arXiv:2606.12714v1 Announce Type: new Abstract: We present an asymptotic analysis and high-order boundary integral method for the three-dimensional Neumann Green's function in general geometries. The Neumann Green's function is a fundamental quantity which arises in numerous fields of science and engineering. In the application of singular perturbation methods to strongly localized reactions and diffusive transport, the Green's function plays the key role in mediating global dynamics. However, this essential quantity can only be determined in closed form for a limited set of geometries. The Green's function for the Laplacian is an elliptic problem with a Dirac forcing term. Accurate resolution of the solution requires a careful decomposition into a singular and a regular part. The bulk scenario is where the source is placed off surface and the singularity is given by the free-space function. In the surface case, where the source is placed at a curved point on the boundary, we use asymptotic analysis to determine a three-term singularity structure. With explicit knowledge of these singularities, we develop a high-order boundary integral method for the determination of the remaining regular part. To resolve the singular boundary data, our integral method uses a custom discretization with Duffy patches near the source. We validate our method using several test cases in which closed form solutions can be developed, including spheres, prolate spheroids and constructed domains. We demonstrate the applicability of our method to address some open problems in narrow capture theory.
Science Journals
arXiv:2606.12731v1 Announce Type: new Abstract: As LLMs increasingly serve in advisory and deliberative roles, users rely on them for non-verifiable reasoning in domains lacking objective ground truths. However, traditional evaluations of LLM reasoning focus almost exclusively on fact-based domains, such as mathematics and science, leaving uncertainty over whether and to what degree models can handle ambiguous, subjective, or value-laden problems over time. To address this concern, we propose moral reasoning as a paradigmatic subdomain of non-verifiable reasoning. We define moral robustness as a model's capacity to exhibit sound moral reasoning across time and contexts, and we introduce a scalable, adversarial, multi-turn evaluation framework to empirically measure this capability. We simulate 48,000 user-agent moral deliberations across four frontier LLMs, varying premise relevance, premise order, conversation duration, and the user's stated moral view. We find that models successfully ignore morally-irrelevant distractors, but shift their reasoning by up to 6.5%, on average, towards the user's stated preferred moral view, and varying their reasoning depending on factors such as order (altering moral judgments by order in 13-22% of the cases) and duration (altering moral judgments between single-turn and multi-turn in 10-24% of the cases). Our analysis indicates that models tailor not just their final verdicts but their underlying justifications to align with a user's moral viewpoint - a failure mode we characterize as moral deliberative sycophancy.
arXiv:2606.12744v1 Announce Type: new Abstract: In-Context Learning (ICL) has become a powerful mechanism for adapting Large Language Models (LLMs) to new tasks without fine-tuning. Extending this concept to Large Multimodal Models (LMMs), Multimodal In-Context Learning (M-ICL) relies on retrieving relevant examples, such as images, captions, or question-answer pairs, to guide predictions across tasks like classification, captioning, and visual question answering (VQA). Most existing approaches select in-context examples based on feature-space similarity, assuming that semantically similar samples provide the most useful context. However, our systematic analysis reveals that this assumption does not always hold: visually similar examples are not necessarily those that most effectively enhance in-context learning performance. To address this, we propose the Guided Retrieval of In-context Prompts (GRIP), a learnable vision-only retrieval framework that leverages feedback from LMMs to identify examples that truly improve model predictions. GRIP learns to distinguish beneficial from detrimental in-context examples through contrastive training, refining retrieval beyond pure similarity. Across three multimodal tasks, namely classification, captioning, and VQA, GRIP improves consistently over similarity-based retrieval on Qwen2.5-VL-7B, with its strongest gains in classification on Idefics2-8B. Moreover, we demonstrate that retrievers trained with feedback from one open LMM can be transferred to other models without retraining, including closed-source GPT-4o and Gemini, enabling scalable and cost-efficient deployment of M-ICL. Code will be published upon acceptance.
arXiv:2606.12747v1 Announce Type: new Abstract: Safety-relevant studies of language models, including alignment and jailbreaking evaluations and AI control protocols, often rely on prefilling model outputs. If AI models can recognize and act on the fact their prior assistant messages have been inserted or edited, the effectiveness and validity of these methods could be compromised. We investigate whether frontier language models can distinguish between tampered and untampered assistant-side context, a capability we call prefill awareness. To do so, we construct a binary preference benchmark across three prefill mechanisms, filtering for cases where models show consistent stances. We find that frontier models show substantial prefill awareness: Claude Opus 4.5 detects prefills opposing its preferences in 9-35% of cases with a 0% false positive rate when prompted; additionally, models often revert towards baseline behavior without explicitly reporting that the prefill was foreign. Controlled ablations later also show that detection and resistance rely on different cues, where stylistic mismatch mainly affects whether models flag a prefill as foreign, while preference mismatch mainly affects whether they revert toward their baseline answer. We also examine more realistic agentic settings such as misalignment-continuation evaluations and SWE-bench trajectories, where frontier models sometimes disavow prefilled assistant turns in ways that depend strongly on dataset, task success, and hidden formatting artifacts. Our results indicate that prefill awareness is already a substantial confound for some prefill-based methods. We recommend that model developers track this capability in frontier systems.
arXiv:2606.12748v1 Announce Type: new Abstract: Why is the past of English "go" the apparently unrelated "went"? Such alternations are frequent in languages. They neither aid communication nor learnability, yet they can be persistent, surviving over centuries or millennia. We present a multi-agent simulation of the emergence of morphological stem and inflection alternations. Alternate forms arise by phonological changes or, as with "go/went", from lexical alternatives associated with a subset of the population. When an agent 'hears' another agent use a novel form for a slot in the paradigm of a word (say, the past tense of go), they will with some probability adopt that form, possibly spreading its use to other slots in the paradigm that shared the same original form. Thus alternative forms can spread through the population and become entrenched as stem or inflectional marker alternants. Unlike many previous computational studies, our system allows for naturalistic lexical forms, realistic phonological rules, lexicons with hundreds or thousands of entries, and agent populations in the tens or hundreds. It supports several network topologies, diffusion patterns and agent adoption policies. One issue with such simulations is evaluation: how realistic is the resulting morphology compared to those of real languages? We introduce the AI Historical Linguist, a novel Large Language Model-driven system that models a debate between two historical linguists. We use this to compare a set of real language morphologies, disguised morphologies, and experimentally evolved morphologies. The results suggest that among the factors that favor more plausible morphologies are scale-free social networks and random Bernoulli adoption of forms. We also present three case studies modeling attested historical changes, allowing us to test what might have happened if history had been different. All code and data are released.
arXiv:2606.12749v1 Announce Type: new Abstract: Bichromatically pumped Kerr microresonators exhibit phase bistability and multistability arising from four-wave mixing between pump and generated modes. Here we analyze the symmetry structure of stationary solutions in the coupled-mode description of such systems. We show that the two-pump coupled-mode equations are a particular case of a more general model with pumps placed at equally spaced modes. For this model, any stationary solution generates a finite family of stationary solutions through a discrete phase transformation. This transformation leaves the equations invariant and preserves the stability type of the corresponding stationary states. As a consequence, the possible stationary values of each individual mode form a regular polygon in the complex plane. The number of vertices is determined by the order of the mode index $\mu$ in the group $\mathbb{Z}_n$, where $n$ is the separation between the pumped modes. Thus, the phase multistability structure depends not only on nonlinear dynamics but also on the arithmetic relation between the mode index and the pump separation. These results provide a simple symmetry-based explanation for the discrete phase structure observed in multimode Kerr resonators under bichromatic pumping.
arXiv:2606.13152v1 Announce Type: new Abstract: We study Fibonacci (staircase) polyominoes, a class of column-convex polyominoes whose lower boundary is a staircase with unit vertical steps. We derive multivariate generating functions that refine Turban's Fibonacci-number enumeration by tracking additional perimeter and area parameters. The proofs use a catalytic functional equation and, in a perimeter specialization, the kernel method, leading to explicit closed forms and Catalan-number coefficient formulas.
arXiv:2606.12752v1 Announce Type: new Abstract: The resilience literature measures urban performance as recovery: the degree to which a city returns to its pre-shock baseline. This paper develops a stronger concept -- civic ascent -- as part of a broader research program on the ethology of coupled agent-environment systems, of which the city is the deepest available empirical instance. Civic ascent is defined as the condition in which a city emerges from shock with higher functional capacity than before. We develop a conceptual framework in the ethological tradition, treating the city as a coupled system of three slow state variables -- topos (physical structure), nomos (institutional structure), and hexis (civic judgment) -- together with a fast affective channel (delta) through which shocks to topos and nomos reach hexis. The framework distinguishes three structurally distinct pressures on civic systems: shocks (discontinuities in T or M), decay (continuous entropy), and leakage (active extraction of civic surplus into non-civic pools). The ascent condition is that reinforcement from cross-coupling of T, M, and H exceeds the combined loss from decay and leakage. Post-shock ascent is measured by a normalised improvement index A(T) applied to a composite civic performance signal P(t) constructed from scale-adjusted key performance indicators, distinguishing intrinsic civic ascent from demographically driven growth. New York City after September 11, 2001, is proposed as the primary empirical case; the operational measurement program is specified in the companion NYC Civic Data Map (Washburn 2026c, 133 KPIs) and executed in Paper 2. The reader for whom only the urban contribution is of interest will find it complete in itself; the reader interested in the larger program will find this paper its formal core.
arXiv:2606.12789v1 Announce Type: new Abstract: Evaluating retrieval-augmented generation (RAG) systems requires benchmarks that capture diverse question characteristics, yet practitioners lack empirical guidance on which dimensions to vary and at what granularity. We present HieraRAG, a hierarchical framework for studying granularity in RAG benchmark construction, defining optimal granularity as the level that maximizes discriminative power (the standard deviation of generation quality across categories) within a given RAG configuration. As a case study, we generate 5,872 synthetic question-answer (QA) pairs from FineWeb-10BT across 3 dimensions (Question Complexity, Answer Type, Linguistic Variation) at 3 granularity levels (2, 4, and 8 categories). With a BM25+Falcon-3-10B pipeline, optimal granularity varies by dimension: complexity benefits from fine-grained distinctions (discriminative power: 0.053) while answer type and linguistic variation peak at medium granularity. We introduce a Coherence Ratio metric to quantify whether fine-grained splits cleanly subdivide parent categories, revealing structural differences across dimensions (Question Complexity: 0.40 vs. Answer Type: 1.44). Human evaluation of 110 stratified QA pairs confirms synthetic quality. While these specific findings reflect a single configuration, HieraRAG provides a portable procedure and validation metric for practitioners to determine evaluation granularity within their own RAG settings.
arXiv:2606.12790v1 Announce Type: new Abstract: Large Language Models have consistently demonstrated a lack of creativity and diversity across tasks. Prior work has focused on addressing whether models are capable of generating creative outputs. Here, we aim to consider novelty and investigate what makes model-generated content novel or not novel in a task-specific manner. We propose a fine-grained evaluation metric GENIE to measure the novelty of responses along task-specific features with respect to a population of responses. We show that unlike GENIE, holistic metrics struggle to capture the high-dimensionality of novelty and do not provide insight on which properties they target. Finally, we use GENIE to measure the effectiveness of mitigation methods that address creativity to better understand where these methods can improve novelty.
arXiv:2606.12797v1 Announce Type: new Abstract: Agentic large language model systems that autonomously invoke tools, maintain persistent memory, and execute multi-step plans are increasingly deployed in public-facing domains, including government services, healthcare triage, and financial advising. We ask whether the frameworks used to build these systems provide architectural-level structural safety guarantees. Applying six containment principles derived from a compositional model of agentic architectures, we audit three dominant frameworks (LangChain, AutoGPT, and OpenAI Agents SDK) and find no native compliance in any of them. Memory integrity, a defense against one of the most prevalent vulnerability classes, is not observed in any of the three evaluated frameworks. We validate these findings empirically: in a simulated government benefits agent built on LangChain, a single memory-poisoning write induces persistent targeted corruption across all tested seeds and backends, increasing the wrongful denial rate for targeted applicants to 88.9%. Under a complex five-factor policy, the same attack preserves aggregate accuracy while increasing targeted wrongful denials by 3.5x, rendering the corruption difficult to detect through standard monitoring. We then introduce two lightweight containment mechanisms: a memory integrity validator and a policy gate, which eliminate both attack vectors with sub-millisecond overhead (<0.2ms per call). We conclude that the current agentic framework ecosystem may not yet meet secure-by-default expectations for public-facing deployments and outline priority architectural interventions to enable trustworthy deployment in high-stakes, socially impactful applications.
arXiv:2606.12800v1 Announce Type: new Abstract: Digital elevation models (DEMs) have reached resolutions and sizes that only parallel computaters can efficiently process. One important application of DEMs is predicting how much water flows where, the so-called ``flow routing problem'' (a variation of which is the problem of determining the drainage area upstream of a point in a DEM). The traditional algorithm for flow routing is sequential, and attempts to parallelize this method have so far only been moderately successful. Herein, we build on earlier work in Richardson et al. (2014) and propose an algorithm and several variations that can efficiently solve the flow routing problem on very large models with very large numbers of parallel processes. For the largest model we use, with 1.88 billion points, the best algorithm herein can route water in 4.0 seconds on 12,288 processes of a computer cluster.
arXiv:2606.12801v1 Announce Type: new Abstract: This study presents AiAWE, an open-source automated writing evaluation system that scores argumentative essays using a LoRA-adapted instruction-tuned large language model (Gemma-3-27B-it). Using a proprietary Educational Testing Service (ETS) dataset of 480 TOEFL Independent Writing essays, we fine-tune Gemma-3-27B and LLaMA-3.3-70B under identical LoRA configurations on a 120-essay training subset and evaluate on the remaining 360 essays under identical inference quantization. The fine-tuned Gemma model achieves a root mean square error of 0.474, a quadratic weighted kappa of 0.828, and an agreement rate of 90.56% within +/- 0.5 of the human score, outperforming both the larger LLaMA-3.3-70B model and the fine-tuned GPT-3.5 baseline reported in prior work on the same dataset. Three findings are of broader interest: open-weight LLMs can match or exceed proprietary fine-tuning for rubric-aligned scoring; model scale is not a reliable predictor of downstream performance under LoRA adaptation; and identical LoRA hyperparameters produce qualitatively different adaptation behaviors across architectures. The production system runs on a consumer-grade server and is publicly accessible at https://app.awade.gec.waseda.ac.jp. LoRA adapters, application code, and fine-tuning YAMLs are publicly available through their respective repositories.
arXiv:2606.12809v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) are trained on massive multimodal data, making data unlearning increasingly important as data owners may request the removal of specific content. In practice, these requests often arrive sequentially over time, giving rise to the challenging problem of MLLM Lifelong Unlearning. However, most existing benchmarks are limited in scale and scope, failing to capture the complexities of MLLM lifelong unlearning. To fill this gap, we introduce the MLUBench, a large-scale and comprehensive benchmark featuring 127 entities across 9 classes under lifelong unlearning requests. We perform extensive experiments using MLUBench and reveal that existing unlearning methods suffer from severe, cumulative degradation. More critically, we further identify the unique challenge of this problem: unlike in unimodal models, MLLM lifelong unlearning is constrained by the need to preserve multimodal alignment. Continually unlearning from one modality could degrade the entire model. To alleviate this challenge, we propose LUMoE, an effective method. Experiments demonstrate that LUMoE significantly mitigates the degradation problem faced by baselines. The source code and the MLUBench dataset are open-sourced in https://github.com/lihe-maxsize/Lifelong_Unlearning_main.
arXiv:2606.12814v1 Announce Type: new Abstract: Recent reinforcement learning approaches have shown great promise in improving humanoid motion tracking performance and achieving fall recovery under disturbances. However, most existing works treat motion tracking and fall recovery as different tasks and require multi-stage training with specialized recovery rewards and/or separate recovery policies. Moreover, existing reinforcement learning-based methods often terminate training episodes immediately after severe tracking failures, limiting recovery-oriented exploration in unstable or fallen states. To address the above issues, we propose Stubborn, a streamlined and unified reinforcement learning framework to achieve robust humanoid motion tracking and fall recovery. Specifically, Stubborn uses an asymmetric Actor-Critic architecture and consists of three major components. First, a yaw-aligned tracking representation is adopted to reduce sensitivity to global drift and heading disturbances while preserving gravity-related balance information. Second, we introduce a Bernoulli-based probabilistic termination mechanism that enables the policy to encourage exploration of fall-recovery behaviors under varying failure modes. Third, we propose a probabilistic termination and tracking-error-driven strategy that dynamically reshapes the sampling distribution based on tracking performance, increasing the training efficiency for difficult motion segments and unstable states. Extensive comparisons with SOTA methods and ablation studies show that Stubborn achieved competitive performance, and the proposed probabilistic termination mechanism and adaptive sampling strategy contributed to the performance and robustness gains. For real-world demonstrations, please refer to https://aislab-sustech.github.io/Stubborn/.
arXiv:2606.12818v1 Announce Type: new Abstract: Irrelevant numbers in a prompt can shift language model judgments, producing anchoring effects in numerical reasoning. We study where this anchor-sensitive signal is carried inside language models using a controlled multiple-choice setup with shared answer options. We define a logit-difference metric comparing the correct answer option with the answer option corresponding to the anchor, and validate that it tracks behavioral anchoring. Using attribution-based circuit localization on 7B--8B Qwen and Llama base and instruction-tuned models, we find that edge-level methods recover this signal more faithfully than node-level methods. Low- and high-anchor circuits transfer strongly within a model, suggesting shared pathway structure across anchor direction. However, sparse transfer across base and instruction-tuned variants is less reliable, indicating that post-training changes which pathways matter most. Overall, our results provide a mechanistic account of how anchoring-related decision signals are carried inside language models.
arXiv:2606.12828v1 Announce Type: new Abstract: Do research topics in artificial intelligence grow gradually, or do they advance through abrupt, detectable jumps? Analyzing 80,814 accepted main-track papers from five premier AI conferences (ACL, CVPR, ICLR, ICML, NeurIPS) spanning 2017 to 2025, we show major AI topics advance through topical phase transitions: remaining marginal for years, then surging across venues within one to three years. Large language models became the dominant cross-venue topic by 2025, diffusion models rose with comparable abruptness, and language-model methods crossed into computer vision via vision-language models, whereas reinforcement learning compounded smoothly, distinguishing genuine phase transitions from ordinary growth. This structure is our primary contribution: a large-scale, cross-venue characterization of how AI research reorganizes. We then ask whether a transition leaves a detectable footprint before it peaks. We define an early-warning signature, four publication-dynamics criteria frozen on 2017-2021 data, and evaluate it out of sample on 2023-2025 transitions, obtaining a precision of 27% and recall of 63% against a 13.5% base rate. Applied to 2025 data, the signature flags reasoning and test-time compute, agentic AI, multimodal LLMs, retrieval-augmented generation, and world models as topics to monitor over 2026-2028. The source code is also publicly available on GitHub at https://github.com/KurbanIntelligenceLab/ai-phase-transitions.
arXiv:2606.12830v1 Announce Type: new Abstract: While recent vision-language models (VLMs) demonstrate strong multimodal understanding, they remain limited in spatial reasoning tasks that require active evidence acquisition and multi-step visual interaction. This limitation suggests that relying solely on implicit visual representations from vision encoders is insufficient for recovering fine-grained spatial evidence. We introduce PERception-Interaction-reason Agent (PERIA), a tool-augmented visual agent for spatial reasoning tasks across map reasoning, visual probing, and vision reconstruction. PERIA uses two lightweight tool families: vision perception tools for exposing textual, symbolic, and spatial evidence, and vision interaction tools for manipulating visual context, tracing paths, and verifying spatial relations. To train PERIA, we develop a unified recipe that combines supervised tool-use trajectory synthesis, composite rewards, and Observation-Relaxed Group-in-Group Policy Optimization (OR-GIGPO) for effective multi-tool behavior. Experiments on 13 benchmarks from 8 datasets show that PERIA-8B improves over the Qwen3-8B backbone by 10.0% on in-distribution benchmarks and 4.4% on out-of-distribution benchmarks, while outperforming previous state-of-the-art baselines of similar size by 7.0%-14.8%. It also achieves performance comparable to much larger models such as Qwen3-VL-235B-A22B-Thinking and GPT-5, demonstrating the effectiveness of PERIA in enhancing spatial reasoning capabilities.
arXiv:2606.12833v1 Announce Type: new Abstract: Spherical Interpolation of orDER $n$ (SIDER-$n$) is a recursive high-order interpolation method for data on the unit sphere $\mathbb{S}^2$, built from repeated spherical linear interpolation (SLERP). This paper develops a quaternion--Lie algebra framework for proving the local consistency of SIDER for smooth spherical curves sampled at equally spaced parameter values. Points on $\mathbb{S}$ are represented as pure unit quaternions, and interpolation errors are measured in fixed-base quaternion logarithmic coordinates. In this setting, each SLERP operation admits an exact Baker--Campbell--Hausdorff (BCH) representation, which converts the geometric interpolation problem into an algebraic problem involving filtered Lie-polynomial expansions. The BCH expansion shows that SLERP is affine to leading order, has no quadratic correction, and has a first nonlinear correction that is cubic and commutator-valued. Using this structure, we prove that SIDER2 has a third-order divided-error form with the same leading nodal factor as ordinary quadratic interpolation. We then show that the recursive SIDER step raises the order by one: the affine part gives the Neville-type finite-difference cancellation, while the nonlinear BCH remainder preserves the sharp filtered degree structure after the nodal factor is removed. Consequently, for every fixed $n\geq2$, $d_{\mathbb{S}^2}\bigl(\gamma(\theta h),P_i^{[n]}(\theta;h)\bigr) = O(h^{n+1}) $under the stated smoothness and small-stencil assumptions. The proof also identifies the shift-invariance of the leading divided-error coefficient as the algebraic compatibility condition underlying the SIDER recurrence.
arXiv:2606.12834v1 Announce Type: new Abstract: As scientific workflows shift from deterministic executables to LLM-based agents, the development practices on offer, such as fine-tuning, reinforcement learning, and prompt-and-go, bury the scientist's judgment. We propose treating agent construction as a workflow stage and introduce AgentBuild, which builds a scientific agent from a contract the scientist authors. The contract is a version-controlled rubric, a difficulty-graded curriculum, and a curated external knowledge base. A rubric-driven judge gates a meta-optimizer coding agent that edits the agent within a declared boundary, so the build compiles the agent, not the scientist's judgment. We instantiate this for Rietveld refinement of X-ray diffraction data through GSAS-II behind MCP and A2A, where a blank-harness construction run progresses through a lithium lanthanum zirconium oxide (LLZO) signal-to-noise ladder, reaches the 4 hour scan as a frontier case, and exposes the workflow-scope limits that remain. The same rubric that rewards credible fits also scores trajectory scope, making the frontier a contract failure rather than a pattern-fitting failure. As base models evolve, re-running AgentBuild is a re-tune, not a rebuild, and the scientist's authored contract remains the durable asset.
arXiv:2606.12836v1 Announce Type: new Abstract: Here we address the problem of estimating the dimensionality and nature of parametric variation in an unknown generative process directly from time-series data, without specifying or fitting a model. In particular we suppose that inter-instance variation in collections of time series is caused by parametric variation in the generating model. We hypothesize that, given a sufficiently large library of time-series features, low-dimensional parametric variation will manifest as low-dimensional structure in feature space, enabling interpretable estimators of the underlying degrees of freedom to be constructed. We test our hypothesis using a library of over 7000 diverse and interpretable time-series statistics and thirteen simulated systems with known parametric variation, spanning linear stochastic processes, nonlinear oscillators, and chaotic dynamics. Our unsupervised, data-driven approach often reconstructs the underlying parametric variation across this extensive range of simulated dynamical systems while also yielding interpretable estimators for each underlying dimension. Applied to the movement dynamics of 1143 fruit flies, we use this method to extract biologically meaningful components corresponding to sex and circadian rhythmicity. Our results pave the way for much-needed data-driven methods to bridge the gap between interpretable theoretical understanding of dynamics and the large and complex datasets that characterize modern scientific problems.
arXiv:2606.12840v1 Announce Type: new Abstract: Regression trees are among the most interpretable yet expressive model classes in machine learning. Historically, greedy induction has been the dominant approach for constructing well-performing regression trees. While optimal methods based on dynamic programming and branch-and-bound exist, they are computationally prohibitive for general linear regression trees, despite often achieving substantially better performance than greedy approaches. Recent work has shown that specialized lookahead strategies can dramatically improve runtime while maintaining near-optimal performance, primarily in classification settings. In this work, we develop a novel algorithm for near-optimal, sparse, piecewise linear regression trees that combines a lookahead-style search strategy with efficient rank-one Cholesky updates of the Gram matrix. We demonstrate, both theoretically and empirically, that our method achieves a favorable trade-off between computational efficiency, predictive accuracy, and sparsity, and scales significantly better than the current state of the art.
arXiv:2606.12847v1 Announce Type: new Abstract: The Abstraction and Reasoning Corpus (ARC) is viewed as a critical avenue to Artificial General Intelligence (AGI), as it enables models to learn abstract transformation rules from few-shot examples and then generalize to new tasks. However, prevalent ARC methodology is either pure language or vision-only (i.e., VARC). The former depends heavily on LLMs, consuming billions of parameters. The latter often struggles to capture high-level semantics, leading to overfitting on pixel-level patterns. To bridge this gap, we propose L-VARC, a novel framework that enhances visual reasoning via a language-guided Learning Using Privileged Information (LUPI) branch. Specifically, we design a Semantic Compression Module by feeding a unified, task-agnostic prompt into DeepSeek-V3. In this way, the raw LARC (a crowd-sourced language description dataset) can be substantially refined and structured, fitting with the context length constraint of standard text encoders (e.g., CLIP). Moreover, we design a Cross-Attention Projector to align visual features with semantic embeddings, aiming to guide the training of the ARC model. Notably, the LUPI branch is taken in the training process and will be discarded during inference, thereby yielding a lightweight model with a mere 18 million parameters. Extensive experiments demonstrate that our L-VARC effectively leverages linguistic priors to boost visual reasoning and outperforms state-of-the-art. Ablation studies further confirm the contribution of the two new designs towards the L-VARC framework. The code is available at https://github.com/GZHU-DVL/L-VARC.
arXiv:2606.12849v1 Announce Type: new Abstract: Semantic mapping is a core service that enables grounded interactions in emerging Extended Reality (XR) applications such as AI assistants and spatial object search. Deploying this capability on mobile XR devices requires a system that is open-vocabulary, real-time, and low-power. Existing approaches are compute-intensive and assume server-class resources. Cloud offloading offers a practical path, but no existing system splits semantic mapping across the device-cloud boundary or manages its communication, execution, and memory footprint. We present SemanticXR, the first device-cloud system for real-time, open-vocabulary semantic mapping and querying under XR power, bandwidth, and memory constraints. Our key insight is to elevate semantically identifiable objects to first-class units of communication, execution, and memory across the device and server. On the server, object-level parallelism and geometry downsampling improve mapping latency, while object-level depth-mapping co-design reduces upstream bandwidth. On the device, an object-level sparse local map with incremental updates and update prioritization enables network-robust querying with bounded memory and downstream bandwidth. Object-level configurable resource usage vs. quality trade-offs let applications and the system adapt mapping to application requirements and operating conditions, respectively. Against a device-cloud baseline with the same perception models, object-level organization improves server-side mapping latency by 2.2X at equal semantic quality. Depth-mapping co-design maintains upstream bandwidth under 2.5 Mbps. On the device, SemanticXR sustains sub-100 ms query latency for up to 10,000 objects even under network drops, supports tens of thousands of objects within 500 MB, and scales downstream bandwidth with map changes, not total scene size. The system adds only 2% device power during normal operation.
arXiv:2606.12852v1 Announce Type: new Abstract: Rapid advances have been made in developing general-purpose embodied agent in environments like Minecraft through the adoption of LLM-augmented hierarchical approaches. Despite their promise, low-level controllers often become performance bottlenecks due to repeated execution failures. We argue that a key limitation is not only the lack of episodic memory, but also the decoupling of \textit{what-where-when} memory from \textit{which-why} reasoning. To address this, we propose \textbf{WISE} (Which-Why Informed Semantic Explorer), a long-horizon agent framework with an enhanced low-level controller equipped with a Causal Event Graph that augments episodic memory with explicit causal structure linking observations to task relevance. Unlike prior work such as MrSteve, which relies on feature similarity for retrieval, WISE enables robust recall under viewpoint changes and supports opportunistic task reordering through causal reasoning. Building on this memory, we propose an Opportunistic Task Scheduler that dynamically re-prioritizes subtasks when causally relevant opportunities are detected. We further equip WISE with a multi-scale progressive exploration strategy to provide spatially comprehensive observations for downstream reasoning. Experiments show that WISE largely improves task success and efficiency on long-horizon sparse tasks, particularly in settings requiring adaptive decision-making.