRA makes this gap measurable, shifting the evaluation question from whether a model can pick an action to whether it can justify an appropriate action for the right visible reasons.