Science & Mathematics

Capability claims in robot manipulation are unverifiable, and existing leaderboards are each closed, simulated, or gameable. We propose HATCH, a robot escape room: an openly specified, physically replicable, sensor-scored, and unrehearsable benchmark. Tasks span five skill categories at three difficulty levels and are chained so that the sensor confirming one task's completion unlocks the next, turning a room into a single long-horizon evaluation. Success is confirmed by instrumentation rather than human judgment, scoring is embodiment-agnostic, and object placements are randomized by a run-day seed. Entrants submit attested self-runs (uncut video, sensor logs, and the seed), with occasional organized events for calibration. Task specifications, build guide, bill of materials, and scoring rubrics are released under an open license.

Atomically thin transition metal dichalcogenides provide a promising material platform for optical phase manipulation, as rapid variations in their dielectric functions can give rise to topological phase singularities (PSs). However, PS evolution while preserving topological charge conservation has not yet been investigated in exciton enabled PS systems. Here, using atomically thin MoS2 films on sapphire as a model system, we theoretically and experimentally demonstrate that the evolution of dielectric dispersion determines the creation, trajectory, merging, and annihilation of optical PSs under topological charge conservation. By varying the MoS2 film thickness, we experimentally realize this evolution through the gradual shrinking of the high energy n-k dispersion loop, which modifies its intersection with the zero reflection condition. Our results provide a general strategy for exploring and manipulating topological optical states.

Practitioner guidance holds that a large language model reviews a document better when the work is divided among agents. We test a narrow form of it in a controlled comparison of single-agent and multi-agent setups on a financial task, reviewing credit assessment memos against requirements. One agent checks six requirements in one call. A team of three on the same model checks the same memo, two requirements each, in the same words, with no roles or later stage. Labels are known for every requirement on 300 synthetic memos, so misses and false alarms are measured separately. Two primary predictions are tested on four models at a Bonferroni-corrected level. Under an instruction to decide each requirement once, the team catches 0.08 to 0.10 more of the defects that carry an explanation on the three models whose single agent misses some, at the corrected level on two. It raises more false alarms on all four, at the corrected level on three. Balanced accuracy, an exploratory measure, changes by 0.02 or less, and the team uses 2.2 to 2.4 times the tokens. Showing each agent the requirements it was not given reduces the added false alarms on the two Claude models, at the corrected level on one. In exploratory comparisons, sampling the single agent three times and flagging on any draw gives higher balanced accuracy than the team on those models, and the team's results depend on the instruction.

Mitigating catastrophic risks from large language models requires excising hazardous operational capabilities, such as cyberwarfare exploitation and pathogen enhancement, without degrading benign scientific utility. In these dual-use domains, legitimate defensive science and malicious attacks share vocabulary, syntax, and internal model representations, causing standard unlearning methods to induce severe collateral damage (concept bleeding). In this work, we present Orthogonal Boundary Unlearning (OBU), a mathematically grounded framework designed to eliminate collateral damage on benign dual-use capabilities. OBU extracts a domain-matched capability subspace from paired calibration queries via singular value decomposition on empirical activation covariances, and restricts parameter updates strictly to the orthogonal null space. This guarantees that benign representations experience zero parameter perturbation along the spanned calibration subspace. We present two complementary realizations of OBU: Projected LoRA, which applies the orthogonal constraint during iterative gradient-based fine-tuning, and Closed-Form Weight Surgery, which directly edits model weights in a single analytical step without training. Across four modern open-weight foundation models (Gemma-3-4B, Qwen3.5-4B, SmolLM3-3B, and Granite-4.2-3B), OBU achieves strong dual-use concept separation, substantially mitigates in-context priming attacks, and demonstrates durable resistance against adversarial relearning without degrading benign domain competence.

Real-world audio typically contains concurrent sound events, yet standard self-supervised methods treat each clip as an isolated training unit without explicitly modeling mixture--constituent relationships. We propose Mixture-to-Constituent Latent Alignment (MiCLA), a two-stage self-supervised pre-training framework that uses synthetic log-mel mixing to align mixture representations with their constituent clips. Stage 1 structures single-clip representations via masked self-distillation and contrastive regularization. Stage 2 introduces bidirectional alignment: for a synthetic mixture, MiCLA suppresses the component aligned with one constituent and aligns the residual with the other, using projective residualization on patch tokens and sample-adaptive soft residualization on the global class token. In downstream evaluations, MiCLA achieves 43.2% mean average precision (mAP) on AS-20K and 50.3% mAP on AS-2M under fine-tuning, and achieves the highest mAP among evaluated models on both AudioSet settings under frozen Convex Gated Probing (CGP). Furthermore, a Joint Semantic Retrieval diagnostic shows that MiCLA improves simultaneous retrieval of both constituent classes over target-averaging baselines. These results indicate that directional mixture-to-constituent alignment provides an effective pre-training signal for polyphonic audio representation learning.

Self-evolving agents aim to improve from their own experience rather than remain fixed after deployment, but naturally collected trajectories often lack ground-truth solutions or verifier-derived rewards. We introduce Retrospective Self-Distillation (RSD), a multi-round framework that turns such experience into dense supervision using privileged retrospective context. For each on-policy trajectory, the same underlying model acts as both student and teacher: the teacher re-scores the student's realized actions while conditioning on cross-task precedents and lessons, together with turn-level reflection and future hindsight from the completed trajectory. The resulting teacher distribution is distilled back into the policy through on-policy self-distillation. Across rounds, the experience bank grows and the retrospective context is refreshed, allowing the model, its data, and its supervision to co-evolve. We further introduce frontier-guided context-policy evolution, where a stronger frontier model proposes constrained edits to how privileged context is constructed. Across ALFWorld, ScienceWorld, and SearchQA with Qwen3-8B and Qwen3-14B, RSD improves performance by up to 59.7% relative to the base model, while frontier-guided context-policy evolution adds up to a further 44.7% relative improvement over a fixed context policy. These results show that naturally accumulated experience can be converted into persistent model capability without ground-truth trajectories or verifier-derived rewards for the core policy update.

Foundation models now serve as reasoning layers for robot teams, but it remains unclear whether embodiment-specialized vision-language models (VLMs) preserve the social-reasoning capabilities of general-purpose VLMs. We evaluate Gemini Robotics-ER on SHREC, a benchmark of real human-robot social interactions, and compare it with published general-purpose VLM baselines. Robotics-ER outperforms GPT-4o-mini on error/competence detection but performs worse on detailed social-attribute judgments; adding a visual frame helps one task while hurting the other. These results suggest embodiment-specialized models may perform differently on social judgments that can be read more directly from observable interaction cues versus those requiring more inference about a person's internal state. Because the comparison spans model families, these differences cannot be attributed to embodiment specialization alone. This matters for multi-robot systems, where a shared foundation model could cause systematic social-perception errors to influence multiple coordination decisions. This work presents an initial evaluation, with broader cross-context and video-based analyses left for future work.

We develop and compare enhanced neural-network solvers for one-dimensional ground-state profiles, which are solutions of $$ Q^{\prime\prime}=bQ-γQ^{p},\quad Q>0,\quad p>1, \quad b, γ>0, $$ the stationary profile equation for the nonlinear Schrödinger, generalized Korteweg-de Vries, and nonlinear Klein-Gordon equations. Its explicit sech-type solutions serve as exact reference for error computation and benchmark hyperparameter selection. We improve on the standard Adam-trained physics-informed neural network (PINN) baseline by introducing higher-order optimization via Gauss-Newton/Levenberg-Marquardt and L-BFGS refinements, as well as Fourier-feature PINNs, and constrained Deep Ritz (DR) formulations. In DR approaches we introduce two new neural-network formulations, which implement two classical variational characterizations: a Nehari-constrained Deep Ritz method, in which the height is learned by projecting onto the Nehari/Pokhozhaev manifold, and a Weinstein Deep Ritz method, in which a normalized profile minimizes the Weinstein quotient. The Deep Ritz formulations perform best overall, in particular, the Weinstein variant gives the best error-runtime balance. The neural-network based solvers remain less accurate and efficient than classical Petviashvili iteration in 1D, however, the current study provides a proof of concept: the mesh-free formulation, non-reliance on exact profiles or the related features make them promising in higher dimensions and for more complicated nonlinearities and dispersive operators.

Heterogeneous multi-robot task allocation depends on estimates of how suitable individual robots are for candidate tasks, yet suitability can be defined in substantially different ways. We compare ten deterministic methods and two LLM-based variants spanning rule-based, similarity-based, requirement-oriented, fuzzy, and semantic approaches. Using common robot and task profiles, we compare how the methods score and rank candidate robots, identify feasibility, and select a top robot, while examining how these outcomes change with team composition, scale, spatial context, capability uncertainty, and task representation. The methods frequently disagree despite operating on identical scenarios. Generic similarity measures can violate properties expected under a one-sided requirement-satisfaction interpretation of suitability, while LLM-based scoring introduces additional stochastic and linguistic sensitivity. These results show that robot--task suitability should be treated as an explicit modeling choice whose assumptions, robustness, and decision-level consequences should be considered before allocation.

4D scene graphs require persistent object identity when an object disappears inside an opaque carrier and later reappears as a fresh detection. When appearance and target-centric motion are unavailable, identity recovery must rely on relational history. We formulate cross-visibility-gap identity persistence as historical-node recovery through a remembered target--carrier binding. We introduce CVG-SGM, a perception-frontend-compatible transfer--maintain--read memory. Geometry-weighted cross-node interaction transfers target-specific evidence into carrier states, a shared per-node GRU maintains this information during transport, and a two-time readout combines stored carrier--identity compatibility with current detection--carrier geometry to recover the historical identity. In the controlled one-identity-per-carrier setting, CVG-SGM achieves 1.000 re-link accuracy, compared with 0.570 for both no-memory and parameter-count-matched no-write controls. It also generalizes to scene-disjoint, unseen longer visibility gaps. On real RGB-D/SAM2 feature-level episodes, it correctly re-links 36/36 held-out detections, compared with 28/36 for the strongest fixed carrier--identity control. Counterfactual carrier-memory swaps redirect predictions according to the swapped bindings, supporting the role of stored relational information in identity recovery. The current approach remains limited by shared-carrier identity ambiguity and does not provide end-to-end raw-video re-identification.

Distance scans in partition-family approximate nearest-neighbor (ANN) search are typically limited by memory bandwidth. Existing systems therefore improve throughput by reading fewer bytes per candidate, often at the cost of recall. We present KatoANN, a CPU ANN search system that removes this tradeoff by batching distance scans on Intel AMX, a CPU matrix engine. Batching allows multiple queries to reuse each fetched vector code, shifting the scan from memory-bound to compute-bound and enabling near-exact distances at the speed of a lossy scan. KatoANN stores each vector once as a 16-bit SQ16 code split into high- and low-byte planes. Its adaptive rank-then-refine scan first ranks all candidates using only the high byte, then refines a small survivor set using the full-byte code. Both stages use the same representation, avoiding a second copy of the raw vectors. Experiments across diverse high-dimensional workloads show that KatoANN is the fastest partition-family ANN search system in our evaluation throughout the high-recall range.

Objectives: Complete Street design is currently guided largely by qualitative principles despite the increasing complexity of multimodal urban traffic. This paper develops a microscopic simulation framework for quantitatively evaluating alternative street designs by explicitly representing heterogeneous road users, including Autonomous Vehicles (AVs), Connected Vehicles (CVs), Human-Driven Vehicles (HDVs), pick-up and drop-off (PUDO) services, and micromobility users. The framework jointly evaluates mobility and safety within a unified testbed. Methods:The framework enriches microscopic simulation with behavioral models for heterogeneous road users. It explicitly captures lateral movements, parking maneuvers, pedestrian crossings and jaywalking, and shared-lane bicycle and scooter interactions. Safety is quantified using a conflict severity measure grounded in collision physics, while candidate street layouts are optimized using a genetic algorithm. Findings: Applied to a segment of Belmont Avenue in Chicago, Illinois, the framework shows that dedicated bicycle lanes reduce conflict severity at the expense of vehicular throughput. Increasing AV penetration mitigates congestion by reducing traffic shockwaves while improving interactions with pedestrians relative to CVs and HDVs. The optimization framework efficiently identifies Pareto-efficient street layouts that balance mobility and safety. Novelty: The proposed framework integrates heterogeneous behavioral modeling, physics-based conflict assessment, stochastic performance evaluation, and design optimization within a single quantitative framework for Complete Street design. Practical Applications: The framework provides transportation planners with a predictive virtual testbed for evaluating street designs and quantifying the tradeoff between mobility and safety before implementation.

Fine-grained liquid identification requires a classifier to distinguish similar substances from compact millimeter-wave measurements. We investigate structured-token attention, which represents a radar feature tensor through complementary antenna, feature, branch, and global views. The proposed StructAttn model embeds these views into eleven tokens and combines them with a three-layer transformer. Experiments on the public FG-LiquID data cover 30 liquids and 3,000 samples. Over five seeds on the released split, StructAttn achieves 96.70\% mean best-epoch accuracy, matching the original-recipe MMNet baseline and exceeding a scalar-token attention model by 1.50 percentage points. A five-seed probability ensemble reaches 97\%, corresponding to 18 errors among 600 test samples. Final-epoch and five-fold results complement the checkpoint-selected comparison, with StructAttn obtaining 95.30\% out-of-fold accuracy. Token and feature analyses show that structured views are useful for prediction, while residual confusions concentrate in water, soft-drink, and edible-oil classes. These results support structured tokenization as a practical design choice for attention-based liquid recognition.

The solar-powered High-Altitude Platforms (HAP) can provide persistent communication and computing services beyond terrestrial coverage. However, changing mission priorities, coupled resource demands, and uncertain solar supply complicate sustainable scheduling, while large language model proposals require independent feasibility checks. This paper proposes SAGE, a Sustainability-certified Agentic Ground-service Engine for a HAP. We formulate an energy-causal stochastic optimization problem incorporating semantic modes, multi-user interference, channel aging, essential platform loads, and battery stress. An agent translates operator intent into typed admission, mode, and grouping candidates, while deterministic convex refinement and an action shield verify resource and reserve feasibility using calibrated solar lower bounds before execution. Agent proposals and deterministic fallbacks share the same certification path. Residual-based forecasting and memory-assisted replanning close the control loop. We establish fixed-candidate convexity, reserve safety under calibrated solar coverage and bounded actuation error, and an online utility bound accounting for candidate-generation error. At nominal load, simulations show 2.38 percentage points higher completion and 5.27% greater useful-workload energy efficiency than energy-harvesting Lyapunov control. Under stressed load and solar uncertainty, reserve violations occur in 1/50 runs, versus 47/50 without energy certification. A 600-record calibrated inference replay further shows 23.5--36.5% utility gains over deterministic-only candidates under compound stress.

Pretrained Vision-Language Models (PVLMs) achieve strong zero-shot recognition but often suffer performance degradation under target-domain shift. Unsupervised adaptation provides a practical way to improve target-domain performance without labeled data, but it remains challenging because pseudo-labels generated by the model can be unreliable and no labeled validation set is available for model selection or early stopping. We propose a simple framework for unsupervised PVLM adaptation that combines progressive margin-based subset selection with cascaded consistency learning. The subset selection strategy constructs a class-balanced pseudo-labeled set using prediction margins and gradually expands it during training, enabling adaptation to start from reliable supervision anchors before incorporating broader target-domain data. Cascaded consistency learning further exploits unlabeled samples by using temporally smoothed teacher predictions on weakly augmented views to supervise student predictions on strongly augmented views, with confidence filtering to reduce the influence of uncertain pseudo-labels. To preserve the pretrained embedding structure, the framework updates only normalization parameters in the visual encoder and class prototypes initialized from text prompts. Experiments across multiple recognition benchmarks with both CLIP and SigLIP backbones show consistent improvements over zero-shot baselines and competitive performance against recent unsupervised PVLM adaptation methods. These results suggest that progressive pseudo-label selection and teacher-guided consistency provide an effective and lightweight strategy for adapting PVLM classifiers without labeled target data.