Papers

Multimodal large language models (MLLMs) incur high inference latency from long visual token sequences. Existing pruning methods commonly use attention maps or output features to estimate token importance or redundancy. Several recent approaches also exploit representation changes, but when and how these changes reflect foreground saliency and semantic consistency remain insufficiently understood. We analyze visual token representation dynamics across encoder depth and uncover two findings. First, the relationship between token update magnitudes and foreground saliency is layer-dependent: large token updates concentrate on foreground regions in two depth intervals, separated by several sink-dominated layers at intermediate depths. Second, similarities between token update directions better distinguish same-class from different-class tokens than those between encoder output features. Building on these findings, we propose MSDG-Prune, a training-free method that uses update magnitudes and directions to preserve salient and diverse visual information. Specifically, we group tokens by update-direction similarity and use query-weighted saliency derived from update magnitudes across a chosen depth window for group-wise token pruning. Extensive experiments across four MLLMs demonstrate the effectiveness and generalizability of MSDG-Prune. On LLaVA-NeXT, it retains 91.9% of uncompressed performance on average with only 5.6% of visual tokens, while achieving a 7.8x prefilling speedup. Code is available at https://github.com/liweixuan-hitsz/MSDG-Prune.

Aerial manipulators extend robotic manipulation into 3D workspaces that are difficult for ground-based robots to access, creating new opportunities for general-purpose manipulation. However, extending Vision-Language-Action (VLA) models to aerial robots introduces distinct challenges due to the tight coupling between manipulation and flight, continuously changing observations, and safety-critical physical interactions. These challenges demand diverse training data and systematic policy evaluation, yet collecting demonstrations and evaluating policies directly on physical aerial platforms are costly, difficult to scale, and hard to repeat under controlled conditions. We present AeroManip-VLA, a scalable benchmark for aerial VLA data generation and policy evaluation. AeroManip-VLA provides a GPU-accelerated simulation framework with low-level payload-aware flight and manipulation control in massively parallel environments. Building on this framework, we combine reusable reinforcement learning policies with expert task rules to automatically generate demonstrations without human teleoperation across diverse objects, environments, and randomized initial conditions. The generated data include basic skills such as grasping and placing, as well as long-horizon tasks that require both navigation and manipulation. We further introduce automated event labeling and trajectory categorization to filter demonstrations. These mechanisms enable fine-grained analysis of task progress, behavioral outcomes, and safety-related failures. Finally, we evaluate a range of imitation learning and VLA baselines across different task settings, revealing their performance characteristics and failure modes. Together, AeroManip-VLA enables scalable aerial manipulation data generation, structured trajectory analysis, and systematic VLA evaluation in simulation prior to real-world deployment.

Post-training has been shown to significantly improve language models' performance on tasks with verifiable outcomes, including mathematical reasoning, software engineering, and computer use. However, whether the same approach can improve forecasting in financial markets is much less clear. Compared with tasks with verifiable outcomes, not only are realized returns noisy, but even what constitutes a relevant information set for making effective predictions is not obvious a priori: the model must decide which observations to gather and then commit to a numerical judgment before the outcome is known. We study this question in a chronological stock-price sandbox, where a language model gathers price, volume, relative-performance, and market-context evidence and predicts a future return. We post-train Qwen3-4B with supervised fine-tuning (SFT) on tool-use demonstrations, then proximal policy optimization (PPO) with a terminal reward given by the forecast score against the realized return. The resulting AURA-4B more than doubles the starting direction–magnitude score, from 20.94 to 43.31, and is comparable to frontier language models on this benchmark. Conditional magnitude agreement rises from 33.3 to 66.2, while directional accuracy changes from 62.9 to 65.4. SFT expands tool use, and PPO further increases the share of ranking and market-context queries. These results show that post-training can substantially improve financial forecasting performance, together with changes in how the model investigates the market, on this outcome-selected benchmark.

Transcribing domain-specific entities and rare proper nouns remains a major challenge in automatic speech recognition (ASR). In this paper, we propose BaLEEN (Biasing with Latent Encoded Entities), a lightweight, hypernetwork-based framework for dynamic contextual adaptation without fine-tuning the underlying ASR model. BaLEEN encodes variable-length contextual keywords using a pretrained language model, compresses them into a fixed sequence of latent vectors via a Perceiver bottleneck, and injects context-dependent bias vectors directly into the intermediate encoder representations of the ASR model. Because both the language model and the backbone ASR model remain entirely frozen during training, BaLEEN operates as a plug-and-play adapter that incurs zero computational overhead at inference time when context biases are precomputed. We evaluate our method on a CTC-based ASR model using a Wikipedia-derived corpus with annotated named entities and synthetic speech. Experimental results demonstrate that BaLEEN reduces keyword miss rate by 8.7% on the test set relative to the unbiased baseline while simultaneously improving overall word error rate by 21% and character error rate by 28%.

Solomonoff's universal inductive inference [Inf. Control. 1964] is one of the most general frameworks for learning from sequential data and predicting future observations. It provides prediction guarantees for any computable stochastic source. We ask whether an analogous universal theory of induction can be developed for quantum systems. To this end, we introduce universal quantum inductive inference, a framework for learning and predicting quantum sources that may exhibit arbitrary correlations and entanglement across time. Given the outcomes of past measurements and the corresponding post-measurement quantum systems, the learner produces a joint state that predicts the next measurement outcome and post-measurement system while preserving correlations with the past systems. We model a quantum source as a multipartite quantum state whose description is generated by an unknown $s$-bit program within a given time bound. The learner is required to produce a state that is $\epsilon$-close in trace distance to the target with probability at least $1-\delta$. We establish an information-theoretic inference algorithm with round complexity $O(s\epsilon^{-2}\delta^{-1})$. We further prove a matching lower bound in the relevant parameter regime, even for classical sources. As a second result, we construct a non-i.i.d. state tomography algorithm that outputs a classical description of the conditional state of the next system given past measurement outcomes. Our information-theoretic tomography algorithm achieves round complexity $O(sn^2\epsilon^{-4}\delta^{-2})$, where $n$ is the number of qubits received in each round. Finally, we investigate the computational hardness of universal quantum inductive inference under quantum cryptographic assumptions.

In this thesis I develop methods for statistical inference when the distributions arising from complex biological systems are multi-modal, geometrically structured, and sometimes only defined up to a normalizing constant. I start from variational inference and, when analytic update equations are unavailable, move to black-box variational inference. To build intuition regarding inference challenges and the proposed methodologies, I introduce a novel unnormalized target density (the CoLN distribution) and reuse it as a controlled test case in the kappa. I then trace a trajectory of increasingly expressive approximations: ensembles evaluated with the multiple importance sampling ELBO (Paper A) and variational mixtures that automate component cooperation and exploration (Paper B). Because expressivity comes at a cost, I develop efficient mixture learning ideas, including Monte Carlo objective estimators to scale mixture learning more efficiently (Paper C). As a new result in the kappa, I overturn a three decades long misconception regarding the potential performance benefits of using mixtures in variational inference. Finally, I move from variational inference to flow matching, where I address the need for specialized treatment of interpolant learning in multi-marginal settings (Paper D). By combining insights from Papers A-D, I derive in Section 5.5 a new method: multi-marginal flow matching with mixtures of variational interpolants. I connect these methodological developments to biological applications, with special emphasis on three-dimensional spatial transcriptomics, where stacked tissue slices induce multi-modal dynamics across space.

We first study the part size statistic in cylindric partitions, listing some specializations of a crude but general formula. Then, we establish bijections between ordinary partitions and cylindric partitions of small profiles with additional restrictions. Lastly, we extend earlier formulas for generating functions of cylindric partitions into distinct parts and functional equations of unrestricted cylindric partition generating functions, involving the part size statistic.

O'Grady's generalized Franchetta conjecture asks whether any codimension two cycle on the universal polarized K3 surfaces restricts to a multiple of the Beauville–Voisin class on a given K3 surface. We apply Bergeron–Li's result on the cohomology of universal K3 surfaces to give an affirmative answer to this conjecture when the universal K3 surface is rationally connected, including the case of genus 22.

Grover's algorithm achieves a quadratic speedup for unstructured search when a global oracle for the target set is available. However, its success probability is generally below $100\%$. To overcome this issue, Long presented a modified version of Grover's search algorithm by introducing a phase-matching condition, which can search for the target set with zero theoretical failure rate. Nevertheless, in many applications, the target set is determined by the intersection of multiple constrained sets. Most existing Grover-based methods are limited to the intersection of only two sets and are not always successful in searching for the target set. In this work, based on Long's method, we propose an algorithm for exactly searching the intersection of multiple sets. We derive and prove a phase-matching condition for searching the intersection exactly. Furthermore, we present a quantum circuit that implements exact intersection search. Finally, we support our theoretical results with numerical simulations.

Quantum steering captures the ability of one party, through local measurements on a shared entangled state, to affect the conditional states held by distant parties in a way that admits no local explanation. Its device-independent (DI) quantification – requiring no characterization of any state or measurement – has been developed in the bipartite setting, notably through the framework of assemblage moment matrices (AMMs) [Phys. Rev. Lett. 116, 240401 (2016)], but has remained largely unexplored beyond it. Here, we extend the AMM framework to tripartite steering, covering both the scenario in which two parties jointly steer a third one (2-steer-1) and that in which a single party steers the remaining two (1-steer-2). For each scheme, we construct a semidefinite program that, from the observed correlations – or merely from the violation of a Bell inequality – yields lower bounds on the steering robustness of the underlying assemblage. As a by-product, our bounds also certify, in a DI manner, the incompatibility of the measurements performed by the steering parties.

Reliable embodied decisions under partial observability require informative observations and sufficient supporting evidence. However, semantic scores alone do not reveal which viewpoints justify a claim or where additional evidence should be acquired. We introduce SafeVantage, a vantage-aware semantic memory and active acquisition framework that retains each claim's supporting views, camera poses, and estimated target location, keeping positive support distinct from search coverage. A learned candidate-observability model uses claim-grounded geometry to predict target visibility at reachable viewpoints. These predictions guide view selection through expected reduction in terminal decision loss, accounting for travel cost and geometrically distinct corroboration. A calibrated head then combines support, spatial consistency, and coverage to produce Yes, No, or Abstain decisions. We evaluate SafeVantage on a category-presence benchmark spanning 232 unseen ProcTHOR houses and 7,424 paired episodes per method and action budget. Compared with validation-selected equal-budget baselines, SafeVantage achieves macro-F1 gains of 24.7% and 12.0% at eight and twelve actions, respectively, with lower risk and higher answer rates at both budgets and 31.7% less travel at eight actions. Equal-input HM3D experiments show lower selective risk under fixed observations, while controlled ScanNet interventions show that restoring supporting views improves downstream VLM answers. Ablations further support the contribution of candidate observability to decision quality and acquisition efficiency. Results demonstrate the value of claim-level viewpoint evidence for connecting semantic memory, active acquisition, and reliable decision-making. Code is available at https://safevantage.github.io

This paper proposes a ternary visible light communication method using an event-based vision sensor (EVS) and a liquid crystal display (LCD). The throughput of optical camera communication (OCC) is limited by the frame rates of conventional frame-based cameras, and EVSs are expected to overcome this limitation because their pixels asynchronously trigger events with polarity in response to brightness changes, with a temporal resolution on the order of microseconds. However, existing event-based OCC relies on binary signaling that uses only the presence or absence of events and leaves the polarity unused. The proposed method maps the three brightness transitions (increase, decrease, and no change) to ternary symbols, thereby increasing the information carried per symbol. At the transmitter, the LCD displays a marker in which the transmitted data are encoded; at the receiver, the EVS captures the marker and records the brightness changes as events with positive and negative polarities. To suppress errors caused by the timing mismatch between display updates at the transmitter and demodulation at the receiver, we also propose a demodulation algorithm based on frame-transition detection. Experimental results demonstrate that the proposed method achieves a throughput of 336.5 kbps, approximately 150% higher than that of its binary counterpart.

Let $d_1,d_2,\ldots, d_r\ge 2$ be $m$-th power free positive integers. In this paper, we provide explicit representations of all irreducible $\mathbb{Z}_+$-modules over $\mathbb{Z}[\sqrt[m]{d_1},\sqrt[m]{d_2},\ldots,\sqrt[m]{d_r}]$, where $\mathbb{Z}_+$ denotes the semiring of non-negative integers.

End-to-end full-duplex speech models have brought open-source machine conversation closer to human-like interaction, yet existing systems remain limited in two intertwined dimensions: long-context robustness and multi-party interaction. Real-world scenarios such as meetings, group lessons, and social-robot reception require a single model to track, contextualize, and respond to multiple speakers over extended durations. Progress is constrained by both data and evaluation: open multi-party speech corpora remain small and are not designed for codec-frame-level full-duplex modeling, while existing long-audio benchmarks focus on passive listening and speech-to-speech benchmarks are mostly short and dyadic. We extend the Moshi paradigm jointly along the long-horizon and multi-party axes in English and Chinese. First, we release 57.6k hours of synthetic training data ($\href{https://huggingface.co/datasets/MultiTalk/MultiTalkPT}{MultiTalkPT}$ and $\href{https://huggingface.co/datasets/MultiTalk/MultiTalkFT}{MultiTalkFT}$) for long-form, multi-party, English-Chinese full-duplex dialogue, with controllable length, participant count, turn-taking, overlap, backchannels, interruptions, addressee shifts, and long-range coreference. Second, we introduce $\href{https://huggingface.co/datasets/MultiTalk/MultiTalkBench}{MultiTalkBench}$, built from real human recordings, for evaluating long-form, multi-party, bilingual full-duplex dialogue. Conversations average 32.6 minutes and include probes for long-range entity tracking, topic coherence, and addressee selection. Third, we train a bilingual Moshi-style model that sustains coherent multi-party English-Chinese conversations over extended durations and substantially outperforms open-source baselines including Moshi, MiniCPM-o-4.5, and Qwen3-Omni-30B-A3B-Instruct on MultiTalkBench.

Existing multimodal fake news detection methods often introduce external information to assist detection. However, most of them rely on entity-level retrieval and are therefore prone to introducing event-irrelevant noise. Meanwhile, existing methods mainly focus on improving overall performance and do not account for differences in the degree of harm posed by different instances of fake news. To address these limitations, we design an Event-Level Evidence Retrieval Framework (ELERF) and propose a Relation-Aware Evidence Graph Network (RAEGNet). ELERF retrieves external evidence based on the complete event semantics of a news item. RAEGNet constructs a directed graph that incorporates news-evidence stance relations and evidence-evidence interaction relations, and introduces a conditional-harm branch to jointly model authenticity and potential harm. Experimental results demonstrate that RAEGNet outperforms multiple baseline methods across all evaluated metrics on Weibo-21, Fakeddit, and our self-constructed SSS dataset.

Modeling the response of driven many-body quantum systems from input–output data is difficult: the dynamics are nonlinear, history dependent, and expensive to simulate as system size grows. A paradigmatic case is High-Harmonic Generation~(HHG), where a strong field drives a medium to emit radiation that is highly sensitive to the drive and encodes long-range temporal correlations. We introduce a dissipative quantum reservoir computing~(DQRC) framework that builds a digital twin of such a system, learning its input–output map directly from data while the reservoir—itself a small open quantum system—stays fixed and only a classical readout is trained. We show that a minimal single-qubit reservoir reproduces the HHG response of a substantially larger Ising spin chain, and on a representative benchmark matches and on several metrics surpasses previously reported temporal convolutional and Kolmogorov–Arnold-network models, while using a simpler, physically realizable system. A single fixed reservoir further generalizes across a broad range of drives, indicating that it learns a shared physical response structure rather than memorizing trajectories. These results establish dissipative quantum reservoirs as compact, physically grounded digital twins for nonlinear, memory-dependent quantum dynamics. Code is available at \href{https://github.com/AI-and-Quantum-Computing/DQuRC}{https://github.com/AI-and-Quantum-Computing/DQuRC}.

Reward hacking occurs when policy optimization exploits a brittle reward interface or an overly permissive proxy objective, improving the training score without improving the underlying response quality. This phenomenon is amplified in group-relative policy optimization: an unsupported reward can shift the group baseline and alter the updates of other rollouts, while post-hoc or purely relative weighting cannot represent group-wide uncertainty. We propose Self-Tuned Anchored Reliability Group-Relative Policy Optimization (STAR-GRPO), a reliability-first advantage estimator based on paired assessments of the same rollout. STAR separates the quality signal from its learning influence: score disagreement determines rollout reliability, relative reliability enters a self-tuned robust location–scale fit before group normalization, and absolute group reliability attenuates the resulting bounded advantage. The analysis establishes coordinate and second-moment bounds, characterizes exact centering through the weighted location equation, and gives reliability-dependent attenuation guarantees for outlying rewards. We evaluate STAR-GRPO in two complementary reward-hacking regimes. In token-interface exploitation, STAR prevents runaway optimization of the deployed-interface score while improving the canonical quality signal. In rubric-proxy overoptimization for medical reasoning, STAR improves independent semantic evaluation, narrows the proxy–judge discrepancy, and reduces overclaim while optimizing the same task proxy. Together, these results show that reliability-first normalization offers a principled way to limit unsupported reward influence on both group baselines and policy updates, while retaining the task reward as the optimization target.

Reinforcement learning (RL) post-training increasingly relies on long-horizon, multi-turn rollouts. As post-training jobs outgrow a single cluster, rollout pools assembled across clusters introduce hardware heterogeneity. Rollout scheduling must serve two stakeholders: the hardware needs high aggregate decode throughput, while each trajectory needs to finish quickly. The tension arises from the memory-bandwidth-bound nature of autoregressive decoding. A large active batch amortizes weight reads for high throughput but leaves each trajectory a smaller bandwidth share and a longer completion time. The scheduling objective is therefore specialization, letting different workers serve different roles. Heterogeneous hardware further enables this specialization. High-bandwidth accelerators favor long-context work, while cost-efficient accelerators sustain large batches. Workload evolution makes this specialization difficult to sustain, and dynamic reassignment faces a circular dependency because a move's benefit depends on subsequent placement decisions. We present CadenceRL, which bypasses this dependency through structural workload reshaping rather than per-move benefit estimation. Pacing replaces long-context trajectories with shorter ones, providing a structurally positive transformation that sustains large active batches for high throughput. When accumulated staleness demands faster completion, concentration directs the residual long-context tail onto high-affinity workers. Late-bound KV preparation stages accumulated prefixes before a destination is selected. On heterogeneous rollout pools, CadenceRL improves decode throughput by up to 48% and reduces P95 trajectory latency by up to 64%. Adding high-bandwidth accelerators reduces tail latency, while adding cost-efficient accelerators increases throughput, without manual routing configuration.

Modern imaging produces terabyte-scale segmentation volumes, assigning each voxel an object label. These categorical, boundary-sensitive and label-rich data underpin connectomics and other imaging-driven fields, yet their scale often forces interpretation through slices, approximate meshes or distributed workflows that obscure spatial context and voxel-level defects. Here we show that such volumes can be explored directly on commodity hardware with Volcanite, an open-source framework for dense-segmentation rendering. Combining compression-aware data handling, a Vulkan GPU backend and segmentation-specific rendering, Volcanite enables low-latency exploration without meshing or distributed infrastructure, including for unpublished, sensitive or proprietary data. Across multi-domain datasets, it preserves voxel labels, adds shadows and global illumination, and renders up to half a trillion voxels at 100 frames per second. By replacing lengthy preprocessing with direct inspection, Volcanite decouples hypothesis generation from data preparation and turns teravoxel label fields into interactive evidence for discovery, validation and cross-domain spatial analysis

We study reduced solutions of Toda systems generated by spherical metrics for an arbitrary complex simple Lie algebra. The main result is an intrinsic centralizer parametrization: for a spherical metric, the reduced family is the positive Iwasawa slice cut out by the centralizer of the principal image of closure of monodromy group. As an application, we transfer existing spherical-metric existence results to the diagonal Toda source tuple.

Recent advances in generative planning have made trajectory inpainting a promising approach to offline goal-conditioned reinforcement learning. However, these methods typically specify the planning horizon before generating plan content, even though the appropriate horizon depends on the route itself. A horizon that is too short can force infeasible transitions, whereas one that is too long can introduce redundant motion. We introduce HorizonFlow, a hierarchical planner that treats plan length as an output of generation rather than a prescribed input. Its subgoal route planner guides its action-prefix controller through a sequence of latent subgoals. Both components combine insertion-based generation with flow matching to jointly generate continuous plan content and length, using the partially generated plan to guide token insertion. HorizonFlow reuses the resulting length information to select candidates and steer generation toward shorter plans without a separate learned value model. Across Maze2D, Multi2D, and OGBench navigation and visual manipulation benchmarks, HorizonFlow achieves the highest average performance among the compared methods.

Let $G$ be a graph of order $n$ with adjacency matrix $A_G$, let $\mathbf e$ denote the all-one vector, and let$W_G=[\mathbf e,A_G\mathbf e,\ldots,A_G^{n-1}\mathbf e]$ be its walk matrix. We consider the case $\operatorname{rank}W_G=n-2$, the first corank for which distinct graphs can have the same walk matrix. We give a complete structural description of such pairs. More precisely, if $G$ and $H$ are distinct graphs on the same labelled vertex set and $\operatorname{rank}W_G=n-2$, then $W_G=W_H$ if and only if $H$ is obtained from $G$ by a reciprocal Wang–Qiu–Hu (WQH) switching. In this case, $A_G-A_H=uv^T+vu^T$, where $u,v\in\{0,\pm1\}^n$ have disjoint supports and form a basis of $\ker W_G^T$. We also determine the minimum order at which a non-isomorphic corank-two walk mate can occur: no such pair exists for $n\le9$, so the previously known $10$-vertex example is sharp. Starting from a labelled realization of that pair, we use singleton union and join operations, together with the graph coronal, to construct connected non-isomorphic pairs with equal corank-two walk matrices for every $n\ge10$.

As a fundamental image processing task, object Re-Identification (ReID) aims to retrieve objects across non-overlapping cameras. Recently, with the development of deep learning, significant advancements have been made in object ReID. However, most existing methods suffer from generalization due to the limited size and diversity of ReID datasets. Meanwhile, current models tend to focus on extracting semantic patterns rather than learning identity-aware feature distributions. To address these issues, we propose a novel feature learning framework named DiffReID for object ReID. It leverages a discriminative diffusion model to gradually learn identity-aware distributions and generate identity-invariant features. More specifically, with the Contrastive Language-Image Pre-training (CLIP) model, we first obtain identity-aware text features by prompt tuning. Then, we propose a Vision-guided Noise Generator (VNG) to initialize probabilistic noises and gradually corrupt identity-aware text features. Afterwards, we take visual features as conditions and propose a Light Weight Denoiser (LWD) to denoise the corrupted text features step-by-step for identity-aware distribution learning. To obtain discriminative features, we further generate identity-invariant guided features from randomly sampling visual-guided noises. Finally, we propose a Mutual Enhancement Constraint (MEC) to facilitate mutual learning between visual features and guided features to enhance the representation robustness and discrimination. Extensive experiments on five object ReID benchmarks demonstrate that our method shows better results than most state-of-the-art methods. The source code is available at https://github.com/AWangYQ/DiffReID.

Detailed image captioning requires accurate and comprehensive descriptions of fine-grained visual content, yet caption quality spans factual accuracy, information coverage, and clarity. Compared with conventional methods that rely mainly on high-quality supervision or holistic rewards, rubric-based reinforcement learning decomposes these requirements into explicit criteria and provides targeted, structured feedback. However, existing methods often use separate models for caption generation, rubric construction, and judging, which may lead to inconsistent interpretations across roles. Some dynamic rubric methods alternate updates between the caption policy and rubric generator while keeping the judge fixed, but staged optimization may still leave rubric construction and judging out of step with policy optimization. We propose MoCo Rubric, a two-stage framework that coordinates these roles. First, role-conditioned, shared-parameter multi-task supervised fine-tuning equips a single vision–language model to serve as the Caption Policy, Rubric Generator, and Rubric Judge. Then, the Generator constructs rubrics online from captions sampled by the current Policy, reference captions, and image evidence. The Judge provides rubric-based rewards, and only the Policy receives GRPO updates. As Policy updates change the candidates being evaluated, we use an exponential moving average of the Policy parameters to update one momentum model shared by the Generator and Judge. This gradual transfer lets both rubric roles track Policy updates without separate RL optimization while smoothing parameter changes that could disrupt their rubric capabilities under direct synchronization. Across five captioning benchmarks, MoCo Rubric achieves an average pairwise win rate of 72.83%, the best mean rank in blind ranking, and the highest average score in caption-based question answering.

As the capabilities of large language models (LLMs) continue to advance, increasing attention is turning to how to translate their abilities into useful behavior. Personal agents bring this question into everyday settings, where models are expected to serve individual users and continually adapt to their preferences. With the underlying model held fixed, such adaptation relies on harness engineering: designing and evolving the surrounding layer that manages context, memory, tools, and execution. Despite rapid progress, the factors governing effective harness evolution remain insufficiently understood. To narrow this gap, we investigate three central questions concerning harness architecture, harness scale, and self-evolution algorithms through complementary empirical and theoretical analyses. Empirically, we introduce a preference-oriented benchmark and systematically characterize the capabilities and limitations of personal agents associated with these three dimensions. Theoretically, we formulate harness evolution as a learning problem and explain these phenomena through approximation, generalization, and optimization errors. Analyses of reachable policies, capacity under finite interaction evidence, and biased update dynamics provide theoretical accounts of the observed phenomena. Together, these results offer a unified perspective on the limits of personalization through harness evolution and inform future harness design.

Panoptic segmentation in forest environments is bottlenecked not by semantic quality but by instance separation; existing unsupervised panoptic approaches produce usable stuff maps but near-zero thing quality. Depth or flow-based instance discovery methods needs sensors that are not always available. We present ProGuT (Prototype Guided Training), which produces panoptic pseudo-labels without per-image training masks, needing only unlabeled images and one-time cluster-to-class mapping. ProGuT clusters CLIP patch features, then recovers trunk instances through multiscale geometric prior that falsifies non-trunk structures via structure-tensor. This is cheap compared to depth, flow or class-supervision methods to create pseudo labels. These are then used for downstream tasks which we evaluate against other unsupervised baselines. ProGuT achieves a Panoptic Quality (PQ) of 65.2 on Our-forest dataset (2.6x improvement over the initial pseudo-label quality) and reaches 65.9 mIoU on Freiburg Forest, outperforming unsupervised baselines like PiCIE (45.3 IoU) and STEGO(57.6IoU). Additionally, ProGuT outperforms existing unsupervised methods for class-agnostic trunk instance benchmark.

In scientific machine learning, physical fields governed by partial differential equations exhibit low-rank structure and scale invariance. When solving equations on coarse grids, missing information leads to the closure problem: modeling unresolved physics to recover lost dynamics. Although closure terms depend on grid resolution, they represent scale-invariant physical laws. A model truly learning physics should capture these mechanisms with low-rank parameterization rather than memorizing grid-specific patterns. Inspired by this, we propose the Scale-Invariant Neural Operator (SINO), which learns on normalized physical scales via a dual-branch architecture operating in spectral and spatial domains. SINO uses bottleneck MLPs to generate continuous convolution kernels, embedding an explicit low-rank inductive bias that concentrates more than 95 percent of variance in 2-3 modes, as validated by PCA across benchmarks, while drastically reducing parameters. This principled design yields 38 times steeper scaling law exponents than FNO, demonstrating superior parameter efficiency. We compare SINO with traditional models (U-Net, DeepONet), Transformer models (Transolver, Oformer, GK-Transformer), and frequency-domain models (FNO, AMFNO, UFNO) on closure problems spanning externally forced Burgers turbulence, decaying Burgers turbulence, KS turbulence, Kolmogorov-forced NS turbulence, and decaying NS turbulence. Experiments show SINO achieves 1.5-38 times error reduction and 2-23 times parameter efficiency over baselines, with superior scaling laws reflecting exceptional data efficiency from principled low-rank design. Code is available at https://github.com/AI4Science-WestlakeU/SINO.

To advance the development of embodied intelligence, Open-Environment Active 3D Mapping has attracted increasing attention, aiming to perform a long-horizon and shortest trajectory exploration for reconstructing unseen scenarios. Since only limited information about unseen environments is available, methods built on the closed-set assumption, i.e., assuming that the test environments are similar to those seen during training, cannot generalize satisfactorily. In existing active mapping methods, long-horizon exploration is often guided by predicting a coarse long-range goal and then converting it into an executable path. However, this stage is usually formulated as single-point prediction. Under partial observability, the same local observation may correspond to multiple plausible exploration directions, making such deterministic prediction prone to brittle decisions and degraded performance in unseen scenarios. Our experiments further verify that this is a key factor underlying their weak generalization. To address this issue, we reformulate long-horizon target prediction as conditional multimodal anchor generation using Conditional Flow Matching. Instead of predicting a single goal, our method learns a conditional distribution over coarse exploration anchors from the current mapping state. These anchors are first converted into executable candidate paths through obstacle-aware planning. We then apply exploration-mode clustering to compress geometrically similar trajectories and reduce candidate redundancy. Finally, a hierarchical selection module selects the most promising mode and reranks paths within it to produce the final executable trajectory. Experiments show that our method improves generalization and reconstruction efficiency in open environments.

Mixed human-LLM documents require locating authorship transitions from detector scores whose reliability varies across text units. Existing weighted mean contrasts are vulnerable to extreme scores, while directly replacing means with robust centers obscures how a misplaced boundary changes the population objective. We propose Robust Weighted Profile-Loss Change Point Detection (RWCP), which combines capped reliability weights, Huber profile gains, and narrowest-over-threshold search in reliability coordinates. Our key analysis expresses the population gap between a true and a displaced split as a merge cost, avoiding a closed-form solution for the nonlinear center of a mixed segment. Under explicit curvature, spacing, and dependence conditions, core RWCP recovers the number of changes and localizes their boundaries; its quadratic-loss limit recovers squared weighted CUSUM. We also study RWCP-R, a separately evaluated decoder that shares source centers across nonadjacent passages. Across five retrospective cached-score benchmark families, core RWCP reduces family-macro WindowDiff by 17.6% relative to weighted change-point detection, and RWCP-R lowers it further. Boundary recovery improves most clearly for isolated changes, while both fixed configurations miss changes in collaborative and densely alternating text.

Recent efforts to scale tool-use post-training have largely centered on the synthesis of executable environments, which constitute only one component of a broader agentic interaction system comprising the environment, task, agent harness, and evaluator. Scaling environments in isolation, however, does not guarantee commensurate gains in model performance, because reliable learning signals depend on coherent interactions among all components of the agentic interaction system. To address this problem, we introduce WEFT (Whole-system Evolution For Tool-use Post-training), which couples scalable agentic interaction system construction, execution-driven self-evolution, and stable post-training. WEFT scales agentic interaction system construction across environment breadth, task complexity, and interaction diversity. Execution-driven self-evolution iteratively uses execution traces and state evidence to attribute failures and revise the responsible components, with fresh rollouts evaluating the changes and providing evidence for subsequent evolution rounds. For stable post-training at scale, WEFT addresses both optimization and execution reliability: prefix-preserving sampling retains verified progress and atomic-turn credit assignment localizes learning signals, while MegaMCP maintains isolated, recoverable state across concurrent rollouts over shared tool services. Extensive experiments across various models and benchmarks demonstrate the effectiveness of WEFT for tool-use post-training. WEFT-8B and WEFT-14B outperform all evaluated matched-size environment-scaling baselines on BFCL V4, $\tau^2$-Bench, and Claw-Eval. In particular, WEFT-14B improves over Agent-World-14B by 6.41, 2.23, and 12.27 percentage points. WEFT-35B-A3B further extends these gains to more challenging long-horizon workflow benchmarks, including Toolathlon-Verified and AutomationBench.