Science & Mathematics

Existing high-speed navigation systems for micro aerial vehicles (MAVs) commonly rely on depth cameras, stereo vision, or LiDAR to obtain explicit geometric information, while monocular cameras offer a lighter and lower-cost alternative. However, monocular flight remains challenging because RGB images offer no direct metric cues and visual appearance varies considerably across environments. We propose a geometry-aware monocular visual planning framework that predicts trajectories from two image frames, their relative pose, and the MAV state. Without explicitly reconstructing depth, the proposed method complements the downstream planning objective with feature-level geometric supervision by transferring the planning-relevant representation learned by a privileged depth expert to the RGB planner through cross-modal distillation. The framework first trains a depth-based geometric expert, then distills its planning features into a monocular branch using paired RGB-D observations, and finally adapts the planner to the downstream trajectory planning objective. This design encourages the policy to learn planning-relevant geometric features, enabling collision avoidance. We further use a 3D Gaussian Splatting (3DGS) simulator to render visual observations for data collection, paired RGB-D supervision, and simulation evaluation. The planner trained in simulation is then deployed in real-world environments without real-world fine-tuning, achieving obstacle avoidance at speeds of up to 6.5 m/s.

Conversational AI systems built on Large Language Models (LLMs) require routing mechanisms that balance answer quality and response latency at scale. However, existing LLM routing benchmarks assume full observability of all candidate model outcomes, under-emphasize latency as a first-class metric, and fragment evaluation across heterogeneous protocols misaligned with real-world settings. We introduce RouterWorld, a comprehensive benchmark for LLM routing in conversational GenAI. RouterWorld unifies 371k single-turn responses and 76k multi-turn agentic trajectories, adopts natural user interactions with a standardized LLM-as-a-judge protocol, and decouples efficiency from transient pricing via hardware-grounded latency annotations. Crucially, RouterWorld formalizes routing under partial observability through two setups: Cold-Start, which restricts learning to sparse, static logs, and Voyager, which models continuous data streams where routing decisions shape future training distributions. Evaluating 13 routing baselines, we find that most routers over-emphasize accuracy while ignoring latency under partial observability. Code and benchmark data are publicly available.

Predicting protein functions from amino acid sequences is essential to understanding biological processes. Large language models (LLMs) offer flexible function descriptions to natural-language queries, yet their limited protein understanding motivates retrieval-augmented generation (RAG). Existing protein RAG methods primarily expand external retrieval signals while overlooking the rich knowledge encoded within a single protein language model (PLM). We show that retrieval performance varies across PLM layers and biological tasks, making conventional final-layer selection suboptimal. To this end, we propose QProtFuse, which uses queries to adaptively fuse multi-layer protein features and jointly learns protein and annotation retrieval through intra-modal and cross-modal contrastive objectives. On Prot-Inst-OOD, QProtFuse achieves the best protein and annotation retrieval scores, reaching 16.4\% Recall@1 with annotation reranking. These gains improve downstream RAG-based function descriptions, yielding the highest LLM-as-a-judge scores for both Qwen3-8B (6.22/10) and DeepSeek-V4-Pro (6.68/10). Ablations and weight analyses demonstrate the value of query-guided fusion and complementary training objectives for exploiting a PLM's internal knowledge.

Foundation models promise reusable capabilities and step-change performance gains across tasks. In finance, large-scale pretraining of domain-specialized foundation models can advance both goals. However, prevailing architectures and objectives do not by themselves address the combined demands of financial systems: partial observability, interacting entities, asynchronous events, changing regimes, action-dependent outcomes, and rare states with disproportionate consequences. Moreover, independently developed specialists may learn incompatible representations, while multi-domain monolithic alternatives couple data access, model updates, and governance. We therefore propose a structured-data-first research agenda centered on modular, domain-specialized financial foundation models connected by a representation fabric for reuse and composition across domains and tasks. These models should maintain persistent financial state across aggregation levels and time horizons, capture its dynamics, and simulate action-conditioned futures, while remaining calibrated across regimes and for tail outcomes. They must also learn efficiently under data and access constraints and provide faithful, traceable explanations. We identify research challenges, directions, and capability-oriented evaluation criteria.

The quality of smartphone portrait photography has become increasingly important to both manufacturers and users. Recent devices use complex computational pipelines to improve facial rendering under varied lighting. These pipelines vary widely across manufacturers, so the same portrait can look noticeably different across smartphones. These differences affect multiple aspects of facial appearance, making perceptual preference difficult to characterize with a single measure. To address this, we propose an interpretable framework that decomposes face contrast into three components: overall facial tonal separation, eye-region contrast, and micro-contrast of fine skin details. We formulate dedicated metrics for each component and validate them via pairwise human judgments, demonstrating that our approach correlates more closely with human contrast perception than traditional contrast measures. To support this analysis, we introduce a realistic mannequin-based dataset designed for systematic facial contrast evaluation, comprising 508 full-resolution captures from 29 smartphones under diverse lighting, with region-specific masks and extensive pairwise perceptual annotations. As a downstream analysis, we show that combining the three component-wise metrics provides a meaningful basis for modeling balanced facial contrast preference in a controlled setting, and we present a preliminary evaluation on a smaller real-portrait dataset indicating that insights derived from the mannequin study can extend to real images. Our work provides a controlled first step toward defining and evaluating face contrast in portrait imaging.

We demonstrate a multi-agent framework that takes an analog circuit from topology synthesis to layout. A researcher agent directs schematic and layout agents, a four-tier memory stores the specification and lessons, and separate agents review the designs. Deterministic tools provide pass/fail simulation reports, symbolic analysis, graph-based netlist equivalence, parasitic injection, and extraction with DRC and LVS in every optimization trial. A human approves the specification and the layout plan. For a fourth-order, 250-Hz subthreshold biopotential low-pass filter, the researcher agent synthesized, through simulation-guided trial and error, a novel current-reuse architecture that shares a single dc bias branch between a super-source-follower biquad and a flipped-voltage-follower biquad. The agents completed a DRC- and LVS-clean layout in 130-nm IHP SG13G2 CMOS that meets the specification in nominal post-layout simulation, despite the noise-power trade-off and threshold-voltage sensitivity of subthreshold operation. The filter's input-referred noise over 0.5-200 Hz is 41.6% below that of the super-source-follower reference design at about equal core power.

Large language models (LLMs) can generate quantum programs, but reliable evaluation remains difficult because benchmarks differ in task composition, scoring, and failure handling. We evaluate 15 hosted frontier, open-weight, quantum-specialized, and retrieval-augmented models on five quantum code-generation suites (530 tasks), plus 10 models on our new QuantumBenchEval (QBE) benchmark (96 tasks across six topics). We report bootstrap confidence intervals, multiplicity-adjusted paired comparisons, and audits of output validity and evaluator behavior. QiskitHumanEval-Standard approaches saturation (98.7% pass@1), while its Hard variant retains headroom (63.8%). None of the five top-two comparisons across primary suites remains significant after multiple-comparison correction. Evaluator design substantially changes scores: removing checks for unspecified quantities raises GPT-6 Astra's QBE pass@1 from 0% to 94.1% (T1) and from 1.2% to 92.9% (T6), without changing model outputs. We identify 17 QuanBench-117 tasks with shared failures linked to benchmark-owned imports. Controlled probes also reveal unexecuted unit tests, leakage of canonical functions into candidate execution, and reference imports that block otherwise valid candidates. These defects can produce both false failures and false passes, so historical QuanBench results require revalidation under an isolated evaluator. Overall, leaderboard scores alone do not adequately characterize progress in quantum code generation. We release our harness, per-sample results, and audit artifacts and recommend validating evaluators, isolating candidate code, and reporting uncertainty and output validity.

Large language models (LLMs) are increasingly used as user simulators, but can they faithfully reproduce an individual's behavior and decision logic over time? We investigate this question in a longitudinal paper-trading study with 80 participants. Using rolling next-day predictions, we evaluate trading activity, buy-sell structure, asset selection, and portfolio outcomes, complemented by controlled input ablations and participant-rated explanation alignment. No evaluated LLM reliably outperforms a simple recent-activity persistence baseline for predicting trade occurrence, while buy-sell structure and asset selection remain difficult to reproduce. Similar activity-level performance can also yield substantially different portfolio trajectories. Ablations show that activity predictions depend primarily on recent trades, whereas asset choices are sensitive to multiple evidence sources. Participants rate explanations as more aligned with their recurring strategies than with their decision-specific rationales. These findings reveal a gap between coarse behavioral agreement and individual-level fidelity, motivating separate evaluation of behavioral accuracy, explanation alignment, and evidence dependence in personalized user simulation.

Consolidating specialized foundation models into a unified generalist is a central goal of post-training. Two approaches are in common use: model merging, which is fast but treated as a heuristic, and multi-teacher on-policy distillation, which matches distributions but demands heavy rollout compute and suffers from cross-task gradient conflict. We show that the two are one objective seen at different orders. Within a shared basin of homologous specialists, the multi-teacher on-policy distillation objective is, to second order, exactly the Fisher-merging objective. Its minimizer is therefore the Fisher merge, a classic model-merging method given by a closed-form combination of the specialists, so merging is optimal, to second order, as a closed-form stage of distillation. Building on this identification, we propose Residual-Governed Branch-and-Merge(RGBM), a two-phase consolidation procedure. Concretely, a full-parameter Fisher merge is intractable on foundation models, so Phase~1 restricts the merge to the task-vector subspace. Extending tangent-space analyses of merging, we show that the same coefficients that mix task vectors also mix implicit rewards, which lets a residual program on unlabelled prompts solve for them. The program minimizes the worst per-domain residual and reports its energy-normalized values as the domain residual spectrum. Phase~2 lets this spectrum govern what may be added: RGBM passivates domains with low normalized residuals, trains isolated branches only for domains above the spectral gap, and writes back the admitted updates with calibrated shrinkage. On two open-source consolidation suites, the Phase~1 merge alone matches or outperforms fully trained multi-teacher distillation, and RGBM improves on it further with zero extra joint-training steps, whereas every un-gated write-back, including continued joint distillation from the merge, falls below it.

The rapid integration of power electronics converters has fundamentally altered the power grid's dynamic landscape, making modularity of operation and interoperability of heterogeneous devices two critical hurdles for power system reliability. In control language, these challenges can be paraphrased as decentralized certificates on the input/output behavior of every device, assured by local control, so that stability and guaranteed performance of the interconnected system emerge. For the power system application, such certificates should eventually be part of the rule book, the so-called grid code, and thus be generic and minimally restrictive for the control architecture of each device. Classical system-theoretic angles of attack to these questions go back to dissipativity concepts, which have historically been co-developed in automatic control, circuit theory, and power systems. Since standard dissipativity (e.g., passivity or small gain) criteria fall short in terms of conservatism, a range of more flexible and advanced tools have been developed to assure less conservative decentralized certificates guaranteeing stability under interconnection as well as dynamic performance following a disturbance. In this article, we review a range of methods that have been put forward to tackle these questions, compare and validate them, and point towards practically relevant research gaps which are amenable to control-theoretic approaches. Moreover, we provide a tutorial exposition of practically relevant power systems/electronics issues accessible to control engineers.

There are many types of path between nodes of a Directed acyclic graph (DAG), including the shortest and longest paths. Here, we ask whether the choice of path used in any analysis reflects the geometry of the latent space in which the DAG is embedded. To do this we define L_p-cube spacetimes, a family of continuous ordered spaces in which a single parameter p determines whether the ordered distance satisfies the ordered triangle identity when p>1, or the reverse ordered triangle identity when 0

Knowledge graphs (KGs) are increasingly used in autonomous driving (AD) to make traffic scene entities and relations explicit for prediction and reasoning. Existing traffic-scene KGs, however, primarily emphasize trajectory forecasting or general scene understanding. This paper presents the nuScenes Intention Knowledge Graph (NIKG), a focused Web Ontology Language (OWL)-based semantic representation and construction pipeline for ego-centric vehicle intention context. NIKG instantiates nuScenes scenes as a self-describing Turtle graph and promotes three intention-relevant relations to first-class properties of each scene participant: ego-relative motion, spatial configuration, and time to collision (TTC) risk. The pipeline uses a small input contract available in many ego-centric driving datasets: ego pose, tracked object annotations, and temporal keyframe order. NIKG is released as an ontology-backed, self-describing graph artifact together with a reproducible construction pipeline and validation protocol for ego-centric intention context. Applying this pipeline to nuScenes mini yields a graph of 211,335 triples across 404 keyframes; under the validation protocol, ego-relative motion labels achieve 92.5% forward consistency. Coverage, semantic consistency, and temporal coherence analyses over this artifact are further reported. NIKG occupies the space between scene-oriented traffic KGs and intention-prediction models: a small, explicit semantic layer intended to support downstream work such as knowledge graph embedding (KGE), SPARQL-based feature extraction, or hybrid neuro-symbolic reasoning.

Neuro-symbolic artificial intelligence (NeSy AI) combines neural perception with symbolic reasoning, enabling logical reasoning. Hyperdimensional computing (HDC) is promising for NeSy AI, owing to its high parallelism and strong robustness to noise. However, the dense neural operations and memory-intensive symbolic inference in NeSy impose a severe "memory wall" bottleneck. Existing NeSy accelerators either focus solely on symbolic kernels or rely on conventional digital architectures, lacking a unified and energy-efficient hardware substrate. Moreover, exploring optimal hardware design configurations often requires costly trial-and-error. In this work, we propose UnO-NeSy, a unified CiM framework for optimizing NeSy computing. UnO-NeSy consists of UnO-CiM, an ultra-low-power NeSy architecture based on a homogeneous 2FeFET--2R compute-in-memory (CiM) array that supports both neural computation and HDC reasoning, and UnO-Sim, a cross-workload, end-to-end CiM simulator that models device non-idealities, neural quantization effects, and HDC execution behavior for accurate system-level evaluation. Built on UnO-Sim, we further develop an automated optimizer to identify workload-specific operating points under coupled energy, accuracy, and hardware constraints. By optimizing the operating configuration of the baseline 2FeFET--2R design, UnO-NeSy achieves $61.8\%$ higher energy efficiency and $38.2\%$ lower full-system inference energy. Across seven RAVEN configurations, UnO-NeSy achieves $96.2\%$ average reasoning accuracy under hardware non-idealities, demonstrating a practical and scalable path toward ultra-low-power NeSy hardware.

Block diffusion drafters accelerate speculative decoding by generating multiple draft tokens in a single forward pass. However, each pass produces only a fixed-length block from a single target-generated anchor, limiting how far the draft can extend. Meanwhile, target verification computes hidden states even at positions beyond the first rejection, but these states are typically discarded rather than reused. We propose ReAnchor, a training-free method that reanchors a pretrained drafter at shifted positions using discarded verifier states to generate parallel draft blocks. These blocks improve predictions within the original block and extend coverage beyond its endpoint. At sampling temperature of T=0, ReAnchor achieves its largest average gains on Gemma 4-12B, improving acceptance length by 19.4% and decoding speed by 17.9% across eight benchmarks. Averaged equally over four target models and eight benchmarks, it improves acceptance length by 15.2% and decoding speed by 13.4% with pretrained DFlash drafters.

The metric distortion framework asks voting rules to elect a candidate close to the voters using only their rankings. SimultaneousPluralityVeto elects from the plurality veto core and achieves the optimal worst-case distortion of 3, but it is not Pareto-efficient: it can declare a tie between two candidates even when every voter prefers one of them. Removing Pareto-dominated candidates afterward would break the convexity of tied winner sets. We instead introduce regularization, which adds a small $\varepsilon$ to every candidate's weight, restoring Pareto-efficiency. To study its effect on distortion, we introduce the rank score veto core, whose candidate weights come from a positional scoring vector $\mathbf{s}$, and prove that its distortion lies in $[1 + 2(||\mathbf{s}|| - s_m),\, 1 + 2||\mathbf{s}||]$; we conjecture that the lower bound is tight. A fixed $\varepsilon$ can increase distortion, but letting $\varepsilon$ depend on the number of voters preserves the distortion guarantee of the original veto core. Regularized SimultaneousPluralityVeto thus has optimal distortion, is Pareto-efficient, and has convex tied winner sets, resolving an open question of Kızılkaya and Kempe.