Searcharxiv⌕ Search

arXiv subjects

Feng Luo

Publications and source records attributed to Feng Luo.

At least 37 records · Page 2Linked to original sources

Demystifying OPD: Length Inflation and Stabilization Strategies for Large Language Models

On-policy distillation (OPD) trains student models under their own induced distribution while leveraging supervision from stronger teachers. We identify a failure mode of OPD: as training progresses, on-policy rollouts can undergo abrupt length inflation, causing truncated trajectories to dominate the training data. This truncation collapse coincides with abrupt repetition saturation and induces biased gradient signals, leading to severe training instability and sharp degradation in validation performance. We attribute this problem to the interaction between student-induced data collection and the distillation objective, which implicitly favors long and repetitive rollouts. To address this issue, we propose StableOPD, a stabilized OPD framework that combines a reference-based divergence constraint with rollout mixture distillation. These together mitigate repetition-induced length inflation and further stabilize OPD training. Across multiple math reasoning datasets, our approach prevents truncation collapse, stabilizes training dynamics, and improves performance by 7.2% on average.

cs.CL↗

MolLangBench: A Comprehensive Benchmark for Language-Prompted Molecular Structure Recognition, Editing, and Generation

Precise recognition, editing, and generation of molecules are essential prerequisites for both chemists and AI systems tackling various chemical tasks. We present MolLangBench, a comprehensive benchmark designed to evaluate fundamental molecule-language interface tasks: language-prompted molecular structure recognition, editing, and generation. To ensure high-quality, unambiguous, and deterministic outputs, we construct the recognition tasks using automated cheminformatics tools, and curate editing and generation tasks through rigorous expert annotation and validation. MolLangBench supports the evaluation of models that interface language with different molecular representations, including linear strings, molecular images, and molecular graphs. Evaluations of state-of-the-art models reveal significant limitations: the strongest model (GPT-5) achieves $86.2\%$ and $85.5\%$ accuracy on recognition and editing tasks, which are intuitively simple for humans, and performs even worse on the generation task, reaching only $43.0\%$ accuracy. These results highlight the shortcomings of current AI systems in handling even preliminary molecular recognition and manipulation tasks. We hope MolLangBench will catalyze further research toward more effective and reliable AI systems for chemical applications.The dataset and code can be accessed at https://huggingface.co/datasets/ChemFM/MolLangBench and https://github.com/TheLuoFengLab/MolLangBench, respectively.

cs.CL↗

DisPatch: Disarming Adversarial Patches in Object Detection with Diffusion Models

Object detection is fundamental to various real-world applications, such as security monitoring and surveillance video analysis. Despite their advancements, state-of-the-art object detectors are still vulnerable to adversarial patch attacks, which can be easily applied to real-world objects to either conceal actual items or create non-existent ones, leading to severe consequences. In this work, we introduce DisPatch, the first diffusion-based defense framework for object detection. Unlike previous works that aim to "detect and remove" adversarial patches, DisPatch adopts a "regenerate and rectify" strategy, leveraging generative models to disarm attack effects while preserving the integrity of the input image. Specifically, we utilize the in-distribution generative power of diffusion models to regenerate the entire image, aligning it with benign data. A rectification process is then employed to identify and replace adversarial regions with their regenerated benign counterparts. DisPatch is attack-agnostic and requires no prior knowledge of the existing patches. Extensive experiments across multiple detectors demonstrate that DisPatch consistently outperforms state-of-the-art defenses on both hiding attacks and creating attacks, achieving the best overall mAP@0.5 score of 89.3% on hiding attacks, and lowering the attack success rate to 24.8% on untargeted creating attacks. Moreover, it strikes the balance between effectiveness and efficiency, and maintains strong robustness against adaptive attacks, making it a practical and reliable defense method.

cs.CV↗

Climate Prompting: Generating the Madden-Julian Oscillation using Video Diffusion and Low-Dimensional Conditioning

Generative Deep Learning is a powerful tool for modeling of the Madden-Julian oscillation (MJO) in the tropics, yet its relationship to traditional theoretical frameworks remains poorly understood. Here we propose a video diffusion model, trained on atmospheric reanalysis, to synthetize long MJO sequences conditioned on key low-dimensional metrics. The generated MJOs capture key features including composites, power spectra and multiscale structures including convectively coupled waves, despite some bias. We then prompt the model to generate more tractable MJOs based on intentionally idealized low-dimensional conditionings, for example a perpetual MJO, an isolated modulation by seasons and/or the El Nino-Southern Oscillation, and so on. This enables deconstructing the underlying processes and identifying physical drivers. The present approach provides a practical framework for bridging the gap between low-dimensional MJO theory and high-resolution atmospheric complexity and will help tropical atmosphere prediction.

cs.CV↗

Locality Sensitive Hashing in Hyperbolic Space

For a metric space $(X, d)$, a family $\mathcal{H}$ of locality sensitive hash functions is called $(r, cr, p_1, p_2)$ sensitive if a randomly chosen function $h\in \mathcal{H}$ has probability at least $p_1$ (at most $p_2$) to map any $a, b\in X$ in the same hash bucket if $d(a, b)\leq r$ (or $d(a, b)\geq cr$). Locality Sensitive Hashing (LSH) is one of the most popular techniques for approximate nearest-neighbor search in high-dimensional spaces, and has been studied extensively for Hamming, Euclidean, and spherical geometries. An $(r, cr, p_1, p_2)$-sensitive hash function enables approximate nearest neighbor search (i.e., returning a point within distance $cr$ from a query $q$ if there exists a point within distance $r$ from $q$) with space $O(n^{1+ρ})$ and query time $O(n^ρ)$ where $ρ=\frac{\log 1/p_1}{\log 1/p_2}$. But LSH for hyperbolic spaces $\mathbb{H}^d$ remains largely unexplored. In this work, we present the first LSH construction native to hyperbolic space. For the hyperbolic plane $(d=2)$, we show a construction achieving $ρ\leq 1/c$, based on the hyperplane rounding scheme. For general hyperbolic spaces $(d \geq 3)$, we use dimension reduction from $\mathbb{H}^d$ to $\mathbb{H}^2$ and the 2D hyperbolic LSH to get $ρ\leq 1.59/c$. On the lower bound side, we show that the lower bound on $ρ$ of Euclidean LSH extends to the hyperbolic setting via local isometry, therefore giving $ρ\geq 1/c^2$.

cs.CG↗

AgenticScholar: Agentic Data Management with Pipeline Orchestration for Scholarly Corpora

Managing the rapidly growing scholarly corpus poses significant challenges in representation, reasoning, and efficient analysis. An ideal system should unify structured knowledge management, agentic planning, and interpretable execution to support diverse scholarly queries - from retrieval to knowledge discovery and generation - at scale. Unfortunately, existing RAG and document analytics systems fail to achieve all query types simultaneously. To this end, we propose AgenticScholar, an agentic scholarly data management system that integrates a structure-aware knowledge representation layer, an LLM-centric hybrid query planning layer, and a unified execution layer with composable operators. AgenticScholar autonomously translates natural language queries into executable DAG plans, enabling end-to-end reasoning over multi-modal scholarly data. Extensive experiments demonstrate that AgenticScholar significantly outperforms existing systems in effectiveness, efficiency, and interpretability, offering a practical foundation for future research on agentic scholarly data management.

cs.DB↗

Decomposition-Driven Multi-Table Retrieval and Reasoning for Numerical Question Answering

In this paper, we study the problem of numerical multi-table question answering (MTQA) over large-scale table collections (e.g., online data repositories). This task is essential in many analytical applications. Existing MTQA solutions, such as text-to-SQL or open-domain MTQA methods, are designed for databases and struggle when applied to large-scale table collections. The key limitations include: (1) Limited support for complex table relationships; (2) Ineffective retrieval of relevant tables at scale; (3) Inaccurate answer generation. To overcome these limitations, we propose DMRAL, a Decomposition-driven Multi-table Retrieval and Answering framework for MTQA over large-scale table collections, which consists of: (1) constructing a table relationship graph to capture complex relationships among tables; (2) Table-Aligned Question Decomposer and Coverage-Aware Retriever, which jointly enable the effective identification of relevant tables from large-scale corpora by enhancing the question decomposition quality and maximizing the question coverage of retrieved tables; and (3) Sub-question Guided Reasoner, which produces correct answers by progressively generating and refining the reasoning program based on sub-questions. Experiments on two MTQA datasets demonstrate that DMRAL significantly outperforms existing state-of-the-art MTQA methods, with an average improvement of 24% in table retrieval and 55% in answer accuracy.

cs.DB↗

A Rigidity Theorem for Convex Sets in Hyperbolic 3-Space

Pogorelov's rigidity theorem states that a compact convex body in the hyperbolic 3-space is determined up to isometry by the intrinsic path metric on its boundary. The main result of this paper addresses a rigidity problem for non-compact closed convex 3-dimensional subsets in hyperbolic 3-space. We show that the intrinsic path metric on the boundary determines a closed convex set up to isometry, provided that the set of limit points of the convex set at infinity of the hyperbolic 3-space has vanishing 1-dimensional Hausdorff measure, i.e., zero length. Furthermore, this zero-length condition is optimal. This can be considered as an analogue of the Painlevé removability theorem in complex analysis, which states that sets of zero length are removable for bounded holomorphic functions. As a corollary, we show that if the underlying complex structure of a connected polyhedral surface is of parabolic type, then it is discrete conformal, unique up to scaling, to a complete flat surface marked with a discrete subset. The proof uses Pogorelov's rigidity theorem for compact convex bodies in $\mathbb R^3$, the Pogorelov map, and the Tabor--Tabor theorem on the extension of locally convex functions.

math.GT↗

DTS: Enhancing Large Reasoning Models via Decoding Tree Sketching

Large Reasoning Models (LRMs) achieve remarkable inference-time improvements through parallel thinking. However, existing methods rely on redundant sampling of reasoning trajectories, failing to effectively explore the reasoning space to uncover high-quality solutions. To address these limitations, we propose Decoding Tree Sketching (DTS), a plug-and-play decoding framework for structural multi-trajectory exploration and reasoning selection. For reasoning exploration, DTS sketches a backbone tree of the reasoning space by selectively branching at decision tokens. For reasoning selection, guided by length-accuracy anti-correlation, DTS designs an early termination to prioritize short and reliable trajectories during decoding. Experimental results across four LRMs and datasets demonstrate that DTS significantly enhances accuracy by 14% and reduces repetitive generation by 8% on average. Notably, DTS enables smaller models to outperform larger models with 10$\times$ the size, highlighting its potential to strengthen reasoning capabilities.

cs.AI↗

Bonnet: Ultra-fast whole-body bone segmentation from CT scans

This work proposes Bonnet, an ultra-fast sparse-volume pipeline for whole-body bone segmentation from CT scans. Accurate bone segmentation is important for surgical planning and anatomical analysis, but existing 3D voxel-based models such as nnU-Net and STU-Net require heavy computation and often take several minutes per scan, which limits time-critical use. The proposed Bonnet addresses this by integrating a series of novel framework components including HU-based bone thresholding, patch-wise inference with a sparse spconv-based U-Net, and multi-window fusion into a full-volume prediction. Trained on TotalSegmentator and evaluated without additional tuning on RibSeg, CT-Pelvic1K, and CT-Spine1K, Bonnet achieves high Dice across ribs, pelvis, and spine while running in only 2.69 seconds per scan on an RTX A6000. Compared to strong voxel baselines, Bonnet attains a similar accuracy but reduces inference time by roughly 25x on the same hardware and tiling setup. The toolkit and pre-trained models will be released at https://github.com/HINTLab/Bonnet.

eess.IV↗

AutoL2S: Auto Long-Short Reasoning for Efficient Large Language Models

Reasoning-capable large language models (LLMs) achieve strong performance on complex tasks but often exhibit overthinking after distillation, generating unnecessarily long chain-of-thought (CoT) reasoning even for simple inputs and incurring high inference cost. However, naively shortening reasoning length can degrade reasoning accuracy, as concise reasoning may be insufficient for certain inputs and lacks explicit supervision. We propose Auto Long-Short Reasoning (AutoL2S), a distillation framework that empowers non-reasoning LLMs to think thoroughly but only when necessary. AutoL2S first learns a lightweight switching token with verified long-short CoTs to enable instance-wise long-short reasoning selection. Then it leverages long-short reasoning rollouts induced by a switching token in a GRPO-style loss to improve reasoning efficiency while maintaining accuracy. Experiments demonstrate that AutoL2S effectively reduces reasoning length up to 71% with minimal accuracy loss, yielding markedly better trade-off in token length and inference time while preserving accuracy.

cs.CL↗

ShadowGS: Shadow-Aware 3D Gaussian Splatting for Satellite Imagery

3D Gaussian Splatting (3DGS) has emerged as a novel paradigm for 3D reconstruction from satellite imagery. However, in multi-temporal satellite images, prevalent shadows exhibit significant inconsistencies due to varying illumination conditions. To address this, we propose ShadowGS, a novel framework based on 3DGS. It leverages a physics-based rendering equation from remote sensing, combined with an efficient ray marching technique, to precisely model geometrically consistent shadows while maintaining efficient rendering. Additionally, it effectively disentangles different illumination components and apparent attributes in the scene. Furthermore, we introduce a shadow consistency constraint that significantly enhances the geometric accuracy of 3D reconstruction. We also incorporate a novel shadow map prior to improve performance with sparse-view inputs. Extensive experiments demonstrate that ShadowGS outperforms current state-of-the-art methods in shadow decoupling accuracy, 3D reconstruction precision, and novel view synthesis quality, with only a few minutes of training. ShadowGS exhibits robust performance across various settings, including RGB, pansharpened, and sparse-view satellite inputs.

cs.CV↗

ChemFM as a Scaling Law Guided Foundation Model Pre-trained on Informative Chemicals

Traditional AI methods often rely on task-specific model designs and training, which constrain both the scalability of model size and generalization across different tasks. Here, we introduce ChemFM, a large foundation model specifically developed for chemicals. By conducting a series of scaling experiments, we identify UniChem as the informative molecular database for pre-training the foundation model. ChemFM comprises 3 billion parameters and is pre-trained on 178 million molecules using self-supervised causal language modeling to extract generalizable molecular representations. This model can be adapted to diverse downstream chemical applications using either full-parameter or parameter-efficient fine-tuning methods. ChemFM consistently outperforms state-of-the-art task-specific AI models across all tested tasks. Notably, it achieves up to 67.48% performance improvement across 34 property prediction benchmarks, up to 33.80% reduction in mean average deviation between conditioned and actual properties of generated molecules in conditional molecular generation tasks, and up to 3.7% top-1 accuracy improvement across 4 reaction prediction datasets. Moreover, ChemFM demonstrates its superior performance in predicting antibiotic activity and cytotoxicity, highlighting its potential to advance the discovery of novel antibiotics. Furthermore, we demonstrate that, as a foundation model, ChemFM exhibits strong data efficiency, requiring significantly fewer labeled training samples to achieve state-of-the-art performance. We anticipate that ChemFM will significantly advance chemistry research by providing a foundation model capable of effectively generalizing across a broad range of tasks with minimal additional training.

cs.CE↗

Johnson-Lindenstrauss Lemma Beyond Euclidean Geometry

The Johnson-Lindenstrauss (JL) lemma is a cornerstone of dimensionality reduction in Euclidean space, but its applicability to non-Euclidean data has remained limited. This paper extends the JL lemma beyond Euclidean geometry to handle general dissimilarity matrices that are prevalent in real-world applications. We present two complementary approaches: First, we show the JL transform can be applied to vectors in pseudo-Euclidean space with signature $(p,q)$, providing theoretical guarantees that depend on the ratio of the $(p, q)$ norm and Euclidean norm of two vectors, measuring the deviation from Euclidean geometry. Second, we prove that any symmetric hollow dissimilarity matrix can be represented as a matrix of generalized power distances, with an additional parameter representing the uncertainty level within the data. In this representation, applying the JL transform yields multiplicative approximation with a controlled additive error term proportional to the deviation from Euclidean geometry. Our theoretical results provide fine-grained performance analysis based on the degree to which the input data deviates from Euclidean geometry, making practical and meaningful reduction in dimensionality accessible to a wider class of data. We validate our approaches on both synthetic and real-world datasets, demonstrating the effectiveness of extending the JL lemma to non-Euclidean settings.

cs.DS↗

FairAgent: Democratizing Fairness-Aware Machine Learning with LLM-Powered Agents

Training fair and unbiased machine learning models is crucial for high-stakes applications, yet it presents significant challenges. Effective bias mitigation requires deep expertise in fairness definitions, metrics, data preprocessing, and machine learning techniques. In addition, the complex process of balancing model performance with fairness requirements while properly handling sensitive attributes makes fairness-aware model development inaccessible to many practitioners. To address these challenges, we introduce FairAgent, an LLM-powered automated system that significantly simplifies fairness-aware model development. FairAgent eliminates the need for deep technical expertise by automatically analyzing datasets for potential biases, handling data preprocessing and feature engineering, and implementing appropriate bias mitigation strategies based on user requirements. Our experiments demonstrate that FairAgent achieves significant performance improvements while significantly reducing development time and expertise requirements, making fairness-aware machine learning more accessible to practitioners.

cs.LG↗

Integrating Rules and Semantics for LLM-Based C-to-Rust Translation

Automated translation of legacy C code into Rust aims to ensure memory safety while reducing the burden of manual migration. Early approaches in code translation rely on static rule-based methods, but they suffer from limited coverage due to dependence on predefined rule patterns. Recent works regard the task as a sequence-to-sequence problem by leveraging large language models (LLMs). Although these LLM-based methods are capable of reducing unsafe code blocks, the translated code often exhibits issues in following Rust rules and maintaining semantic consistency. On one hand, existing methods adopt a direct prompting strategy to translate the C code, which struggles to accommodate the syntactic rules between C and Rust. On the other hand, this strategy makes it difficult for LLMs to accurately capture the semantics of complex code. To address these challenges, we propose IRENE, an LLM-based framework that Integrates RulEs aNd sEmantics to enhance translation. IRENE consists of three modules: 1) a rule-augmented retrieval module that selects relevant translation examples based on rules generated from a static analyzer developed by us, thereby improving the handling of Rust rules; 2) a structured summarization module that produces a structured summary for guiding LLMs to enhance the semantic understanding of C code; 3) an error-driven translation module that leverages compiler diagnostics to iteratively refine translations. We evaluate IRENE on two datasets (xCodeEval, a public dataset, and HW-Bench, an industrial dataset provided by Huawei) and eight LLMs, focusing on translation accuracy and safety.

cs.SE↗

A Time-Triggered Communication Method Based on Urgency-Based Scheduler in Time-Sensitive Networking

The development of the automotive industry and automation has led to a growing demand for time-critical systems to have low latency and jitter for critical traffic. To address this issue, the IEEE 802.1 Time-Sensitive Networking (TSN) task group proposed the Time-Aware Shaper (TAS) to implement Time-Triggered (TT) communication, enabling deterministic transmission by assigning specific time windows to each stream. While Fixed Routing and Waiting-Allowed (FR-WA) scheduling algorithms offer flexibility, they suffer from inefficiencies in solution time and scalability. This study analyzes TAS implementation challenges, emphasizing how network scale expansion increases computational constraints. We propose an Urgency-Based Scheduler method (TT-UBS) to address these limitations to enhance deterministic transmission and computational efficiency under anomalies. A novel scheduling algorithm for TT-UBS parameter determination is developed, alongside simulations and comparative evaluations. Results show that TT-UBS guarantees deterministic traffic delivery while reducing solution time by 98.22% in test scenarios compared to traditional approaches. The methodology is extended to other scheduling algorithms to assess efficiency improvements. This advancement supports TSN's application in mission-critical systems by optimizing time-triggered communication performance and enabling reliable network deployment. The framework demonstrates significant potential for real-time in-vehicle networks requiring latency and jitter control.

cs.NI↗

Ultrafast and reliable domain-wall and skyrmion logic in a chirally coupled ferrimagnet

Unlocking the spin degree of freedom in addition to the electron's charge, spin-based logic offers an in-memory computing architecture beyond-CMOS technology. Here, we encode information into chiral spin textures (e.g., chiral domain-wall and skyrmion) and achieve an ultrafast and reliable all-electrical logic by exploiting the Dzyaloshinskii-Moriya interaction-induced chiral coupling. Taking advantage of fast spin dynamics in antiferromagnetically coupled systems, we achieved a fast domain-wall motion passing through the logic gate, exceeding 1 kilometre per second, yielding an operation time of 50 picoseconds for a 50 nanometres-long logic gate. Furthermore, we present a fast logic operation with skyrmion bubbles in a racetrack that exhibits a topologically protected computation scheme. Our work demonstrates a viable approach for advanced microchips with high operation frequency and ultralow power consumption, paving the way for next-generation computing technologies.

cond-mat.mes-hall↗