arXiv:2606.03807v1 Announce Type: new
Abstract: We initiate the study of the algorithmic complexity of finding collisions in single-layer binary neural networks. Given a random matrix $\mathbf{A} \in \mathbb{R}^{m\times n}$, an input $\mathbf{x} \in \{-1,1\}^n$ is mapped to a binary output vector $\varphi(\mathbf{A}\mathbf{x})\in \{-1,1\}^m$, where $\varphi$ is an activation function with constant behavior on $[\kappa, \infty)$ for some threshold $\kappa \geq 0$.
We identify the threshold scale $\kappa=\Theta(1/\sqrt{\alpha})$, where $\alpha=m/n$, as separating two complementary phenomena. When $\kappa \ll 1/\sqrt{\alpha}$, we give a simple online algorithm that efficiently produces extensive collisions. When $\kappa \gg 1/\sqrt{\alpha}$, for a natural \emph{randomized} non-periodic activation and suitable oscillation complexity, we prove that the extensive-collision space exhibits an overlap gap property (OGP), yielding an exponential lower bound against online algorithms.
Ours is the first work to use the overlap gap property as a rigorous criterion for collision resistance. The key difference between collision finding and average-case search is that collision finding has a new ``worst-case'' aspect: the collision finder has full control over the choice of colliding pairs. Our lower bound is proved in the online model; extending such guarantees to broader classes of algorithms, including spectral, algebraic, lattice-based, or quantum methods, remains an open direction.
Science Journals
arXiv:2606.02954v1 Announce Type: new
Abstract: We report the demonstration of regrown epitaxial AlN gate interlayers with ultra-wide bandgap (UWBG) AlGaN polarization-graded field effect transistors (PolFETs). The introduction of the epitaxial AlN gate interlayer enables significant improvement in breakdown strength, with average breakdown field exceeding 6.94 MV/cm, which represents state-of-the-art for lateral field effect transistors, while maintaining excellent on-state current density exceeding 1 A/mm. The integration of epitaxial AlN enables state-of-the-art power-switching figure of merit exceeding 1.87 GW/cm2 at a breakdown voltage exceeding 1.45 kV. This work shows the potential of UWBG AlGaN for next-generation high-power switching and RF applications with enhanced device performance established by a high-quality epitaxially regrown AlN gate interlayer.
DeskCraft: Benchmarking Desktop Agents on Professional Workflows and Human-in-the-Loop Collaboration
arXiv:2606.03103v1 Announce Type: new
Abstract: Real-world professional desktop workflows in specialized creative and engineering software unfold over long horizons and often require human-in-the-loop coordination, where agents proactively seek necessary information and users provide additional instructions, clarifications, feedback, or corrections as the task progresses. Yet existing desktop GUI benchmarks mostly reduce this setting to short, simplified tasks with all user instructions provided upfront. To address this issue, we introduce DeskCraft, a desktop GUI benchmark targeting long horizon creative and engineering workflows and proactive human-agent collaboration. DeskCraft organizes tasks into a multilevel difficulty taxonomy, with long horizon tasks requiring over 50 execution steps, and covers professional creative software across design, video, audio, and 3D creation. Furthermore, DeskCraft formalizes human-agent collaboration into an interaction protocol covering mid-turn and post-turn exchanges. Mid-turn interaction captures both agent-initiated clarification under uncertainty and user-initiated interruption during execution, while post-turn interaction accommodates user-driven feedback after the agent signals completion, together spanning the full space of realistic collaboration patterns. We evaluate 18 proprietary and open source agents on 538 tasks and find that GPT-5.4 reaches 31.6% on standard tasks and 27.6% on interactive tasks. Further analyses reveal persistent failures in long horizon workflow delivery and proactive clarification. We will open-source all evaluation codes, tasks, and data at https://github.com/mrwwk/DeskCraft.
arXiv:2606.00809v2 Announce Type: replace
Abstract: Many real-world conversational settings for knowledge discovery, including podcasts, hiring screens, and marketplaces, require a purpose-driven understanding of a person. We study the Next-Best-Question (NBQ) problem: at each turn, an interviewer should ask the question with the highest expected information gain given what has already been learned and the conversation goal. We propose NBQ, a plug-and-play framework that seeds a diverse pool of candidate questions, maintains a compact and continuously updated user state, adaptively selects the next question within a turn budget, and distills the resulting free-form dialogue into a structured vector-based user profile. As a demanding application, we instantiate NBQ for reciprocal matchmaking, where compatibility must be mutual and each person is modeled by both self-description and counterpart-preference representations. To support large-scale matching, we further introduce QuickMatch, an efficient retrieval layer that recasts reciprocal matching from quadratic pairwise scoring to approximate vector search. Experiments show that NBQ improves user profiling quality by up to 13.6% and 14.0% in AC@T and AR@T, respectively, while QuickMatch accelerates retrieval by up to 22.9x with recall up to 0.989.
arXiv:2601.00353v4 Announce Type: replace
Abstract: Resource-constrained Internet of Things (IoT) devices, from medical implants to small drones, must transmit sensitive telemetry under adversarial wireless channels while operating under stringent computing and energy budgets. Authenticated Encryption (AE) is essential to ensure confidentiality, integrity, and authenticity. However, existing lightweight AE standards lack forward-security guarantees, compact tag aggregation, and offline-online (OO) optimizations required for modern high-throughput IoT pipelines. We introduce Diamond , the first provably secure Forward-secure and Aggregate Authenticated Encryption (FAAE) framework that extends and generalizes prior FAAE constructions through a lightweight key evolution mechanism, an OOoptimized computation pipeline, and a set of performance-tier instantiations. Diamond substantially reduces amortized offline preprocessing (up to 47%) and achieves up to an order-of-magnitude reduction in end-toend latency for large telemetry batches. Our comprehensive evaluation on 64-bit ARM Cortex-A72, 32-bit ARM Cortex-M4 and 8-bit AVR architectures confirms that Diamond outperforms baseline FAAE variants in authenticated encryption throughput and end-to-end verification latency while maintaining compact tag aggregation and strong breach resilience. Diamond outperforms NIST lightweight AE candidates for medium and large payloads, while remaining competitive for small messages when amortized across batches. We formally prove the security of Diamond and provide two concrete instantiations optimized for compliance and high efficiency. Our open-source release enables reproducibility and seamless integration into IoT platforms.
arXiv:2606.02332v2 Announce Type: replace
Abstract: Combining attention's global retrieval with the sequential importance signal of state space models (SSMs) is the open challenge of hybrid language modeling. Transformers see everywhere but cannot prioritize; SSMs know what matters but cannot revisit. Existing hybrids -- Jamba (block level) and Hymba (head level) -- place the two in separate compartments, so neither informs the other during the attention computation itself. We propose SISA (SSM-Informed Softmax Attention), which adds an SSM-derived importance term directly inside the attention score and realizes the full operation as a single SDPA call on augmented query/key vectors -- no recurrent state, no custom kernel. At 152M / 5B tokens, SISA reaches LAMBADA-greedy 17.3% (vs. Transformer 13.9 and Mamba-3 15.5) and attains NIAH 100% from step 1K, 7x faster than Transformer's retrieval convergence; at 369M, Mamba-3 leads LAMBADA while SISA preserves perfect NIAH and stock-SDPA execution. SISA thus defines a third design axis for SSM-attention hybrids -- score-level fusion -- beyond the block-level and head-level paradigms that have dominated the field.
arXiv:2606.02616v1 Announce Type: new
Abstract: Background in General Relativity (e.g. black holes, wormholes, or spacetime models in general) is needed to comprehend more recent efforts around understanding quantum phenomena like entanglement (e.g. >>It from qubit<< as well as >>ER = EPR<<). The former in turn requires a lot of knowledge from differential topology and differential geometry. While this knowledge is available in very good mathematics textbooks, it is scattered i.e. quite a bunch of sources need to be consulted to acquire it. The goal of this contribution is to provide the major background in a single place; in this sense, this contribution is some sort of guided tour through the corresponding literature.
arXiv:2606.03528v1 Announce Type: new
Abstract: This paper addresses distributed throughput optimization for dense multi-AP IEEE P802.11bq networks. We develop a packet-level model that jointly captures cross-link carrier-sense multiple access with collision avoidance (CSMA/CA), sub-7GHz RTS/CTS exchange, beam-training overhead, directional mmWave interference, signal-to-interference-plus-noise-ratio (SINR)-based MCS selection, and retransmissions. The resulting configuration problem is formulated as a multi-group combinatorial multi-armed bandit (CMAB), where each AP selects its contention window, clear-channel assessment threshold, beamwidth, and MCS reservation margin from finite candidate sets. Inspired by combinatorial successive accept-reject methods, we propose a group-wise feasible CSAR variant that uses Hadamard-guided feasible exploration to estimate empirical ranking scores and eliminate low-performing candidates within each parameter group. Simulations show that the proposed scheme improves aggregate and per-AP throughput over the considered Thompson-sampling baseline across most AP densities and reduces throughput stabilization time by approximately 49$\%$ under the evaluated settings. The learned configurations reveal that high throughput requires a balance among control-channel aggressiveness, mmWave spatial reuse, beam-training cost, and MCS robustness, rather than simply minimizing collisions or maximizing the PHY rate.
arXiv:2303.15619v2 Announce Type: replace
Abstract: The choice of \emph{which} tokens to mask is a central, under-examined design decision in masked language modeling (MLM). Standard pretraining masks tokens uniformly at random, but several studies show that more informative masking targets can improve downstream performance. We study masking as a \emph{task-adaptive} component of the fine-tuning pipeline and introduce \textbf{Typhoon}, a masking strategy that uses the gradient of the task loss with respect to one-hot token inputs to estimate, online, how much each token type contributes to the objective. Typhoon maintains an exponential moving average of per-token-type saliency and calibrates these scores into a masking distribution whose expected masking rate matches a target budget, under a token-independence approximation. We formalize the method and evaluate it against random masking and whole-word masking on two GLUE tasks, MRPC and CoLA, across three BERT-family backbones (TinyBERT, DistilBERT, and BERT-base) and five random seeds per configuration ($90$ training runs in total). Our main finding is that, once seed variance is accounted for, no masking strategy is reliably better than the others on these tasks: on MRPC the gap between Typhoon and the best baseline stays within $0.004$ $F_1$, across all twelve Typhoon comparisons no paired test reaches significance, and every $95\%$ confidence interval contains zero. Typhoon's apparent advantage in single-run experiments does not survive this more careful evaluation. We read this as a cautionary, reproducibility-focused result -- gradient-based task-adaptive masking is competitive but not clearly better than resource-free random masking at this scale -- and we describe a clean modern reimplementation to support follow-up work.
arXiv:2606.02981v1 Announce Type: new
Abstract: Best-of-$N$ inference scaling (drawing $N$ candidate answers from a language model and returning the one a reward model ranks highest) improves accuracy by an amount that varies across models, but predicting that amount in advance currently requires running the procedure end-to-end. Prior work links cheap statistics of a model's sampled outputs and validation-set correctness (how often samples agree, how diverse they are, how confident the model is, and where correct samples appear) to model behavior, but does not isolate which of these form a stable, compact predictor of best-of-$N$ gain. We fit ridge predictors on features computed from a single labeled validation-set sampling pass, use bootstrap-Lasso as a stability analysis of the candidate feature set, and give a concentration analysis with an explicit linear-approximation residual. Across three base-model families, six post-training methods, and math and reasoning task domains, the stability analysis identifies a strict three-feature core spanning prompt-level agreement spread, label-assisted first-correct-sample position, and completion-length variance; a compact ridge predictor built from this core plus an entropy add-on reaches Spearman $\rho = 0.90$ with actual best-of-$N$ gain under a reward-model verifier. The intended use is labeled validation-set screening of candidate configurations before paying the full reward-model scoring cost.
arXiv:2606.02993v1 Announce Type: new
Abstract: Understanding how structured internal structure emerges during neural network training is central to the study of deep learning. We investigate this phenomenon through the group composition task, where a two-layer neural network is trained to predict $g_1 \star g_2$ for elements of a finite group $G$. By lifting the projected gradient flow to the Fourier domain, we demonstrate that the training dynamics are governed by a Riemannian gradient ascent on a representation-theoretic energy functional. We prove that, under random initialization, this flow drives each neuron to converge almost surely toward a single irreducible representation, while the cross-layer Fourier coefficients achieve a rotational rank-one alignment. This framework provides a representation-theoretic account of feature learning and characterizes a novel low-rank compression phenomenon for matrix-valued group representations. Moreover, for Abelian groups, we provide a complete population-level description: random initialization promotes uniform diversification across nontrivial representations and induces Haar-uniform phases, jointly approximating the indicator via a majority-vote mechanism. We further prove that both phase alignment and representation competition emerge with exponential convergence rates.
arXiv:2602.10949v2 Announce Type: replace-cross
Abstract: Effective initialization in deep networks requires an understanding of random neural networks. In this work, a rigorous probabilistic analysis of deep bias-free random Leaky ReLU networks is provided. We prove a Law of Large Numbers and a Central Limit Theorem for the logarithm of the norm of network activations, establishing that, as the number of layers increases, their growth is governed by a parameter called the Lyapunov exponent. This parameter characterizes a sharp phase transition between vanishing and exploding activations, and we calculate the Lyapunov exponent explicitly for Gaussian or orthogonal weight matrices. Our results reveal that standard methods, such as He initialization or orthogonal initialization, do not guarantee activation stability for deep networks of low width. Based on these theoretical insights, we propose a novel initialization method, referred to as Lyapunov initialization, which sets the Lyapunov exponent to zero and thereby ensures that the neural network is as stable as possible, leading empirically to improved learning.
arXiv:2606.03929v1 Announce Type: new
Abstract: Colinear chaining is a classical heuristic for sequence alignment: it enables scalable genome comparison and is a main component of many state-of-the-art read mappers based on seed-chain-extend. The earliest $O(n \log \log n)$ time algorithms by Eppstein et al. (J. ACM, 1992) chained $n$ fragments between two sequences $T$ and $Q$ while minimizing a gap cost based on the diagonal distance $\Delta_{\text{diag}}$ between consecutive fragments. They also forbid fragment overlaps, which are essential in current chaining formulations: in long-read mapping, overlaps improve sensitivity and avoid restrictions on the fragment class considered. Jain, Gibney, and Thankachan (J. Comput. Biol. 2022) recently combined a $\Delta_{\text{diag}} = |\Delta_T -\Delta_Q|$ overlap cost with the classic $L_\infty = \max(\Delta_T , \Delta_Q)$ gap cost that takes the maximum between the horizontal and vertical gap between the fragments and they proved that chaining under this cost model is equivalent to the anchored edit distance. We improve the existing $O(n \log^3 n)$-time algorithm for anchored edit distance to $O(n \log \log n)$ time in $O(n)$ space, by combining the gap-cost computation of Chao and Miller (Algorithmica, 1995) with the overlap-cost computation of Baker and Giancarlo (ESA, 1998). By developing llchain, a simpler $O(n \log n)$-time implementation of our method, we show how chaining algorithms that might have been recently overlooked by the bioinformatics community scale competitively to millions of fragments and large genomes. On average, llchain is $10\times$ faster than other methods on instances with $3\,000\,000$ anchors, and over $3\times$ faster on MEMs between HiFi reads and a reference human genome.
arXiv:2606.03825v1 Announce Type: new
Abstract: Transformers have become the dominant architecture for large language models, largely due to the scalability and flexibility of attention, feed-forward layers, residual connections, and normalization. This paper introduces dynamic short convolutions as an additional neural network primitive for improving Transformers. Unlike static short convolutions, dynamic convolutions use input-dependent filters, which preserves the locality bias of convolution while increasing expressivity. Motivating experiments show that applying dynamic short convolutions to key, query, and value representations improves performance on challenging associative recall tasks compared with static convolutional variants. Across language-modeling experiments ranging from 150M to 2B parameters, dynamic convolutions consistently outperform standard Transformers and Transformers augmented with static short convolutions. Fitting scaling laws indicates a 1.33$\times$ compute advantage over compute-matched Transformers when dynamic convolutions are applied to the key, query, and value vectors, and a 1.60$\times$ advantage when adding dynamic convolutions after every linear layer. Dynamic convolutions also offer improvements on linear RNNs (Mamba-2/Gated DeltaNet) and mixture-of-experts architectures. We make these gains practical with custom Triton kernels that enable efficient training with a manageable end-to-end slowdown. These results suggest that dynamic short convolutions are a scalable, hardware-efficient, and expressive primitive for advancing Transformer-based language models.
arXiv:2606.03194v1 Announce Type: new
Abstract: The Membership Problem for Pedigree Polytope (M3P) asks, given $X\in\mathbb{Q}^{\binom{n}{3}}$, whether $X\in\mathrm{conv}(P_n)$, where $P_n$ is the set of all pedigrees. A pedigree is a structured encoding of a Hamiltonian cycle construction in $K_n$. We establish that M3P is solvable in strongly polynomial time via a recursively constructed layered network $(N_k, R_k, \mu)$ and a multicommodity flow problem MCF$(k)$. The necessary and sufficient condition for membership established is that the optimal total flow in MCF$(n-1)$ equals the maximum possible flow $z_{\max}$. The complexity analysis, grounded in Tardos's strongly polynomial algorithm for combinatorial linear programs (1986), shows that this condition can be checked in strongly polynomial time in the dimension of the matrix involved. By sufficiency, this implies M3P~$\in$~P. Since the Symmetric Travelling Salesman Problem (STSP) reduces to M3P via the Multistage Insertion (MI) formulation (Arthanari 1983), STSP is solvable in polynomial time, and the P vs.NP question is resolved. The proofs leading to this result are fully machine-verified in Lean~4/Mathlib4, with zero unresolved \texttt{sorry}s in the main proof chain. The main contribution is the Lean~4 machine verification of all proofs in the main chain, resulting in \texttt{theorem p\_equals\_np}: P = NP. The Lean~4 formal verification covers the sufficiency of MCF(n-1) for membership in $\mathrm{conv}(P_n)$, and the P = NP chain via Maurras (2002), Gr\"{o}tschel--Lov\'{a}sz--Schrijver (1988), Cook (1971), and Karp (1972). The complete lean project (36 Lean~4 files, 2968/2968 build targets clean) is available at https://github.com/TiruArt/Pedigree-Polytopes-Lean4.
arXiv:2606.03264v1 Announce Type: new
Abstract: We introduce PaddleOCR-VL-1.6, an upgraded compact document parsing model built upon PaddleOCR-VL-1.5. Although PaddleOCR-VL-1.5 establishes a strong 0.9B baseline, its remaining errors concentrate in under-optimized regions where model behavior is unstable, data coverage is sparse, or supervision is unreliable. Rather than expanding the training corpus indiscriminately, PaddleOCR-VL-1.6 introduces a region-aware data optimization framework that identifies weak regions from the previous model, applies targeted enhancement to these regions, and improves the reliability of supervision signals. It further adopts a progressive post-training recipe based on curated data selection and reinforcement learning, pushing model performance to a higher level through staged optimization. PaddleOCR-VL-1.6 achieves a new state-of-the-art score of 96.33% on OmniDocBench v1.6, demonstrates strong competitiveness against top-tier VLMs, and provides a practical post-training recipe for the PaddleOCR-VL series.
arXiv:2606.03144v1 Announce Type: new
Abstract: Large language models (LLMs) are increasingly used as self-study assistants in technical disciplines, yet their reliability as mathematical reasoning assistants remains poorly understood. We introduce GTBench, a curriculum-grounded benchmark for evaluating LLMs as mathematical research assistants in graph theory, comprising 63 problems organized into three groups of increasing difficulty: undergraduate definitions and basic properties (Group 1), algorithm tracing and structural reasoning (Group 2), and graduate-level proof construction (Group 3). Problems are sourced from verified academic materials including Diestel's Graph Theory. We evaluate five frontier models -- GPT-5, Claude Sonnet 4.6, Gemini 2.5 Flash-Lite, Llama 3.3 70B, and Mistral Large 3 -- under zero-shot and chain-of-thought prompting, using exact-match and LLM-as-judge evaluation for Groups 1 and 2, and a hybrid human expert and LLM-as-judge protocol for Group 3. Our results reveal a pronounced performance hierarchy: GPT-5 approaches ceiling on Group 1 (95.8% zero-shot) and maintains meaningful accuracy on graduate proofs (82%), while all other models degrade substantially with difficulty, with Llama achieving 0% under human evaluation on Group 3 zero-shot. Failure mode analysis shows that correct algorithm, wrong execution errors dominate Groups 1 and 2, while Group 3 additionally surfaces incomplete reasoning failures and reveals systematic disagreement between human evaluators and the automated judge, particularly on verbose or near-complete proofs (kappa = 0.48-0.83 across human pairs). GTBench provides the first curriculum-grounded evaluation framework for graph-theoretic reasoning in LLMs, with direct implications for the governance of AI tools in mathematical education and scientific research.
arXiv:2606.03847v1 Announce Type: new
Abstract: Action chunking has become a common inference strategy for flow-based robot policies, improving action coherence by modeling multi-step temporal dependencies in demonstrations. However, the execution horizon is still typically set as an empirical fixed value, overlooking that predictable free-space motions and precision-critical interaction phases often require different replanning frequencies. In this work, we first show that the denoising process of flow-based policies contains an intrinsic signal of task phases: clean-action estimates remain stable during predictable motion phases, but fluctuate more strongly around contact-rich or precision-sensitive operations. Motivated by this observation, we propose DVAC (Denoising-Variance Adaptive Chunking), a test-time method that adaptively determines how many actions to execute from each predicted chunk. DVAC measures the variance of clean-action estimates over the final denoising steps, executes the stable low-variance prefix, and replans before high-variance future actions are committed. To transfer across tasks and rollouts, DVAC further calibrates the threshold with a rolling estimate of the local variance scale. Experiments on LIBERO, RoboTwin, CALVIN, and real-world manipulation show that DVAC improves task success while reducing replanning frequency. With a $\pi_{0.5}$-based policy, DVAC improves LIBERO success from 94.75% to 98.00% and reduces replanning by 43.0%, while also yielding aggregate gains on RoboTwin and CALVIN and improving real-world execution efficiency.
arXiv:2606.03943v1 Announce Type: new
Abstract: Video-Action Models (VAMs) leverage the broad visual dynamics captured by pre-trained video diffusion models, offering a promising path toward generalizable robot manipulation. However, RGB-only video rollouts are not directly actionable: they leave metric 3D motion, contact geometry, and fine-grained spatial constraints under-specified, making action grounding ambiguous. Meanwhile, scaling action supervision across diverse tasks and embodiments remains costly. We present PointAction, a framework that bridges video predictions to robot actions through explicit point-based 4D modeling. PointAction fine-tunes a foundation video generation model to jointly predict future RGB frames and dynamic 3D pointmaps, producing temporally consistent 3D motion of task-relevant scene geometry. These point dynamics serve as a structured, embodiment-agnostic action interface, which a diffusion-based action decoder maps to executable robot actions. By using metric 3D point dynamics as the interface between video prediction and control, PointAction reduces the ambiguity of RGB-only action grounding and supports transfer across tasks and embodiments with limited action supervision. Experiments show that PointAction achieves state-of-the-art 4D generation quality on robot scenes, outperforms existing baselines in simulation, and generalizes to two real robot arms unseen during pretraining.
arXiv:2606.03567v1 Announce Type: new
Abstract: This paper introduces a systematic gray-box identification framework for voltage-source converter models based solely on terminal time-series data. The proposed approach combines a physically informed white-box standard model with iterative time-domain calibration to estimate controller parameters that mimic the behavior of the black-box model in electromagnetic transient (EMT) simulations. Unlike conventional frequency-domain identification methods, the framework leverages time-domain data more effectively to better constrain the surrogate model across a broader operating range and capture reference-signal dynamics. To evaluate the accuracy of the identified model, the paper presents additional frequency-domain validation metrics based on Nyquist analysis and singular value decomposition, allowing for both quantitative assessment of model divergence and qualitative classification of mismatch types. The methodology is tested on cases with increasing structural uncertainty, from exact parametric recovery to an actual detailed EMT black-box model. Results demonstrate that the proposed framework can accurately recover parameters when the internal structure is known, adjust for moderate structural mismatch with extra degrees of freedom, and offer a reliability measure for small-signal stability analysis of converter models protected by intellectual property
arXiv:2512.21235v3 Announce Type: replace
Abstract: Imitation learning from human demonstrations has become a dominant approach for training autonomous robot policies. However, collecting demonstration datasets is costly: it often requires access to robots and needs sustained effort in a tedious, long process. These factors limit the scale of data available for training policies. We aim to address this scalability challenge by involving a broader audience in a gamified data collection experience that is both accessible and motivating. Specifically, we develop a gamified remote teleoperation platform, RoboCade, to engage general users in collecting data that is beneficial for downstream policy training. To do this, we embed gamification strategies into the design of the system interface and data collection tasks. In the system interface, we include components such as visual feedback, sound effects, goal visualizations, progress bars, leaderboards, and badges. We additionally propose principles for constructing gamified tasks that have overlapping structure with useful downstream target tasks. We instantiate RoboCade on three manipulation tasks -- including spatial arrangement, scanning, and insertion. To illustrate the viability of gamified robot data collection, we collect a demonstration dataset through our platform, and show that co-training robot policies with this data can improve success rate on non-gamified target tasks (+16-56%). Further, we conduct a user study to validate that novice users find the gamified platform significantly more enjoyable than a standard non-gamified platform (+24%). These results highlight the promise of gamified data collection as a scalable, accessible, and engaging method for collecting demonstration data.
arXiv:2606.03887v1 Announce Type: new
Abstract: The growing penetration of distributed energy resources (DERs) intensifies congestion in distribution networks by introducing bidirectional power flows and increasing competition for limited network capacity, underscoring the need for effective and efficient congestion management, including flexible grid-access schemes. This paper proposes a bilevel optimization model for the dynamic allocation of connection capacity to DERs under non-firm connection agreements, aligning the objectives of distribution system operator (DSO) and DER owners. The upper-level problem, representing the DSO, determines the allocated connection capacity for all DERs, defined as maximum time-varying power limits, subject to distribution system constraints and the last-in-first-out (LIFO) allocation rule. The lower-level problem, representing DER owners, maximizes the profit of each DER within the allocated power limits. The proposed model is tested on a modified CIGRE medium-voltage (MV) network, demonstrating a balanced trade-off between grid utilization and economic efficiency. Furthermore, the model enhances DER integration, enforces transparent allocation rules, reduces variability in allocation patterns, and achieves up to an 80% reduction in total curtailment costs compared with benchmark methods.
arXiv:2606.02747v1 Announce Type: new
Abstract: Planning records define restrictions over geographic areas, but their source documents often provide only indirect spatial evidence rather than machine-readable boundaries. We introduce Plan2Map, a 208-case multimodal benchmark for document-grounded geospatial boundary reconstruction from UK planning records. Given only a source planning document, systems must reconstruct a valid geospatial boundary from notice text, schedules, map plates, map labels, and boundary annotations; the reference GeoJSON is held out for scoring. We propose GeoPlanAgent, a document-grounded, geospatial-tool-in-the-loop system that decomposes the task into evidence extraction, localisation, map registration, boundary segmentation, projection, and verification. On Plan2Map, GeoPlanAgent achieves 0.736 mean IoU and 0.904 median IoU, with 67.8\% of predictions at or above 0.8 IoU, substantially outperforming direct VLM-to-GeoJSON baselines. Diagnostic analysis shows that direct VLM prediction remains unreliable, while remaining errors are concentrated in localisation and map registration, and supervised boundary segmentation substantially improves pixel-level mask quality. Plan2Map provides a concrete testbed for multimodal geospatial reconstruction from public planning records. Project page: https://odeb1.github.io/Plan2Map_Project_Page/.
arXiv:2606.02755v1 Announce Type: new
Abstract: Large language model (LLM) applications are increasingly expected to satisfy deterministic institutional requirements while relying on probabilistic generative components. This mismatch makes ordinary post-hoc benchmarking insufficient for systems that must be safe, reliable, auditable, and economically useful. This paper contributes an evaluation-protocol extension for operational LLM systems grounded in acceptance-test-driven development, safety engineering, and business-centric validation. The extension translates stakeholder goals into executable behavioral contracts, release gates, monitoring signals, and evidence artifacts before prompt, model, retrieval, or agent changes are accepted. It adapts the red-green-refactor discipline of test-driven development to a red-train-green lifecycle: first define failing acceptance tests for desired behavior, then improve the LLM system through prompt changes, retrieval design, fine-tuning, guardrails, or data augmentation, and finally release only when multidimensional gates are satisfied. The contribution is a governance-oriented metric stack, reference architecture, and empirical protocol for comparing acceptance-test-driven LLM development against prompt-first and benchmark-after workflows.
arXiv:2606.03939v1 Announce Type: new
Abstract: Federated Learning (FL) of foundation and edge models increasingly targets deployments where client data distributions drift over time, yet existing forgetting-mitigation methods assume each client's distribution is stationary. Flashback, the strongest recent FL method against cross-client (spatial) forgetting, uses monotonically accumulating per-class label counts as a knowledge proxy; this proxy becomes miscalibrated under temporal distribution shift and anchors the global model to an outdated class balance. We formalise temporal forgetting in FL with a per-phase metric isolated from protocol-level fluctuations and propose Flashback Continual Learning (FlashbackCL), a drop-in extension of Flashback with (i) temporally-decayed label counts; (ii) a device-aware replay buffer with Class-Balanced Reservoir Sampling (CBRS); and (iii) server-side active coreset curation on the public distillation set. The results show that FlashbackCL achieves 6.9% to 10.0% relative improvement relative to Flashback, on CIFAR-10 with 50 clients and three controlled temporal shift modes, while simultaneously reducing temporal forgetting by up to 68%. A 5-variant ablation identifies CBRS replay as the critical component. FlashbackCL also improves Flashback by 3.5 points on stationary CIFAR-100, suggesting that class-balanced replay regularises spatial heterogeneity as well as temporal shift.