”三合一并购”后中金公司首份三季报：净利破130亿，实现翻倍以上增速
华尔街见闻投行与自营挑大梁
Pratt parsing is a black box
Lobsters
On the Design Space of Discrete Diffusion Online Adaptation for Molecular Optimization
arXiv cs.LGarXiv:2607.02834v2 Announce Type: replace
Abstract: Molecular optimization often starts from a pretrained generative model that captures a broad prior over valid molecular structures. At test time, however, the goal is not to sample from this prior, but to use a limited oracle budget to shift generation toward task-specific high-reward molecules. We study this adaptation problem for discrete diffusion models. Each online round couples several choices. The loop must decide which candidates to evaluate, how rewards become model updates, which feedback to reuse, and how far to move beyond the pretrained prior. These choices have mostly been studied in isolation, leaving open whether they complement one another, become redundant, or interfere inside a full online adaptation loop. We conduct controlled studies across six small-molecule binding-affinity tasks and three protein-fitness tasks. We find that acquisition, reward shaping, and model debiasing provide complementary routes to higher reward, especially for small molecules. Replay further stabilizes learning, while validity penalties keep small-molecule exploration on the valid molecular manifold. Together, these findings point to a practical recipe for feedback-efficient molecular optimization: online fine-tuning with acquisition, reward shaping, debiasing, replay, and validity control. This recipe outperforms offline fine-tuning and inference-time search baselines under matched oracle-call budgets and GPU-hour accounting. The gains are largest when high-reward candidates require larger shifts from the pretrained prior.
Cyclical Entropy Eruption: Entropy Dynamics in Agent Reinforcement Learning
arXiv cs.LGarXiv:2605.27954v2 Announce Type: replace
Abstract: Agentic large language models are increasingly used to solve real-world tasks by reasoning over goals, invoking tools, and interacting with external environments. Reinforcement learning provides a natural framework for improving these behaviors, and recent agent RL methods have achieved strong results across domains. However, the training dynamics of agent RL remain poorly understood, limiting our ability to diagnose instabilities and design more effective training algorithms. In this work, we identify a previously underexplored phenomenon in agent RL, which we term cyclical entropy eruption. Unlike single-turn reasoning RL, where entropy typically collapses and stays low, agent RL training exhibits unique recurring cycles of sharp entropy eruption and gradual subsidence. We decompose this dynamic into three phases and provide theoretical and empirical analyses of each, explaining the mechanisms underlying its cyclical oscillation. We further show that degenerate patterns such as sentence duplication and hallucination, once acquired during eruption, can persist and accumulate across cycles. Motivated by these findings, we propose SEAL (Separation-Enhanced Agent Learning), a lightweight auxiliary loss that separates correct and incorrect trajectories in representation space, directly targeting the root cause of entropy eruption. Experiments across multiple benchmarks, models, and RL algorithms demonstrate that SEAL stabilizes training and yields stronger downstream agent performance.
INLA for Adaptive UAV Sampling of Radio Environment Maps in IoT Networks: Prior Width Governs Accuracy and Calibration
arXiv cs.LGarXiv:2610.40723v1 Announce Type: cross
Abstract: We study adaptive UAV sampling for radio environment map (REM) construction from non-Gaussian IoT observations. Integrated nested Laplace approximation(INLA) updates a Mat\'ern GMRF posterior, with waypoints selected by entropy pre-screening and Sherman--Morrison lookahead over the reachable set. We show that the likelihood-induced Skew-Normal correction has a sextic entropy effect, explaining its limited influence on waypoint selection. Across 19 retained ray-traced windows,Random and lawnmower sampling cost a median 18.0\% and 14.0\% more residual NMSE than posterior-informed sampling, equivalently reductions of 15.3\% and 12.3\% when referred to those baselines, while differences among informative rules remain unresolved. Widening the prior shadowing standard deviation from 6.0 to 10.025\,dB improves NMSE and credible-interval coverage, with monotone improvement over a 4--16\,dB sweep on this corpus and no observed turning point, although overconfidence persists. Online width estimation from the UAV's own counts yields about 6.5\,dB without recovering the gain. Across 30 measured LoRaWAN propagation windows paired with simulated counts and scored only on received cells, the margin over non-adaptive sampling reproduces, the training-derived width minimizes the tested interval score, and GP-UCB has nominally $3$--$5\%$ lower NMSE. The results show substantial, corpus-dependent sensitivity to prior specification, while fine differences among acquisition rules have limited or unresolved effects, the preferred width depends on the corpus, scored cells and model mismatch.
CAP: Learning to Optimize from Sparse Observations of Related Tasks
arXiv cs.LGarXiv:2610.41173v1 Announce Type: new
Abstract: In experiment-driven black-box optimization, each evaluation can consume materials, equipment time, and other resources, limiting the number of experiments available for a new task. Records from completed related optimization campaigns could help, but learning a sequential query policy remains challenging when task-specific experience consists of only a few sparse histories. We propose CAP (Candidate Acquisition Policy), which models shared structure and task-specific variation in fixed historical records to sample interactive practice functions. On these functions, CAP learns a feedback-conditioned query policy aimed at final optimization performance under a fixed evaluation budget. At deployment, its parameters remain fixed while each new observation informs the next query. Across five synthetic task families spanning 6 and 20 dimensions and two engineering simulation benchmarks, BOPTEST and SU2, CAP attains the lowest median final-budget value among the reported methods on four synthetic families and both engineering benchmarks. These results indicate that sparse records from completed related tasks can support adaptive query decisions on new tasks, making more effective use of limited, costly experimental evaluations.
Before We Adapt: The Zero-Shot Generalization Bottleneck in Continual World Models
arXiv cs.LGarXiv:2610.41041v1 Announce Type: new
Abstract: Continual reinforcement learning (CRL) requires agents to operate under environmental changes while reusing prior knowledge whenever possible. We argue that model-based reinforcement learning provides a promising framework for this setting because learned world models (WMs) can, in principle, support change detection, knowledge reuse, adaptation, and acquisition of new knowledge. However, such benefits depend heavily on whether the learned model can generalize beyond the contexts in which it was trained. In this work, we propose to formulate this problem with a contextual Markov decision process in which context is decomposed into behavioral and physical components. This view organizes continual WM learning into reuse, adaptation, and acquisition of new knowledge. As a first step, we study direct reuse when behavioral rules remain unchanged. Using Dyna-Q and TD-MPC in controlled FourRooms tasks, we find that zero-shot reuse can degrade substantially under physical-context changes and that this behavior depends strongly on the observation representation used by the WM. Additional ablation studies and latent-space diagnostics suggest that learned representations do not automatically provide the invariances needed for reuse. These results identify representation generalization as an important bottleneck for reliable world-model reuse in CRL.
A Lightweight Fiducial-Based Pipeline for 3D Hyperspectral Mapping of ex-vivo Lumpectomy Specimens
arXiv cs.CVarXiv:2606.14534v3 Announce Type: replace
Abstract: Hyperspectral Imaging (HSI) is a promising modality for intraoperative assessment of resection margins in Breast-Conserving Surgery (BCS), but its clinical translation requires aligning the inherently 2D spectral information onto the 3D shape of the excised tissue so that suspicious regions can be precisely localized for targeted follow-up. We present a fully automated, calibration-free pipeline that produces a 3D hyperspectral point cloud of an ex-vivo lumpectomy specimen from a set of consumer-camera RGB images and a single top-down HSI acquisition. The 3D geometry is reconstructed with a deep-learning Structure-from-Motion backbone, stabilized in a metric reference frame by a custom bundle adjustment that enforces consistency on the corners of four ArUco markers placed around the specimen. The HSI cube is then registered to the reconstruction without recovering the HSI camera pose: the markers, visible in both modalities, define 16 corner correspondences that drive a planar homography, and 3D coordinates are recovered by lookup on an orthographically rendered depth map. Evaluated on two ex-vivo lumpectomy specimens, the pipeline achieves a median 3D registration error below 1 mm and a 2D reprojection error below 0.02 mm, with a total per-specimen processing time under 4 minutes on accelerated hardware. These results support the feasibility of integrating HSI-guided spatial localization into intraoperative margin assessment workflows for breast-conserving surgery.
Bayesian PET Reconstruction with Learned Flow Matching Priors via Langevin Sampling
arXiv cs.CVarXiv:2610.41337v1 Announce Type: cross
Abstract: Image reconstruction for positron emission tomography (PET) is a challenging inverse problem due to its inherent ill-posedness and Poisson counting noise, particularly in low-dose acquisitions. In this work, we introduce a novel Bayesian PET reconstruction framework that leverages a flow-matching generative prior learned from high-quality PET images. By incorporating this prior, we formulate PET reconstruction as a Bayesian inference problem and solve it via latent-space posterior sampling using the unadjusted Langevin algorithm. We evaluate the proposed method through numerical experiments and benchmark it against classical reconstruction methods and state-of-the-art deep learning-based baselines. The proposed method achieves improved reconstruction performance while enabling posterior uncertainty quantification.
A Patch is More Than a Token
arXiv cs.CVarXiv:2610.41814v1 Announce Type: new
Abstract: Vision Transformers represent each image patch through a canonical token produced under one observation condition. That token is a powerful interface, but it is still one realized state of the patch. We introduce PMTT, which establishes two foundations: a patch representation object can extend from one canonical state to a canonical-centered Representation State Family, and identity-preserving variation opens Patch-Level Local Feature Space. We develop four parts of this representation program. Patch Information Acquisition and Dense Orbital Views expose lawful same-patch states. Visible-to-Hidden prediction tests whether observed states reveal structure beyond themselves. Intrinsic analysis identifies low-dimensional, energy-concentrated, patch-specific, predictive geometry. Representation Organization separates acquiring more information from organizing that information, with an analytic Manifold as the central structured representation. We realize PMTT on frozen DINOv3 S/B/L/H+ backbones without neural post-training of the representation. Across two independent 4,096-anchor cohorts, Manifold explains 70.8-79.6% of Hidden variation, compared with 55.0-72.3% for directly Observed states, and wins on more than 95% of anchors at every scale. Effective rank falls from about 2.5 to 2.0 while dominant energy concentration rises with backbone scale. Across six visual capabilities and four scales, Observed exceeds the canonical Center in 21 of 24 conditions, Manifold exceeds Center in 23, and Manifold exceeds Observed in 16. The strongest current consequence appears in physical correspondence, where MegaDepth satisfies Center < Observed < Manifold at every scale. These results show that pretrained visual backbones contain richer patch information than the standard single-token interface directly exposes. Code and experimental assets are available at https://github.com/zixiongwang-ai/PMTT.
Beyond Trial-and-Error: A Geospatial-Guided Active Reasoning Framework for Robust Aerial Vision-and-Language Navigation
arXiv cs.CVarXiv:2610.41446v1 Announce Type: new
Abstract: Aerial Vision-and-Language Navigation (VLN) requires UAVs to locate language-specified goals in large-scale urban environments, where objects often appear small and visually ambiguous from high-altitude viewpoints. Existing methods largely follow a reactive paradigm common in indoor VLN, repeatedly updating navigation states from local observations while underutilizing the geospatial structure that remains stable during navigation. As a result, visually similar distractors can cause a reactive goal search to repeatedly reinforce locally plausible goal hypotheses based on passively accumulated evidence, leading to unreliable localization. To address this, we propose GeoReasoner, a robust framework that shifts the paradigm of Aerial VLN from reactive trial-and-error to geospatial-guided active reasoning. Specifically, GeoReasoner first performs pre-search geospatial reasoning to derive region-level priors and constrain the goal hypothesis space. Building on these, it further conducts hypothesis-driven exploration to actively acquire discriminative evidence and refine competing goal hypotheses. Finally, consistency-aware goal verification adversarially assesses complementary evidence sources before stopping, hence reducing premature or erroneous goal confirmations. Experiments on the real-world CityNav benchmark demonstrate that GeoReasoner sets a new SOTA among zero-shot Aerial VLN methods while competing favorably with supervised approaches.
Parse Before You Execute: Efficient Spatial Reasoning with Executable Primitives
arXiv cs.CVarXiv:2610.41389v1 Announce Type: new
Abstract: Spatial reasoning in Vision-Language Models (VLMs) has increasingly benefited from tool-integrated pipelines that ground computation in explicit geometry. However, existing systems often compile queries into relatively open-ended artifacts, such as code, task constraints, tool plans, or gathered evidence, leaving what must be resolved before execution under-specified. This can lead to unnecessary tool acquisition and iterative refinement, while still producing plausible reasoning traces from incomplete spatial commitments. We identify this failure mode as under-committed query compilation. To address it, we propose PEPS, a training-free agentic paradigm Parsing Executable Primitives for Spatial reasoning under a typed FESM schema over Frame, Entity, State, and Metric. PEPS follows a parse-before-execute principle: the Parser commits each query to a compact set of executable requirements; the Executor selectively grounds the required primitives through fixed geometric APIs and deterministic computation; and the Verifier checks execution sufficiency, enabling early termination or targeted repair. Extensive experiments across challenging spatial reasoning benchmarks show that PEPS consistently improves reasoning accuracy while substantially reducing refinement rounds, API cost, and end-to-end latency over strong tool-integrated baselines. These results demonstrate that explicit requirement commitment can make tool-integrated spatial reasoning not only more reliable, but also more execution-efficient.
Shifting Focus by Reliability: Geometric Fusion Modulation for Robust RGB-T Detection
arXiv cs.CVarXiv:2610.41204v1 Announce Type: new
Abstract: RGB-thermal (RGB-T) object detectors combine visible and thermal imagery for robust detection, but signal corruption in either stream can substantially degrade accuracy. Restoration-based methods incur significant latency and do not consistently improve downstream accuracy, raising a more direct question: rather than restoring corrupted inputs, can we efficiently adapt the detector to their varying signal quality? Conventional PEFT offers an efficient route, but its input-independent corrections cannot track diverse test-time corruptions. Our controlled corruption sweeps show that corruption types have similar effects near clean conditions but diverge increasingly with severity, motivating corruption-aware, per-input adaptation. We therefore propose PG-FM (Prompt-Guided Feature Modulation), which adapts RGB-T object detectors to the estimated corruption type and severity of each modality. Under controlled RGB-T corruption simulations, PG-FM preserves clean-set performance, improves mAP50 by up to 20.8 points on FLIR and 19.9 points on KAIST, and transfers without retraining to unmodified, naturally acquired external RGB-T datasets.
Teaching Large Language Models When Not to Know: Learning Temporal Critique for Ex-Ante Reasoning
arXiv cs.AIarXiv:2605.14636v2 Announce Type: replace
Abstract: Large language models (LLMs) often fail to reason under temporal cutoffs: when prompted to answer from the standpoint of an earlier time, they exploit knowledge that became available only later. We study this failure through the lens of ex-ante reasoning, where a model must rely exclusively on information knowable before a cutoff. Through a systematic analysis of prompt-level interventions, we find that temporal leakage is highly sensitive to cutoff formulation and instruction placement: explicit cutoff statements outperform implicit historical framings, and prefix constraints reduce leakage more effectively than suffix constraints. These findings indicate that prompting can steer models into a temporal frame, but does not endow them with the ability to verify whether a response is temporally admissible. We further argue that supervised fine-tuning is insufficient, since ex-ante correctness is not an intrinsic property of an answer, but a relation between the answer and the cutoff. To address this gap, we propose TCFT, a Temporal Critique Fine-Tuning framework that trains models to acquire cutoff-aware temporal verification. Given a query, a cutoff, and a candidate response, TCFT teaches the model to identify post-cutoff leakage, explain temporal boundary violations, and judge temporal admissibility. Experiments with Qwen2.5-7B-Instruct and Qwen2.5-14B-Instruct show that TCFT consistently outperforms prompting and SFT baselines, reducing average leakage by 41.89 and 37.79 percentage points, respectively.
Implicit Assumptions About Tasks in Continual Learning
arXiv cs.AIarXiv:2610.41792v1 Announce Type: cross
Abstract: Continual learning is the study of how a model accumulates knowledge from data received over time, rather than all at once. In practice, this is typically achieved by dividing a dataset into tasks and presenting them sequentially. Explanations of the resulting measurements take into account factors such as the similarity between segments, the rate at which the data changes and the order of presentation. Each of these factors can be defined in terms of the underlying data, the partition imposed on it, or the learner that encounters it. These levels do not necessarily coincide, and it is usually left implicit which level a reported quantity refers to. We separate these and identify which properties acquire a value at each level. In controlled teacher--student experiments, we demonstrate that partitioning the same underlying trajectory at different levels of granularity, changing the order of presentation, and varying the similarity structure of the underlying data all result in different measured outcomes. We review representative benchmarks to determine which of these properties are stated. Where the underlying data cannot be inspected, as in foundation models built on undisclosed pre-training corpora, its properties are inherited by the model's initial state, and the training applied on top of it can be described.
Pinching Antennas for ISAC: Waveguide Architectures, Solutions, and Future Directions
arXiv cs.AIarXiv:2610.41380v1 Announce Type: cross
Abstract: Integrated sensing and communication (ISAC) is a core design principle for sixth-generation (6G) wireless systems, yet fixed antenna geometry limits its performance. Pinching antenna systems (PASS) guide electromagnetic waves along dielectric waveguides and radiate them through position-adaptive antennas, allowing communication links and sensing observations to benefit from spatial reconfiguration. This article reviews PASS-enabled ISAC through single-waveguide, multi-waveguide, and segmented-waveguide architectures, comparing their feed structures, spatial degrees of freedom, operating protocols, and uplink/downlink support. We organize solution pathways around physics-consistent modeling, joint spatial-signal optimization, dynamic channel acquisition, and stable real-time control. A segmented waveguide-enabled pinching-antenna array (SWAN) case study demonstrates joint segment selection, antenna po?sitioning, and transmit beamforming using segment-hysteresis reinforcement learning. The results illustrate how waveguide loss and per-segment aperture gain create communication-sensing tradeoffs. We conclude with research directions in joint transmit?receive operation, measurement-calibrated digital twins, and trustworthy control.
From Evaluation to Evolution: Benchmarking Image Editing and Evolving Image Models
arXiv cs.AIarXiv:2610.41210v1 Announce Type: cross
Abstract: Real-world image editing requires coordinating multiple visual transformations with content preservation and reference constraints, yet aggregate scores can readily obscure unmet local requirements. Effective evaluation must therefore identify local failures and translate them into priorities for subsequent model improvement. We introduce WeGenEditBench, comprising 2,740 evaluation instances and 148 task types across single-image, text-intensive, and multi-reference editing. Pre-generated checklists shared across models decompose each request into atomic requirements; item-level deductions and visual evidence are established first and then summarized into scores for edit execution, preservation, visual quality, and task-specific correctness. To support the experience-acquisition component of recursive self-improvement (RSI), we develop WeGenEvoStudio, a unified multi-agent harness for RSI-oriented learning that integrates model-conditioned failure diagnosis, retrieval-grounded data construction, supervised adaptation, and re-evaluation. Among 15 evaluated configurations, GPT-Image-2 leads single-image editing with an Overall score of 4.076, while Seedream-5.0-Pro leads text-intensive editing at 4.291. Evolution experiments across four generation and editing benchmarks demonstrate substantial targeted gains: on their report-selected Top-5 single-image targets, FLUX.2 [dev] improves Overall from 1.083 to 3.517, while Qwen-Image-Edit-2511 gains 13.2%. Together, our benchmark and harness provide a foundation for developing more reliable image models through interpretable evaluation and feedback-driven learning.
Sensorium: An Open-Source Multimodal Data Collection and Curation Library for Embodied Surgical AI
arXiv cs.AIarXiv:2610.41946v1 Announce Type: new
Abstract: Embodied surgical AI requires reliable multimodal data, yet surgical data collection remains costly, difficult to repeat, and strongly dependent on the hardware setup. We present Sensorium, an open-source Python library for recording, processing, annotating, and curating multimodal surgical data from configurable robot and sensor streams using the Robot Operating System (ROS) and ROS2 middleware. The library combines operator-facing data acquisition, configurable processing, semantic annotation, and episodic dataset generation. We evaluate Sensorium across 1516 recordings of phantom, ex vivo, and in vivo porcine experiments involving multiple surgical tasks and sensing configurations, and report preliminary usability feedback.
What Should LLMs Forget: Reinforcement Unlearning of Knowledge Associations
arXiv cs.AIarXiv:2610.41616v1 Announce Type: new
Abstract: Large language models (LLMs) acquire extensive factual knowledge during pre-training, which may include sensitive, private, or copyrighted information, motivating the development of unlearning methods that selectively remove specific knowledge while preserving general capabilities. However, existing approaches typically optimize entire responses without explicitly identifying the knowledge to be forgotten, resulting in coarse and uncontrolled forgetting. To address this, we revisit the forgetting target of LLM unlearning and introduce Knowledge Association (KA), which characterizes the association between a query context and its target knowledge and is instantiated by the corresponding conditional predictive distribution, as the fundamental target of forgetting. Based on KA, we propose Knowledge Association Reinforcement Unlearning (KARU), a unified reinforcement learning framework that suppresses forgotten knowledge associations while preserving retained ones. Specifically, KARU derives complementary mechanisms from the KA-based formulation: token-level reward construction for finer-grained forgetting supervision, distribution-level regularization for precise knowledge suppression, and proximity-aware retain sampling to mitigate collateral degradation. Experiments across diverse LLMs on representative benchmarks demonstrate that KARU consistently improves forgetting effectiveness over existing baselines while maintaining comparable retention performance and general utility.
Agentic World Modeling for Agent Test-time Scaling without Environment Replica
arXiv cs.AIarXiv:2610.40927v1 Announce Type: new
Abstract: Test-time scaling (TTS) improves large language model agents by sampling and selecting among multiple candidate trajectories, but common realizations assume that these trajectories can be executed in parallel. This assumption breaks in realistic deployments, where enterprise workspaces and cloud services are shared and persistent, costly or infeasible to replicate, and actions may introduce irreversible side effects or cross-trajectory interference. Existing generative world models avoid direct execution, but typically compress environment state into a single prompt, limiting their ability to reliably simulate transitions under partial observability. Instead, we introduce Agentic World Modeling (AWM), which reformulates next-observation prediction as a multi-turn evidence-acquisition task. For each hypothetical action, an agentic simulator adaptively queries the real environment through safe read-only interfaces, reasons over the returned evidence, and uses executable computation to construct the resulting observation without mutating shared state. Integrated with trajectory-level TTS, AWM enables parallel exploration over a single shared environment. Across $\tau$-bench, AppWorld, and OfficeBench, AWM outperforms a single-shot simulation baseline by 40.07% on average, and surpasses a trained world-model baseline by 12.36% on average. These results establish interactive, read-grounded simulation as a practical foundation for safe and scalable agentic test-time scaling in shared, non-resettable environments.
继续滚动加载更多…