Search arXiv⌕ Search

arXiv subjects

Search papers

Search indexed arXiv papers on artificial intelligence, large language models, computer vision and robotics. Read source abstracts and follow links to arXiv.

At least 829 records · Page 46Linked to original sources
arXiv2026

Morphology-aware deep learning predicts colorectal cancer survival from whole-slide images

Background: Five-year survival prediction from hematoxylin-and-eosin whole-slide images could support risk stratification, but the task is weakly supervised, and the tissue morphologies carrying prognostic information remain incompletely characterized. Methods: We conducted a retrospective secondary analysis using TCGA-COAD/READ (n=562) for morphology-specific analysis, Alliance/ CALGB 89803 (n=424) for model development, and SurGen (n=218) for external evaluation. We propose MorphContrast, a morphology-aware contrastive multiple-instance learning framework that uses tissue labels as structural supervision for five-year survival prediction. MorphContrast combines gated-attention pooling with three auxiliary objectives: patch-level morphology labels define within-patient contrasts, outcome concordance or discordance defines cross-patient contrasts within matched morphologies, and diversity regularization discourages representational collapse. Morphology-matched patches from patients with discordant outcomes serve as hard negatives, encouraging prognostic discrimination within comparable histologic contexts. Results: Morphology-specific models show weak but distributed prognostic discrimination: adipose tissue achieves an area under the receiver operating characteristic curve (AUC) of 0.622, stroma achieves 0.604, and both perform comparably to tumor epithelium (0.601-0.605). On Alliance/CALGB 89803, the single model achieves an AUC of 0.642, while the post hoc ensemble reaches a pooled out-of-fold AUC of 0.70 and a risk-group hazard ratio of 3.79 (95% confidence interval: 2.55-5.63). On the external SurGen colon cancer cohort (n=218), the single model and ensemble achieve AUCs of 0.669 and 0.678, respectively. Conclusions: These findings support tissue morphology as structural supervision for outcome-aware representation learning in colorectal cancer prognosis.

cs.CV↗
arXiv2026

Stability Transfer for Autonomous Systems via Higher-Order Taylor-Truncated Lyapunov Certificates

We develop analytical sufficient conditions for certifying the local asymptotic stability of nonlinear autonomous systems based on Lyapunov functions derived from their finite-order Taylor polynomial approximations. This approach translates the stability analysis problem from the original nonlinear dynamics to a more computationally tractable polynomial domain. The derived conditions allow for the explicit calculation of a guaranteed inner approximation of the region of attraction ($\mathrm{ROA}$). We show how sum-of-squares (SOS) programming and grid search can be used to systematically search for polynomial Lyapunov functions that yield a large certified threshold radius and extract a certified invariant set inner approximation of the $\mathrm{ROA}$ of the nonpolynomial original system. The framework is then applied to a control tuning problem: finding reduced-effort Linear Quadratic Regulator (LQR) controllers that ensure that a given initial state of interest belongs to the certified $\mathrm{ROA}$ of the nonpolynomial closed loop system. The effectiveness of the overall methodology is demonstrated using a pendulum with nonlinear damping and a nonlinear control system.

eess.SY↗
arXiv2026

PLEDGE: Progressive Lossy Error-Bounded Compression of 3D Gaussian Splatting

Existing 3D Gaussian Splatting (3DGS) codecs report reconstruction quality after compression, and give users little control over the distortion of individual primitives. We present PLEDGE, a training-free, progressive, error-bounded codec that translates four user-declared tolerances into per-primitive quantization bounds: screen-space position, footprint deformation due to rotation and scale, view-dependent color, and opacity. The bounds apply throughout a declared camera domain, including previously unseen viewpoints. Position and opacity are bounded directly; footprint deformation is bounded under the rasterizer's affine projection model, and color is bounded over all viewing directions. PLEDGE organizes quantized residuals into a progressive bitstream with explicit error bounds for every valid prefix. Experiments across thirteen scenes verify all declared per-mechanism bounds and demonstrate competitive rate-distortion performance. The four tolerances incur substantially different rate costs, so users can spend rate where their application needs it.

cs.CV↗
arXiv2026

Constrained Grid-Forming Control for Photovoltaic Systems: A Control Barrier Function Approach

This paper leverages control barrier functions (CBFs) to systematically account for constraints of voltage source converters (VSCs) and photovoltaic (PV) resources in grid-forming (GFM) control. The proposed control combines (unconstrained) dual-port GFM control with control barrier functions (CBFs) to ensure reliable and safe operation of a GFM PV system under grid faults, sudden changes in irradiance, and load perturbations. First, CBFs are used to encode physical and operational limits of the PV system, namely current and voltage limits at its ac and dc terminals. Then, CBFs are used to restrict the dynamics of dual-port GFM control to within the VSC and PV limits during both transient and steady states. To this end, a cost function is defined that encodes trade-offs between forming the ac voltage phase angle/frequency and voltage magnitude under constraints. Then, real-time optimization is used to make the least modification, as measured by the cost function, to the (unconstrained) dual-port GFM control state to remain within the safe set encoded by the CBFs. Finally, the performance of the developed constrained GFM control is evaluated through EMT simulations in MATLAB/Simulink.

eess.SY↗
arXiv2026

Facts Don't Speak Louder than Words: The Behavioral Essence of Long-CoT Distillation

While long chain-of-thought (CoT) distillation significantly improves the reasoning abilities of language models, its underlying mechanism remains poorly understood. Current pipelines generally rely on rejection sampling to gather reasoning traces across diverse prompts, assuming that factual correctness and broad problem coverage are indispensable. Surprisingly, we find that training exclusively on flawed reasoning trajectories or a restricted prompt set yields nearly comparable performance. Driven by this counterintuitive finding, we propose the Behavioral Hypothesis: the essence of Long-CoT distillation is behavioral alignment with structured reasoning patterns, rather than factual knowledge memorization. We then validate this by demonstrating substantial reasoning gains even when models are fine-tuned exclusively on previously mastered problems-effectively isolating behavioral alignment from novel knowledge injection. Having established that reasoning patterns are the true bottleneck, we investigate how to learn them across varying levels of complexity optimally. We reveal a critical interplay: while the recent probability-based loss excels at learning simple patterns, the standard cross-entropy loss is essential for scaling to complex CoTs. We believe these insights demystify CoT distillation and provide a principled foundation for the future development of reasoning models. Code is available on https://github.com/yuyongcan/Long-CoT-Distillation.

cs.CL↗
arXiv2026

$3$-manifold groups are linear

We prove that the fundamental group of every compact graph $3$-manifold is linear over the complex numbers. As a corollary, all fundamental groups of compact $3$-manifolds are linear over the complex numbers.

math.GT↗
arXiv2026

Verifier-Harness Co-Evolution for Training-Free Self-Improvement on Open-Ended Tasks

Training-free self-improvement lets an agent use feedback to revise the response it is about to return and to update the harness state that shapes how it handles later tasks. On open-ended tasks that feedback rarely comes from reference answers or executable tests, so the agent has to rely on a model-based verifier. When the verifier is wrong, the current response can get worse, and misleading experience can persist across the task stream. A verifier that never changes also grows stale as the agent's responses improve and task demands shift. We introduce Unified Co-Evolution (UCoE), a training-free method that evolves an executable verifier program together with the agent's harness state while all model parameters stay fixed. UCoE synthesizes candidate verifier programs and tests them on response contrasts, pairs of responses that differ in one declared property and have a known preferred side. It selects the candidate whose discrimination on its weakest contrast family is highest, a quantity we call driving capacity. The selected verifier diagnoses one weakness in each draft, guides a single repair, decides which response is served, and determines whether the accepted repair may enter the harness state. As tasks complete, their diagnostics seed new candidate verifiers, which must beat the current verifier on the same contrasts before adoption. Across six open-ended benchmarks, UCoE averages 59.00% preference against matched direct generation with Qwen3-8B and 57.19% with Qwen3.8-27B, margins of 9.00 and 7.19 percentage points. On three verifiable benchmarks, the same method raises the mean from 36.06% for direct generation to 38.48%. Code is available at https://github.com/tianyi0216/Unified-Co-Evolution.

cs.AI↗
arXiv2026

OpenDepth: Repurposing Image Editing Priors for Robust Open-World Monocular Depth Estimation

Recently, the integration of generative priors has revolutionized Monocular Depth Estimation (MDE), achieving state-of-the-art (SOTA) performance with remarkably high data efficiency. However, although methods based on Text-to-Image (T2I) models excel in general scenes, their efficacy degrades sharply when confronted with open-world environmental perturbations, such as illumination variations and adverse weather. Since T2I priors are optimized for semantic synthesis rather than preserving pixel-level structure, they struggle to decouple geometry from appearance under severe visual shifts, easily triggering structural hallucinations. Therefore, our core insight is to model open-world depth estimation as consistent geometric inference across multi-style images, fully leveraging the powerful structure-preserving editing priors of Image-to-Image (I2I) models. Based on this, we propose OpenDepth, a robust open-world MDE framework that introduces "Intra-Domain Awareness" -- the capability to map diverse stylistic inputs sharing identical geometry into a unified latent space. To achieve this, we first align shallow features within the Diffusion Transformer (DiT) to anchor degraded observations to unified geometric representations. Additionally, we introduce a joint RGB latent reconstruction task, requiring the model to recover clean RGB information from degraded inputs. This multi-stage strategy effectively decouples environmental style interference, compelling the model to focus strictly on underlying geometric recovery. Extensive evaluations show that OpenDepth sets new SOTA records for robust zero-shot depth estimation across standard and extreme weather benchmarks, demonstrating exceptional structural integrity and adaptability.

cs.AI↗
arXiv2026

A grand potential for molten-salt electrochemical phase diagrams from first principles

The clean-energy transition is driving unprecedented demand for metals, whose extraction is among the largest industrial sources of carbon emissions. Molten-salt electrolysis offers a path toward carbon-free production of aluminum, iron, titanium, and critical metals. However, corrosive operating conditions and complex, multicomponent feedstocks make experimental development costly and challenging. Here, we introduce a first-principles framework for constructing electrochemical phase diagrams of multicomponent systems to accelerate electrode and electrolyte design and guide the selection of compatible materials and operating conditions for electrochemical cells. We derive a grand potential that generates these diagrams at scale, and supply free energies using first-principles data obtained through a fine-tuned foundation machine-learned interatomic potential. For silicon in molten CaCl$_2$, the construction not only recovers the experimental diagram when using the first-principles framework on a limited set of structures, but also predicts further domains when extended to a larger set of crystal structures. Finally, we demonstrate the predictive power of the approach by screening ten alkali and alkaline-earth halide electrolytes for aluminum electrowinning. Three thermodynamic criteria identify NaF, where cryolite emerges as the stable aluminum-bearing phase, recovering the chemistry that Hall and Héroult established empirically in 1886. The framework provides a first-principles basis for selecting compatible electrodes, electrolytes, and operating conditions toward low-carbon metal production from complex source materials.

cond-mat.mtrl-sci↗
arXiv2026

From Syndromes to Computation: A General Framework for Logical Operations in Quantum Error-Correcting Codes

Fault-tolerant quantum computation requires error-correcting codes that not only protect quantum information but also support useful logical operations with practical physical realizations. Here we develop a general framework in which the stabilizer syndromes of commuting Pauli rotations directly organize logical-gate design. We derive an exact codespace-preservation condition valid at arbitrary rotation angles, providing a common stabilizer-level description for CSS and non-CSS codes and for both Clifford and non-Clifford operations. A particularly useful equal-syndrome specialization reduces Clifford gate design to stabilizer factorization, yields a general generator weight bound of $\lceil d/2\rceil$ for distance-$d$ codes, and systematically extends to logical rotations higher in the Clifford hierarchy. Applied to bivariate-bicycle (BB) and non-CSS Mirror qLDPC codes, the framework gives explicit low-weight Clifford and controlled-phase gates. Our implementations yield new low-weight gates, including an addressed $T$ gate on the $[[16,4,4]]$ BB code, $T$-type rotations on the $[[64,8,8]]$ BB code, and CCZ-, CS-, and $T$-type operations on the $[[36,6,6]]$ Mirror code. We also develop analytic constructions using Clifford algebra bases, which turn the syndrome constraints into explicit algebraic relations and provide a systematic route to logical addressability. We use this approach to analytically construct addressed CNOT, SWAP, controlled, and CCZ-type gates on recursive hypercube codes. An independent construction on middle quantum Reed--Muller codes yields C${}^r$Z gates that saturate the equal-syndrome weight bound. These results establish stabilizer syndromes as a constructive link between the stabilizer structure of a quantum code, the logical computations it can directly support, and the fault-tolerance properties of their physical realizations.

quant-ph↗
arXiv2026

MAS-Gate: Unlocking Multi-Agent System Lifecycle via Node-Centric Quantification

Large language model (LLM)-based multi-agent systems (MAS) can solve complex problems beyond the reach of a single agent by decomposing tasks and coordinating complementary capabilities. Despite extensive research on MAS, existing work largely focuses on specific aspects or individual phases of their development and operation, resulting in fragmented perspectives with objectives and signals that are difficult to reconcile across these different settings. To bridge these disconnected perspectives, we conceptualize MAS development and operation as a three-phase lifecycle spanning design-time orchestration before execution, runtime intervention during execution, and post-hoc attribution after execution, and introduce MAS-Gate, a unified node-centric quantification applicable across these phases. Modeling each execution as a directed acyclic graph, MAS-Gate combines within-node utility, which estimates a node's contribution through changes in reference-answer likelihood, with inter-node strength, which approximates semantic uptake and potential downstream reach. We adapt this shared quantification to three objectives: providing a structure-aware dense reward for Orchestrator training, training an Editor to act on execution prefixes using paired rewards from completed edited and unedited trajectories, and supplying risk-ranked nodes as auxiliary evidence for LLM-based post-hoc attribution. Experiments spanning interaction environments, knowledge-intensive multi-hop question answering, general reasoning, and post-hoc attribution demonstrate that MAS-Gate improves workflow construction, targeted runtime editing, and LLM-assisted error localization over corresponding phase-specific baselines. Together, these results establish MAS-Gate as a unified structural interface for improving MAS generation, dynamic editing, and error attribution throughout the lifecycle.

cs.AI↗
arXiv2026

Extending Layered Attestation into ARM Application Run-Times

Modern mobile and edge platforms provide hardware-backed attestation of application and platform integrity at load-time. However, they offer little support once application execution begins. We present ARA, an Application-Runtime Anchor for ARM that extends this load-time chain of trust into execution. ARA bootstraps from the underlying ARM platform Root of Trust (RoT) through a measured loading chain and establishes an attested protected run-time within the application address space using recently introduced ARM Permission Overlay Extensions (POE). The resulting run-time domain can host programmable RoTs for Measurement and Storage (RTM/RTS), enabling application run-time integrity verification, attestation evidence generation, and protected use of secrets while retaining efficient access to application run-time state. As exemplary uses, we instantiate ARA for protected media decryption and application control flow measurement. Our prototype and evaluation show that ARA supports frequent run-time RoT-backed operations at near in-process cost, with roughly 10x and 50x lower overhead than out-of-process equivalents. This offers a practical bridge from layered attestation to application run-time trusted computing on ARM.

cs.CR↗
arXiv2026

Background2Background: a self-supervised background removal framework for high dimensional microscopy

High-dimensional microscopy captures features across space, time, and spectrum, but complex backgrounds obscure signals. We present Background2Background (B2B), a self-supervised background removal framework leveraging the backgrounds larger correlation relative to the signal. Local pixel shuffling disrupts signal correlations while preserving background structure, transforming signals into effective noise for Noise2Noise-based background removal. Across fluorescence, stimulated Raman scattering and mid-infrared photothermal imaging, B2B achieves over 40-fold background suppression and reveals hidden signatures.

eess.IV↗
arXiv2026

StateHarness: Maintaining Evolving Project State for Long-Horizon Coding Agents

Long-horizon coding requires agents to maintain an evolving view of the project they are changing: what problem is currently being solved, what has already been learned or modified, which other artifacts may be affected, and what work remains unresolved. When this information is scattered across a growing interaction history, local edits can become disconnected from their later consequences and progress can be difficult to carry forward. We introduce StateHarness, a runtime framework that maintains evolving project state without updating model parameters. The working agent records its current problem and hypothesis, important findings and decisions, discovered repository relationships, and unfinished work. Runtime support prompts record updates, searches for callers, tests, and documentation that may be affected by a change, and maintains a compact working brief for subsequent decisions. When behavioral signals accumulate, a separate diagnostic session revisits the recorded history and surfaces overlooked evidence. Across three repository-level evaluations, StateHarness improves task completion and several measures of change consistency. On RefactorBench, full-chain completion increases from 29.4% to 35.3% and cleanup-check accuracy from 87.0% to 95.4%. On five LHTB software-engineering tasks, mean reward improves from 0.427 to 0.576, exceeding LongHorizon-Harness at substantially lower recorded execution cost. On ChainSWE, bug-fix rate increases from 35.1% to 39.5% and residual debt decreases from 60 to 43 items, although pass-to-pass regressions increase. These results show that maintaining project state can improve long-horizon coding while introducing measurable execution and regression trade-offs.

cs.SE↗
arXiv2026

What Do Deepfake Audio Detectors Hear? From Self-Supervised Models to Audio LLMs

Real-world audio recordings for deepfake detection rarely consist solely of speech; they often are mixtures of speech and background events that themselves may also be synthetic. When processing these mixtures, it remains unclear whether detectors attend to the speech or the background event, and where their spectral evidence lies. We study this on the CompSpoof V2 database, whose mixtures pair independently real-or-fake speech and event components. Our results show that all systems detect fake events more accurately than fake speech. However, this performance gap is largely driven by poor generalization to an unseen fake speech domain, rather than a true difference in difficulty. Furthermore, sub-band probing reveals that the front-end encoder dictates where a countermeasure focuses its frequency evidence. To enable reliable deployment in real-world scenarios, future countermeasures should address out-of-domain generalization failures and carefully select front-end encoders aligned with their target acoustic evidence. Our source code and checkpoints are released at https://github.com/nii-yamagishilab/AntiDeepfake-AudioLLM.

cs.SD↗
arXiv2026

Observational evidence of black hole event horizon from ultra-compact X-ray binaries

The event horizon is the most crucial feature that distinguishes black hole (BH) from all other objects in the universe. However, since no radiation can escape from the event horizon, its direct observational evidence remains lacking. Although both are compact objects, the event horizon leads to differences in the radiative behavior between black hole and neutron star (NS) X-ray binaries. Historically efforts to probe the existence of event horizons revealed that the BH X-ray binaries are fainter than NSs with similar orbital periods in their quiescent states. Here, we use a special class of X-ray binaries namely the ultra-compact X-ray binaries (UCXBs), to reveal the existence of the event horizon near stellar-mass black holes. In short-period UCXBs, gravitational radiation dominates the orbital angular momentum loss, allowing us to accurately determine their accretion rates. Based on five UCXBs with orbital periods shorter than 40 min, located in globular clusters, and identified accretors (NS or BH), we obtained their accretion rates. Combining their accretion rates and X-ray luminosities, we calculated the X-ray radiative efficiencies of NS and BH. The radiative efficiency of BH is about two to three orders of magnitude lower than that of NS systems, indicating the accretion energy is swallowed by the BH, thereby indirectly probe the presence of the event horizon.

astro-ph.HE↗
arXiv2026

When Does an ASR Error Matter? Task-Role Diagnostics for Tool-Using Voice Agents

In cascaded voice agents, ASR errors can have markedly different downstream consequences: a small transcription error may leave a tool call unchanged or alter its intent or arguments. Conventional word error rate (WER), however, does not capture these differences. We introduce SpeechEvalX, a paired diagnostic protocol that evaluates how task roles affected by ASR errors influence structured tool calls. Controlled edits are matched by source utterance, edit count, and operation, while task-role labels are compared with lexical severity, BERT cosine distance, Semantic-WER, and binary criticality. Across 594 source utterances, joint call accuracy falls from 82.3% with reference transcripts to 71.9-74.9% across three ASR front ends. Number and negation edits cause 21.9-25.7 percentage-point losses, versus 1.4-2.9 points for control edits. Detailed task-role labels improve source-grouped ROC-AUC from 0.73 to 0.78 over binary criticality, showing that downstream impact depends strongly on the task role affected.

cs.CL↗
arXiv2026

Smartphone-Based, Field-Deployed AI Inspection System for Power Distribution Infrastructure

Routine inspection of power distribution infrastructure remains challenging for utilities operating geographically dispersed networks, where conventional manual inspection can be difficult to perform consistently and at scale. This work presents a smartphone-based AI inspection system for assessing targeted distribution-infrastructure conditions from images captured by field personnel. The proposed workflow integrates image-quality and viewing-perspective validation with AI-assisted assessment of pole lean, crossarm and top-cleat tilt, voltage level, vegetation-clearance risk, and distribution-transformer (DTR)-ground clearance. The system integrates deep-learning models with monocular depth estimation, geometric analysis, and rule-based processing within a common multi-stage inspection workflow designed for practical field operation. The system was pilot deployed across five selected distribution feeders within the Eastern Power Distribution Company of Andhra Pradesh Ltd. (APEPDCL), where 1,222 defect instances were identified across the assessed categories and subsequently tracked through a rectification and verification workflow, with 84 defect instances recorded as rectified during the study period. Feeder-level interruption records were also examined over the study period; however, the available observations do not provide sufficient evidence to establish a reliability improvement attributable to the proposed system. A curated multi-module field-captured dataset, PD-Defect, is also publicly released to support reproducible research in distribution-infrastructure inspection. The field deployment demonstrates the practical integration of smartphone-based AI-assisted inspection into field inspection and maintenance activities, providing a mechanism for systematic identification, documentation, rectification tracking, and verification of targeted distribution-infrastructure defects.

cs.CV↗