Search arXiv⌕ Search

arXiv subjects

Xin Su

Publications and source records attributed to Xin Su.

At least 19 recordsLinked to original sources
arXiv2026

Post-Hoc Conformal Prediction for Reliable Wireless Communications

Deploying artificial intelligence (AI) in high-stakes wireless applications such as autonomous transportation requires guarantees of reliable operation. Such guarantees can often be obtained by designing systems that act on a set of predictions via conservative policies catering to all possible outcomes within the set. For instance, in location-based beam selection, a base station may identify a set of plausible locations and select a beam that ensures high capacity over the set. The miscoverage probability of the set with respect to the true outcome (e.g., the true user location) then quantifies the outage probability, while the size of the set determines the final performance or resource budget (e.g., the transmission capacity). Conventional conformal prediction (CP) applies when the target miscoverage level is prescribed in advance. However, in practical wireless systems, prediction sets may instead be selected under prescribed operational constraints, requiring the resulting miscoverage probability to be quantified post hoc. This paper develops a formal statistical framework for the data-driven selection of prediction sets that provides a reliable estimate of the resulting miscoverage probability. The methodology builds on backward CP and probably approximately correct CP, yielding distribution-free reliability guarantees on the miscoverage probability. We apply the framework on three distinct wireless applications: narrowband interference detection, near-field localization, and codebook-based beam identification. Numerical results validate the reliability guarantees and show that the proposed post-hoc conformal methods achieve accuracy comparable to the naive probability-based approach across all considered applications.

eess.SP↗
arXiv2026

Rachel: A general-purpose language model directs and revises retrosynthetic routes

Retrosynthetic planning advances through decisions that reshape the remaining chemical problem: a locally plausible disconnection can leave precursors whose chemoselectivity constraints complicate the rest of the route. Existing planners often channel model proposals through search or template procedures, leaving open whether a general-purpose large language model (LLM) can itself sustain and revise route strategy. We developed Rachel, a stateful environment that executes and checks LLM-directed chemistry but prescribes neither a search policy nor a stopping rule. Without supplied reference routes or route-level solutions, GPT-5.5 achieved strict closure for 111 of 120 PaRoutes120 targets and 24 of 25 targets in the separate RF25 difficult-target cohort. RF25 was drawn largely from studies published after GPT-5.5's reported knowledge cutoff. Closure required complete routes and independent source resolution of every terminal precursor after planning. On a shared PaRoutes subset, forward-model support exceeded that of most comparator methods, and Rachel received the highest mean overall route score from both method-blinded LLM evaluators. Recorded trajectories showed continued model-proposed chemistry, with revised strategies carried into subsequent steps. Replacing LLM route decisions with fixed policies reduced strict closure to 6-15/120 despite continued local chemical execution; restricting planning support also reduced closure in RF25. Within Rachel, a general-purpose LLM coordinated successive chemical choices and revised its strategy as earlier decisions reshaped the remaining problems.

physics.chem-ph↗
arXiv2026

G-ray: Ray-Level Relative Geometric Position Encoding in Multi-View Vision Transformers under Camera Heterogeneity

We study relative position encoding for multi-view vision Transformers under camera heterogeneity, including varying fields of view (FoVs) or projection models. Existing rotary relative position encodings commonly use image-plane positional coordinates, producing projection-dependent relative phases and inconsistent geometric cues for cross-projection attention. We introduce G-ray, a ray-level relative position encoding whose rotary phases are parameterized by camera-local ray angles. The same camera-local ray pair induces the same relative phase across projections, providing projection-invariant positional consistency. G-ray can be used directly or integrated with existing encodings, retaining complementary geometric cues without additional learned parameters. We validate G-ray in three host encodings, RoPE, GTA, and RayRoPE, across 3D reconstruction and novel-view synthesis (NVS). Across three heterogeneous 3D reconstruction benchmarks at 50 views, G-ray leads all six averaged metrics and reduces mean pointmap relative error by 45.8% over MapAnything, with calibration supplied to both. Trained exclusively on homogeneous pinhole images, the 3D reconstruction model handles mixed pinhole and non-pinhole inputs without retraining and remains competitive on homogeneous pinhole 3D reconstruction protocols. For NVS, GTA and RayRoPE improve with G-ray under joint viewpoint and FoV variation. The project's webpage is available at https://g-ray-project.github.io/.

cs.CV↗
arXiv2026

Balanced truncation for bilinear time-delay systems with approximate Gramians

A balanced truncation (BT) procedure is presented to achieve a structure-preserving model order reduction (MOR) for bilinear time-delay systems in this paper. We attempt to define Gramians for bilinear time-delay systems based on Volterra series theory. The controllability and observability Gramians in the frequency domain are given explicitly, which align with the ones of bilinear systems and time-delay systems. Based on the derived Gramians, a BT procedure is provided to produce reduced order models with the same structure. We also propose a numerical quadrature rule based on the truncated Laguerrre expansion to give an accurate approximation to Gramians. The resulting low-rank structure of approximate Gramians benefits a lot the efficient execution of the whole MOR procedure. Two numerical examples are simulated to showcase the efficiency of our approach.

math.DS↗
arXiv2026

Heralded Non-Gaussian Squeezed-State Inputs for Parity-Detection SU(1,1) Interferometry

Non-Gaussian operations can reshape the photon statistics of continuous-variable probes, but their metrological advantage is meaningful only when heralding probability and photon-number resources are counted consistently. We compare photon subtraction, photon addition, and photon catalysis as input-side heralding operations in a balanced SU(1,1) interferometer with parity detection. A unified finite-transmissivity map supplies closed conditional moments and the corresponding quantum Fisher information at arbitrary operation order; internal loss is absorbed into a single effective parity observable whose lossless limit recovers the ideal pulled-back measurement. At fixed preparation parameters, single-photon subtraction and addition improve the conditional phase information over the Gaussian reference across most of the high-transmissivity regime, while multi-photon catalysis opens useful low-transmissivity windows. However, when the coherent--squeezed allocation is independently optimized at fixed conditional-probe energy and fixed interferometer gain, the success-weighted Fisher information of all three non-Gaussian operations remains below the optimized Gaussian benchmark. This conclusion is subject to the tested constraints: single-photon operations, a coherent-plus-squeezed-vacuum Gaussian family, fixed gain, and parity readout. Photon catalysis separately generates a conditional branch with high local quantum Fisher information that dark-point parity extracts poorly, identifying a measurement mismatch rather than a state-preparation failure. The result draws a sharp boundary between conditional non-Gaussian enhancement and practically available precision under explicitly stated resource constraints.

quant-ph↗
arXiv2026

CoRE-UIR: Prior-guided common and residual experts for efficient all-in-one remote sensing image restoration

Remote sensing images acquired by unmanned aerial vehicles (UAVs) and satellites are often degraded by adverse weather, illumination variation, and imaging artifacts, which may co-occur and jointly induce global distribution shifts and local structural corruption. Although All-in-One image restoration offers an appealing unified alternative to task-specific pipelines, existing methods still suffer from weak or implicit degradation cues and parameter redundancy caused by full-rank multi-expert designs with overlapping restoration behaviors. We propose CoRE-UIR (Common and Residual Experts for Universal Image Restoration), a prior-guided global-local framework centered on the Common-and-Residual Expert Block (CoRE). CoRE explicitly decomposes restoration capacity into a common dense expert for degradation-invariant restoration and low-rank residual experts for degradation-specific compensation, enabling adaptive specialization without redundant expert replication. Built on this design, Degradation Prior Embedding (DPE) adapts frozen CLIP features into an explicit restoration-oriented prior, while Global Feature Modulation (GFM) aligns global feature statistics before local residual compensation. We also construct MDVD-108K (Multi-Degradation VisDrone), a large-scale UAV restoration dataset covering both single and compound degradations, together with a real-world test set. Extensive experiments on multiple datasets show that CoRE-UIR improves the overall average PSNR by 1.05 dB while running 11.83$\times$ faster and reducing peak memory by 85.3% relative to the strongest baseline, BaryIR, thereby maintaining a favorable quality-efficiency trade-off. Evaluations on downstream tasks and unseen degradation also validate the generalizability of CoRE-UIR. The code and dataset will be released at https://github.com/zzaiyan/CoRE-UIR.

cs.CV↗
arXiv2026

Refusal-Gated Decoding: Preserving Refusal Behavior Under High-Temperature Sampling

High-temperature sampling is one of the primary mechanisms for increasing diversity in LLMs. Recent advances in truncation-based sampling techniques have helped mitigate drawbacks of high-temperature sampling such as neural text degeneration, thereby enabling greater diversity in LLM outputs without sacrificing coherence. However, increasing the entropy of the token probability distribution via high temperatures has also been shown to weaken model guardrails by reducing the model's refusal response in the presence of harmful prompts. Despite the potential benefits of high-temperature sampling and the importance of maintaining model safety, there is a lack of existing solutions for maintaining the refusal behavior of LLMs under a higher entropy regime. To address this gap, we systematically study how temperature influences refusal behavior in LLMs and propose an efficient sequential decoding approach which preserves a model's greedy decoding refusal response at high temperatures while incurring minimal additional latency. Through extensive experiments, we show that our approach preserves 91-99% of the greedy decoding refusal behavior across three benchmark datasets without compromising the model's high-temperature response for safe prompts. Our work demonstrates how refusal behavior can be maintained in an efficient manner for applications which require high-temperature sampling.

cs.AI↗
arXiv2026

Cross-Cultural Value Attribution in Large Vision-Language Models

The rapid adoption of large vision-language models (LVLMs) in recent years has been accompanied by growing fairness concerns due to their propensity to reinforce harmful societal stereotypes. While significant attention has been paid to such fairness concerns in the context of social biases, relatively little prior work has examined the presence of stereotypes in LVLMs related to cultural contexts such as religion, nationality, and socioeconomic status. In this work, we aim to narrow this gap by investigating how cultural contexts depicted in images influence the judgments LVLMs make about a person's moral, ethical, and political values. We conduct a multi-dimensional analysis of such value judgments in nine LVLMs using counterfactual image sets, which depict the same person across different cultural contexts. Our evaluation framework pairs descriptive analyses (Moral Foundations Theory categorization, lexical analyses, and value sensitivity) with a novel grounding analysis that compares LVLM cross-context variation against two large-scale human surveys (MFQ-2 and WVS Wave 7). Across 4.8 million LVLM generations, we identify three bias patterns that replicate across architecturally diverse models: an inversion of the socioeconomic-status-to-Authority relationship found in WVS, and two race-conditional failures that override cultural context cues when depicting Middle Eastern persons. Additional ablations show that the socioeconomic-status-to-Authority inversion bias is amplified by image conditioning and persists across different model sizes.

cs.CV↗
arXiv2026

Cultural Counterfactuals: Evaluating Cultural Biases in Large Vision-Language Models with Counterfactual Examples

Large Vision-Language Models (LVLMs) have grown increasingly powerful in recent years, but can also exhibit harmful biases. Prior studies investigating such biases have primarily focused on demographic traits related to the visual characteristics of a person depicted in an image, such as their race or gender. This has left biases related to cultural differences (e.g., religion, socioeconomic status), which cannot be readily discerned from an individual's appearance alone, relatively understudied. A key challenge in measuring cultural biases is that determining which group an individual belongs to often depends upon cultural context cues in images, and datasets annotated with cultural context cues are lacking. To address this gap, we introduce Cultural Counterfactuals: a high-quality synthetic dataset containing nearly 60k counterfactual images for measuring cultural biases related to religion, nationality, and socioeconomic status. To ensure that cultural contexts are accurately depicted, we generate our dataset using an image-editing model to place people of different demographics into real cultural context images. This enables the construction of counterfactual image sets which depict the same person in multiple different contexts, allowing for precise measurement of the impact that cultural context differences have on LVLM outputs. We demonstrate the utility of Cultural Counterfactuals for quantifying cultural biases in popular LVLMs.

cs.CV↗
arXiv2026

Synthetic Contrastive Reasoning for Multi-Table Q&A

Multi-table question answering requires models to retrieve relevant evidence, link schemas, and perform compositional reasoning across relational tables. Existing multi-table Q&A resources typically provide questions and final answers but lack reasoning supervision that explains how answers are derived. To address this gap, we construct a synthetic contrastive reasoning-trace dataset for MMQA by generating validated positive traces and plausible negative traces with heterogeneous LLMs. We then use the resulting preference pairs to fine-tune open-weight LLMs with Contrastive Preference Optimization (CPO). Across Qwen3-14B, Mistral-8B, and Llama-3.1-8B, CPO achieves absolute average improvements over Q&A supervised fine-tuning ranging from 9.7%-16.3%, with gains up to 21 percentage points on MMQA. Ablations show that heterogeneous positive and negative trace generators strengthen the contrastive signal, and automated as well as human evaluations indicate that the generated pairs are largely faithful, coherent, and meaningfully contrastive.

cs.AI↗
arXiv2026

Hybrid Verified Decoding: Learning to Allocate Verification in Speculative Decoding

Large Language Model (LLM) generation remains expensive because autoregressive decoding calls the model once for each new token. Speculative decoding reduces this cost by drafting multiple tokens and verifying them with the target model in one step, but its speedup depends on how many drafted tokens are accepted. Parameter-free draft sources can propose long continuations at low cost in structured and agentic workloads, yet a cache match that looks promising at one generation step may have low payoff at the next. We propose Hybrid Verified Decoding, which predicts the accepted length of a cache draft before verification and uses this payoff estimate to choose between cache verification and a model-based drafter. Across three LLMs and sixteen datasets, Hybrid Verified Decoding is especially effective on agentic workflows, where it outperforms EAGLE3 in every setting with a 2.73x average speedup. Our analysis shows how prompt structure creates cache opportunities, how high-payoff cache drafts concentrate in a small part of the draft space, and how payoff-guided selection reduces sequential decoding work, pointing to runtime draft selection as a promising direction for speculative decoding.

cs.CL↗
arXiv2026

Enhancing Phase Estimation in a Hybrid Interferometer via Kerr Nonlinearity and Photon Subtraction

We propose a high-precision phase estimation scheme in a hybrid interferometer by synergistically combining a Kerr nonlinear phase shifter and multi-photon subtraction operations. Using a coherent state and a vacuum state as input resources, we systematically evaluate the phase sensitivity via homodyne detection and analyze the quantum Fisher information as well as the quantum Cramér-Rao bound under both ideal and lossy conditions. Our results show that the joint integration of Kerr nonlinearity and multi-photon subtraction yields remarkable advantages over either technique used alone. The proposed scheme enables the phase sensitivity to surpass the standard quantum limit, exceed the conventional Heisenberg scaling ($1/N$), and approach the super-Heisenberg scaling ($1/N^{2}$)-a direct consequence of Kerr nonlinearity. More precisely, the super-Heisenberg scaling $\propto $ $1/N^{2}$ is the ultimate precision limit permitted by the $k=2$ Kerr nonlinearity and does not violate the fundamental Heisenberg limit for linear phase accumulation. Even under moderate internal photon loss, the system maintains high precision and exhibits enhanced robustness to decoherence. The Kerr nonlinearity introduces an intensity-dependent phase shift proportional to the squared photon number, while multi-photon subtraction tailors non-Gaussian states to strengthen phase information extraction. Compared with existing schemes based on hybrid interferometers or SU(1,1) interferometers, our architecture achieves superior precision and stronger loss resilience. All components are experimentally accessible with current quantum optical technologies. This work provides a promising route for practical high-precision quantum metrology and quantum sensing.

quant-ph↗
arXiv2026

Dynamic Latent Routing

We investigate the temporal concatenation of sub-policies in Markov Decision Processes (MDP) with time-varying reward functions. We introduce General Dijkstra Search (GDS), and prove that globally optimal goal-reaching policies can be recovered through temporal composition of intermediate optimal sub-policies. Motivated by the "search, select, update" principle underlying GDS, we propose Dynamic Latent Routing (DLR), a language-model post-training method that jointly learns discrete latent codes, routing policies, and model parameters through dynamic search in a single training stage. In low-data fine-tuning settings, DLR matches or outperforms supervised fine-tuning across four datasets and six models, achieving a mean gain of +6.6 percentage points, while prior discrete-latent baselines consistently underperform SFT. Mechanistic analyses and targeted code ablations show that DLR learns structured routing behaviors with distinct causal roles.

cs.LG↗
arXiv2026

Reliable Narrowband Interference Detection via Backward Conformal Prediction

Narrowband interference can severely degrade the performance of WiFi links by concentrating significant power on a small portion of the channel. Machine learning (ML) detectors trained on baseband I/Q samples can identify the affected subcarriers with high accuracy, surpassing model-based detectors that rely on hand-crafted statistics. The predictive probabilities produced by such detectors are, however, typically poorly calibrated, and downstream mitigation modules generally operate under strict resource budgets that limit the number of candidate interference states that can be acted upon. Conformal prediction (CP) provides a distribution-free framework for constructing prediction sets that control the probability of excluding the true output, i.e., the miscoverage level, at a prescribed level. However, this target miscoverage level must be fixed in advance, while the resulting prediction-set size remains uncontrolled, which is misaligned with operationally constrained settings. To address this issue, we develop a backward conformal prediction (BCP) framework in which the prediction-set size is fixed by the operational budget and the corresponding per-input miscoverage level is estimated from calibration data with provable reliability guarantees. We instantiate the framework for narrowband interference detection in WiFi systems and show through simulations that BCP yields reliable miscoverage estimates whose accuracy approaches that of an uncalibrated baseline as the calibration set grows.

eess.SP↗
arXiv2026

Geometry-Aware CLIP Retrieval via Local Cross-Modal Alignment and Steering

CLIP retrieval is typically framed as a pointwise similarity problem in a shared embedding space. While CLIP achieves strong global cross-modal alignment, many retrieval failures arise from local geometric inconsistencies: nearby items are incorrectly ordered, leading to systematic confusions (e.g., pentagon vs. hexagon) and produces diffuse, weakly controlled result sets. Prior work largely optimizes for point wise relevance or finetuning to mitigate these problems. We instead view retrieval as a problem of neighborhood alignment. Our work introduces (1) neighborhood-level re-ranking via Hungarian matching, which rewards structural consistency; (2) query-conditioned local steering, where directions derived from contrastive neighborhoods around the query reshape retrieval. We show that these techniques improve retrieval performance on attribute-binding and compositional retrieval tasks. Together, these methods operate on local neighborhoods but serve different roles: re-ranking rewards alignment whereas local steering controls neighborhood structure. This shows that retrieval quality and controllability depend critically on local structure, which can be exploited at inference time without retraining.

cs.CV↗
arXiv2026

IOTEL: A Tool for Generating IoT-enriched Object-Centric Event Logs

Integrating Internet of Things (IoT) data with business process event logs is crucial for analysing IoT-enhanced processes, yet remains challenging due to differences in abstraction levels and the separation of data sources. Simply incorporating raw IoT data increases the size and complexity of the resulting log, often requiring additional processing before process analysis can be performed. While tools for generating IoT-enriched event logs exist, they either rely on specialised schemas or focus on extracting event logs from sensor data, offering limited support for integrating process-relevant IoT data into existing event logs. To address this gap, we present IOTEL, a tool for systematically generating IoT-enriched object-centric event logs (OCEL). By building on the OCEL schema, IOTEL enables structured IoT data integration compatible with existing process mining tools. It support practitioners and researchers in analysing IoT-enhanced business processes, as demonstrated in a real-world scenario. A video demonstrating the tool is available online.

cs.SE↗
arXiv2026

Scaling Knowledge Graph Construction through Synthetic Data Generation and Distillation

Document-level knowledge graph (KG) construction faces a fundamental scaling challenge: existing methods either rely on expensive large language models (LLMs), making them economically nonviable for large-scale corpora, or employ smaller models that produce incomplete and inconsistent graphs. We find that this limitation stems not from model capabilities but from insufficient training on high-quality document-level KG data. To address this gap, we introduce SynthKG, a multi-step data synthesis pipeline that generates high-quality document-KG pairs through systematic chunking, decontextualization, and structured extraction using LLMs. By fine-tuning a smaller LLM on synthesized document-KG pairs, we streamline the multi-step process into a single-step KG generation approach called Distill-SynthKG. Furthermore, we repurpose existing question-answering datasets to construct KG evaluation datasets and introduce new evaluation metrics. Using KGs produced by Distill-SynthKG, we also design a novel graph-based retrieval framework for RAG. Experimental results demonstrate that Distill-SynthKG not only surpasses all baseline models in KG quality (including models up to eight times larger) but also consistently improves in retrieval and question-answering tasks. Additionally, our proposed graph retrieval framework outperforms all KG-retrieval methods across multiple benchmark datasets.

cs.CL↗
arXiv2026

WARM-CAT: Warm-Started Test-Time Comprehensive Knowledge Accumulation for Compositional Zero-Shot Learning

Compositional Zero-Shot Learning (CZSL) aims to recognize novel attribute-object compositions based on the knowledge learned from seen ones. Existing methods suffer from performance degradation caused by the distribution shift of label space at test time, which stems from the inclusion of unseen compositions recombined from attributes and objects. To overcome the challenge, we propose a novel approach that accumulates comprehensive knowledge in both textual and visual modalities from unsupervised data to update multimodal prototypes at test time. Building on this, we further design an adaptive update weight to control the degree of prototype adjustment, enabling the model to flexibly adapt to distribution shift during testing. Moreover, a dynamic priority queue is introduced that stores high-confidence images to acquire visual prototypes from historical images for inference. Since the model tends to favor compositions already stored in the queue during testing, we warm-start the queue by initializing it with training images for visual prototypes of seen compositions and generating unseen visual prototypes using the mapping learned between seen and unseen textual prototypes. Considering the semantic consistency of multimodal knowledge, we align textual and visual prototypes by multimodal collaborative representation learning. To provide a more reliable evaluation for CZSL, we introduce a new benchmark dataset, C-Fashion, and refine the widely used but noisy MIT-States dataset. Extensive experiments indicate that our approach achieves state-of-the-art performance on four benchmark datasets under both closed-world and open-world settings. The source code and datasets are available at https://github.com/xud-yan/WARM-CAT .

cs.CV↗