Searcharxiv⌕ Search

arXiv subjects

Xiangyu Zhou

Publications and source records attributed to Xiangyu Zhou.

At least 19 recordsLinked to original sources

On strongly pseudoconvex Kähler--Berwald manifolds

In this paper, we establish the equivalence between the Kähler--Finsler and Kähler--Berwald conditions for complex Finsler metrics. That is, a Kähler--Finsler metric always satisfies the complex Berwald condition: its Chern--Finsler nonlinear connection reduces to the Chern connection of a Kähler metric. We also provide, for the first time, smooth weakly Kähler--Finsler examples which are not Kähler--Berwald; and propose a condition that is formally stronger than weakly Kähler--Finsler, with equivalence holding in complex dimension two. Then we prove the de Rham decomposition and uniformization theorems for Kähler--Berwald manifolds, and show that every strongly pseudoconvex locally symmetric complex Finsler metric is both Kähler--Berwald and complex Einstein--Finsler. Additionally, we establish several rigidity theorems for the Kobayashi and Carathéodory metrics: $F_K$ is Kähler--Berwald if and only if it is Kähler; if the manifold is simply connected, then $F_C$ is Kähler--Berwald if and only if it is Kähler.

math.DG↗

ExGrad: Beyond Suppression - Explicit Clean-Gradient Subtraction in Adversarial Training

Adversarial training derives updates from surrogate losses, but source-loss descent does not directly characterize an update's effect on adversarial margins. We introduce ExGrad (extrapolative gradient recomposition), which constructs clean-input, adversarial-input, and adversarial-weight-perturbed (AWP) gradients. Virtual steps measure their margin responses, and a globally solvable three-variable problem selects coefficients over a bounded affine domain without nonnegativity constraints. Across affine radii on CIFAR-10, negative clean coefficients rapidly predominate and increase in magnitude with the radius. Cross-zero interventions show that setting the clean coefficient to zero is insufficient: mild subtraction improves PGD-20, while stronger subtraction trades clean accuracy for further PGD-20 gains. In CIFAR-10 WRN-28-4 controls, none of the three candidate-removal variants matches full ExGrad. Projection diagnostics show that negative clean coefficients usually coexist with positive net clean-axis projections, although both decrease as the radius grows. Norm-matched controls indicate that update magnitude alone does not explain the moderate-radius gains. Experiments on CIFAR-10, CIFAR-100, and ImageNet-100 show radius-dependent improvements over AWP. On CIFAR-100, ExGrad at $R=0.8$ improves all five reported metrics, including clean and AutoAttack accuracy by 3.84 and 1.75 percentage points. On ImageNet-100, the same radius improves all four robust metrics over both AWP and PGD-AT. These results identify clean-gradient subtraction as a response-revealed update preference whose strength and clean-robust operating point are controlled by the affine radius.

cs.LG↗

SpatialCORE: Confidence-Aware Grounded Spatial Reasoning in Large Vision--Language Models

Large Vision-Language Models (LVLMs) have made remarkable progress across visual perception tasks, yet spatial reasoning remains a persistent weakness, especially for questions that require reasoning over visual space. Recent spatial-reasoning methods incorporate generated grounding, where models predict bounding boxes, masks, or other localization outputs for task-relevant objects as part of their reasoning trace. However, these approaches typically optimize final-answer correctness alone, allowing correct answers to be rewarded even when the model does not reason from confidently localized task-relevant objects. We introduce SpatialCORE (Spatially COnfident REasoning), a post-training framework that turns the model's own confidence in generated grounding into a learning signal for spatial reasoning. Its central idea is to reinforce grounding that is both accurate and confident, encouraging the model to reason from confidently localized task-relevant objects. SpatialCORE realizes this through a self-regulating spatial reward that weights each predicted bounding box's matching quality by its coordinate-token confidence. An answer gate further ties grounding optimization to final-answer correctness. SpatialCORE achieves state-of-the-art results among open-source and specialized spatial reasoning models across diverse benchmarks, and transfers effectively in zero-shot settings to unseen data distributions. The source code is available at https://github.com/rafiibnsultan/SpatialCORE.

cs.CV↗

Invariance of plurigenera for Kähler families with nef canonical bundles

Let $p:X\rightarrowΔ$ be a proper holomorphic submersion from a Kähler manifold $X$. We prove that, if $K_{X_t}$ is nef for every $t\inΔ$, then $\dim_{\mathbb C}H^0(X_t,mK_{X_t})$ is independent of $t$ for every integer $m\geq1$. This gives a partial answer to Siu's conjecture on the invariance of plurigenera for Kähler families.

math.CV↗

Towards Mitigating Deceptive Safety Alignment in Large Reasoning Models

Large Reasoning Models (LRMs) are commonly trained with reinforcement learning (RL) to improve their generation of chain-of-thought (CoT) reasoning before producing final answers. However, RL rewards are typically assigned based on final answers, providing little or no direct supervision over intermediate reasoning. This can lead to deceptive safety alignment, where the reasoning trace and final answer convey inconsistent safety signals. To systematically investigate this phenomenon, we introduce DSAR (Deceptive Safety Alignment Rate), a metric that jointly assesses reasoning traces and final answers to quantify their safety inconsistency. Across multiple LRMs and benchmarks, we find that deceptive safety alignment is pervasive under standard prompting conditions and is substantially amplified under prefilling attacks. We further provide a hidden representation analysis showing that models exhibit stronger safety discrimination at the final-answer stage than during intermediate reasoning. To close this gap, we propose SARA (Safety-Aware Reasoning Alignment), an RL-based method that rewards both safety-aware reasoning and safe final answers, encouraging early harmful intent recognition and enforcing reasoning-answer consistency. Experiments show that SARA significantly mitigates deceptive safety alignment under both standard and adversarial settings while preserving helpfulness and utility. Code is available at https://github.com/xzhou98/SARA.

cs.AI↗

A Converse to the Skoda $L^2$ division theorem II

We prove a converse to the original Skoda $L^2$ division theorem, asserting that Skoda $L^2$ division estimates for a single nonzero holomorphic function force the weight to be plurisubharmonic. In addition, we establish corresponding results for higher-rank quotient bundles. Finally, we derive an optimal $L^2$ extension theorem directly from the division estimates.

math.CV↗

Uniform RC-positivity of tangent bundles

In this paper, we prove that every rationally connected projective manifold admits a smooth uniformly RC-positive Hermitian metric on its tangent bundle, answering Yang's question and giving a characterization of rational connectedness by uniform RC-positivity. We find an example to show that RC-positivity alone does not characterize rational connectedness. We also prove that uniform RC-positivity of the tangent bundle is preserved under blow-ups along connected smooth centers on compact complex manifolds. In the non-Kähler setting, we construct such metrics on all Hopf and Kato surfaces and obtain classification results for compact complex surfaces.

math.CV↗

A remark on Bogomolov type vanishing theorem

Several vanishing theorems for pseudo-effective line bundles are presented. All related results in the present note can be derived from already known theorems, in particular from the Steenbrink vanishing theorem and Boucksom's generalization of Bogomolov's vanishing theorem. No originality or priority is claimed for these statements; they are listed here merely as conjugate forms of the Kawamata--Viehweg vanishing theorem, for convenience of reference and comparison with existing literature.

math.AG↗

A logarithmic Bogomolov--Sommese vanishing theorem on compact Kähler manifolds

In this paper, we establish a logarithmic Bogomolov--Sommese vanishing theorem in terms of numerical dimension for pseudo-effective line bundles on compact Kähler manifolds. As an application, we obtain a rigidity result with vanishing second Chern class for logarithmic cotangent bundles by combining the vanishing theorem with a structure theorem of Iwai and Matsumura.

math.AG↗

Alignment of LRMs via Counter-Aligned Few-Shot Conversation Exposure

Large Reasoning Models (LRMs) rely on explicit chain-of-thought (CoT) reasoning and large context windows to achieve strong performance on complex tasks, but these features also introduce new attack surfaces. We show that LRMs' reasoning processes can be systematically steered by prepending counter-aligned few-shot conversations containing explicit CoT traces, leading to unsafe generations on harmful queries and unwarranted refusals on benign ones. We formalize this attack as SRCF (Steering Reasoning via Counter-Aligned Few-shot Conversations) that operates solely through a flexible conversational interface and requires no access to the model's parameters and gradients. Our key insight is that SRCF exploits an adversarial generalization issue that induces a representation drift, causing the representations of benign and harmful inputs to shift in a similar direction. This observation motivates our post-training defense, ARCF (Aligning Reasoning via Counter-Aligned Few-Shot Conversations), which exposes models to counter-aligned conversational contexts while enforcing aligned targets. ARCF is compatible with existing post-training methods and consistently improves safety and helpfulness without degrading utility.

cs.AI↗

On the extension of Kähler currents on compact complex manifolds

Let $(X,ω)$ be a compact Kähler manifold and let $V\subset X$ be a closed complex submanifold. Coman-Guedj-Zeriahi proposed the problem: is every $ω|_V$-plurisubharmonic function on $V$ the restriction of an $ω$-plurisubharmonic function on $X$? In this paper, we solve this problem affirmatively, even for a compact Hermitian manifold.

math.CV↗

Capacity Stability of Complex Monge-Ampère Equations with Moving Prescribed Singularities

For complex Monge-Ampère equations with moving big cohomology classes and prescribed model singularities of positive Monge-Ampère mass, we prove that, under total variation convergence of the right-hand side non-pluripolar positive Radon measures, convergence of the prescribed model potentials in Monge-Ampère capacity is equivalent to convergence in capacity of the associated normalized solutions. We further prove that the ceiling operator coincides with the singularity envelope for potentials associated to a big $(1,1)$-class, regardless of their Monge-Ampère mass, thereby resolving a conjecture of Darvas-Di Nezza-Lu. Consequently, the singularity envelope is idempotent without the positivity assumption on the mass.

math.CV↗

Vanishing theorems for pseudo-effective line bundles

In the present paper, we establish a general Kawamata-Viehweg-Kollár-Nadel type vanishing theorem for higher direct images in terms of numerical dimension for closed positive currents on compact Kähler manifolds, unifying a number of important vanishing theorems.

math.CV↗

Reconsidering Overthinking: Penalizing Internal and External Redundancy in CoT Reasoning

Large reasoning models (LRMs) often exhibit overthinking, producing verbose Chain-of-Thought (CoT) traces that increase inference cost and obscure the underlying reasoning process. Existing CoT compression methods mainly rely on global length rewards, which conflate necessary intermediate reasoning with redundant text and may therefore compromise reasoning fidelity. This paper revisits overthinking from a semantic-efficiency perspective and decomposes CoT redundancy into two distinct forms: internal redundancy, defined as informational stagnation before the first correct answer, and external redundancy, defined as superfluous continuation after the first correct answer. Based on this decomposition, we propose a dual-penalty reinforcement learning framework that separately optimizes reasoning progress and termination behavior. Specifically, a sliding-window semantic similarity metric penalizes low-progress reasoning segments, while a normalized external-redundancy metric discourages post-answer continuation. Experiments on GSM8K, MATH500, and AIME24 across different model scales show that our method reduces average reasoning length by 41.3% on the 1.5B model and 40.1% on the 7B model, while preserving competitive accuracy and achieving the best overall accuracy-efficiency score among evaluated baselines. The learned compression behavior further transfers to out-of-domain reasoning tasks, including GPQA and LiveCodeBench. More importantly, our analysis reveals a clear asymmetry between the two redundancy types: external redundancy can be largely removed with little performance loss, whereas internal redundancy compression follows a sensitive accuracy-efficiency trade-off. These results suggest that effective CoT compression should optimize semantic efficiency rather than sequence length alone, offering a principled route toward more concise, efficient, and interpretable LRMs.

cs.AI↗