arXiv subjects

Hanlin Xu

Publications and source records attributed to Hanlin Xu.

11 recordsLinked to original sources

arXiv2026

MATCH: Modulating Attention via In-Context Retrieval for Long-Context Transformers

The quadratic computational cost of traditional attention mechanisms poses a major bottleneck to the scalability and practical deployment of large language models (LLMs), particularly in long-context scenarios. To improve efficiency, existing approaches often enforce rigid structural constraints such as local attention windows. However, these strategies typically lead to substantial performance degradation on tasks requiring precise long-range recall. In this work, we propose MATCH, a scalable and efficient framework that augments sparsified attention mechanisms with dynamically integrated in-context information through an efficient retrieval system. Empirical results show that MATCH significantly improves the performance of sparse-attention models on both synthetic and real-world natural-language tasks. These findings highlight the versatility of MATCH as a general approach for enhancing in-context retrieval capabilities while maintaining the efficiency benefits of sparse attention architectures.

cs.CL↗

arXiv2026

Domino: Decoupling Causal Modeling from Autoregressive Drafting in Speculative Decoding

Speculative decoding accelerates LLM inference by drafting multiple tokens and verifying them in parallel with the target model. However, its practical speedup is constrained by the trade-off between draft quality and drafting cost: autoregressive drafters model causal dependencies among draft tokens but incur sequential overhead, while parallel drafters reduce drafting cost but weaken intra-block dependency modeling. In this paper, we propose Domino, a speculative decoding framework that decouples causal dependency modeling from expensive autoregressive draft execution. Domino first uses a parallel draft backbone to produce preliminary draft distributions for the entire block, and then applies a lightweight Domino head to refine them with prefix-dependent causal information. To stabilize teacher-forced causal encoding, we further introduce a base-anchored training curriculum that first strengthens the parallel backbone and then gradually shifts optimization toward the causally corrected final distribution. Experiments on Qwen3 models show that Domino achieves up to $5.49\times$ end-to-end speedup under the Transformers backend and up to $5.8\times$ throughput speedup under SGLang serving.

cs.CL↗

arXiv2026

MacNeille completions of parabolic quotients

Alternating sign matrices (ASMs) arise as the Dedekind-MacNeille completion of the Bruhat order on the symmetric group. They enjoy fruitful combinatorial and geometric properties, with a particularly rich history on enumerations and bijections. In this paper, we explicitly describe the Dedekind-MacNeille completion of the Bruhat order on any parabolic quotients of the symmetric group. It is naturally a subposet of the alternating sign matrices, with different lattice operations. Moreover, we demonstrate the relations between the meet and join operations in this lattice with taking unions and intersections of the corresponding ASM varieties, respectively. Finally, we conclude with a more detailed discussion of special cases.

math.CO↗

arXiv2026

Bridging Visual Representation and Reinforcement Learning from Verifiable Rewards in Large Vision-Language Models

Reinforcement Learning from Verifiable Rewards (RLVR) has substantially enhanced the reasoning capabilities of large language models in abstract reasoning tasks. However, its application to Large Vision-Language Models (LVLMs) remains constrained by a structural representational bottleneck. Existing approaches generally lack explicit modeling and effective utilization of visual information, preventing visual representations from being tightly coupled with the reinforcement learning optimization process and thereby limiting further improvements in multimodal reasoning performance. To address this limitation, we propose KAWHI (Key-Region Aligned Weighted Harmonic Incentive), a plug-and-play reward reweighting mechanism that explicitly incorporates structured visual information into uniform reward policy optimization methods (e.g., GRPO and GSPO). The method adaptively localizes semantically salient regions through hierarchical geometric aggregation, identifies vision-critical attention heads via structured attribution, and performs paragraph-level credit reallocation to align spatial visual evidence with semantically decisive reasoning steps. Extensive empirical evaluations on diverse reasoning benchmarks substantiate KAWHI as a general-purpose enhancement module, consistently improving the performance of various uniform reward optimization methods. Project page: KAWHI (https://kawhiiiileo.github.io/KAWHI_PAGE/)

cs.CV↗

arXiv2025

Diffusion LLM with Native Variable Generation Lengths: Let [EOS] Lead the Way

Diffusion-based large language models (dLLMs) have exhibited substantial potential for parallel text generation, which may enable more efficient generation compared to autoregressive models. However, current dLLMs suffer from fixed generation lengths, which indicates the generation lengths of dLLMs have to be determined before decoding as a hyper-parameter, leading to issues in efficiency and flexibility. To solve these problems, in this work, we propose to train a diffusion LLM with native variable generation lengths, abbreviated as dLLM-Var. Concretely, we aim to train a model to accurately predict the [EOS] token in the generated text, which makes a dLLM be able to natively infer in a block diffusion manner, while still maintaining the ability of global bi-directional (full) attention and high parallelism. Experiments on standard benchmarks demonstrate that our method achieves a 30.1x speedup over traditional dLLM inference paradigms and a 2.4x speedup relative to autoregressive models such as Qwen and Llama. Our method achieves higher accuracy and faster inference, elevating dLLMs beyond mere academic novelty and supporting their practical use in real-world applications. Codes and models have been released.

cs.CL↗

arXiv2025

Self Speculative Decoding for Diffusion Large Language Models

Diffusion-based Large Language Models (dLLMs) have emerged as a competitive alternative to autoregressive models, offering unique advantages through bidirectional attention and parallel generation paradigms. However, the generation results of current parallel decoding methods deviate from stepwise decoding, introducing potential performance degradation, which limits their practical deployment. To address this problem, we propose \textbf{S}elf \textbf{S}peculative \textbf{D}ecoding (SSD), a lossless inference acceleration method that leverages the dLLM itself as both speculative decoding drafter and verifier without auxiliary modules. SSD introduces a self-drafting mechanism where the model generates predictions for multiple positions, then verifies them through hierarchical verification trees in a single forward pass. Unlike traditional speculative decoding that requires separate draft models, SSD eliminates model redundancy and memory overhead by exploiting the dLLM's inherent parallel prediction capability for multiple positions. This self-speculative approach allows the model to progressively verify and accept multiple tokens in a single forward pass. Our experiments demonstrate that SSD achieves up to 3.46$\times$ speedup while keeping the output identical to stepwise decoding on open source models such as LLaDA and Dream. Code will be made publicly available on GitHub.

cs.CL↗

arXiv2025

GLASS: Global-Local Aggregation for Inference-time Sparsification of LLMs

Inference-time sparsification is a promising path to deploy large language models (LLMs) on resource-constrained devices, yet existing training-free methods typically estimate feedforward network (FFN) neuron importance from the input prompt alone. We show this prompt-only signal is often unreliable, especially for short prompts and long-form decoding, leading to inaccurate masks and degraded generation fidelity. We propose GLASS, a plug-and-play, training-free framework that stabilizes dynamic FFN pruning by aggregating two complementary views of neuron criticality: local prompt-specific activations and a global model-intrinsic prior. GLASS fuses global and local signals via rank aggregation, yielding robust critical-neuron selection even when the prompt is short. We interpret GLASS as the maximum-a-posteriori consensus ranking under a permutation-based probabilistic model, providing a principled foundation for its weighted rank-aggregation rule. We apply GLASS to a diverse set of open-source LLMs, and show that it yields substantial improvements over prior training-free baselines in the challenging short-prompt, long-generation scenarios, achieving up to 45.10% lower perplexity and 25.73% lower KL divergence, while delivering significant on-device decoding speedup.

cs.LG↗

arXiv2024

Engineering Carbon Credits Towards A Responsible FinTech Era: The Practices, Implications, and Future

Carbon emissions significantly contribute to climate change, and carbon credits have emerged as a key tool for mitigating environmental damage and helping organizations manage their carbon footprint. Despite their growing importance across sectors, fully leveraging carbon credits remains challenging. This study explores engineering practices and fintech solutions to enhance carbon emission management. We first review the negative impacts of carbon emission non-disclosure, revealing its adverse effects on financial stability and market value. Organizations are encouraged to actively manage emissions and disclose relevant data to mitigate risks. Next, we analyze factors influencing carbon prices and review advanced prediction algorithms that optimize carbon credit purchasing strategies, reducing costs and improving efficiency. Additionally, we examine corporate carbon emission prediction models, which offer accurate performance assessments and aid in planning future carbon credit needs. By integrating carbon price and emission predictions, we propose research directions, including corporate carbon management cost forecasting. This study provides a foundation for future quantitative research on the financial and market impacts of carbon management practices and is the first systematic review focusing on computing solutions and engineering practices for carbon credits.

cs.CY↗

arXiv2024

Scalar sector of Type 2 Seesaw model explorations with multi-lepton final states

Originally motivated for the generation of (Majorana) neutrino masses, the Type 2 Seesaw Model has also a rich extended Higgs sector with, if accessible at the LHC, a distinctive phenomenology of neutral, charged and doubly-charged states. The goal of the work is to present an exhaustive phenomenological study of the most promising production and decay channels of pair or associated scalars, decaying directly or in cascades to Standard Model particles at the LHC. The study is complementary to the literature in that it highlights a previously unnoticed important sensitivity to a mixing angle. The ensuing uncertainty calls for a comprehensive experimental search strategy for the various processes. These processes can be studied within LHC energies reach, by comparing cutflow results for different final states. We carried out prospective search analyses with multi-lepton, jets and missing energy configurations, assuming an ATLAS-like detector at LHC and HL-LHC, for charged, doubly-charged, and for the first time neutral scalar productions. The work is a collaboration between ATLAS experimentalists and theoreticians in continuation of an endeavor that lead to previous published ATLAS analyses for the search of (doubly)charged Higgs bosons, aiming at proposals for future experimental searches.

hep-ph↗

arXiv2024

Type-II Seesaw Higgs triplet productions and decays at the LHC

The Type-II Seesaw Model provides an attractive scenario to account for Majorana-neutrino masses. Its extended Higgs sector, if sufficiently light, can have a rich and distinctive phenomenology at the LHC while yielding automatically an essentially Standard-Model-Higgs-like state. Several phenomenological studies have been devoted to the scalar sector of this model, as well as experimental searches focusing mostly on the (doubly-)charged states. In this paper we present an exhaustive study of the main production and decay channels of all the non-standard scalar states originating from the $SU(2)_L$ doublet and a complex triplet of the model. We stick to scenarios where lepton-number-violating decays are suppressed, for which present experimental limits are still weak, highlighting theoretical parameter sensitivities that were not previously emphasized in the literature and the uncertainties they can induce for the experimental searches at the LHC. A comprehensive classification of the various cascade decays and corresponding Standard Model particle multiplicities is provided. As an illustration, a detailed prospective search study at the LHC with an ATLAS-like detector is carried out on some benchmark points, for charged, doubly-charged, and, for the first time, neutral state productions

hep-ph↗

arXiv2024

OAKINK2: A Dataset of Bimanual Hands-Object Manipulation in Complex Task Completion

We present OAKINK2, a dataset of bimanual object manipulation tasks for complex daily activities. In pursuit of constructing the complex tasks into a structured representation, OAKINK2 introduces three level of abstraction to organize the manipulation tasks: Affordance, Primitive Task, and Complex Task. OAKINK2 features on an object-centric perspective for decoding the complex tasks, treating them as a sequence of object affordance fulfillment. The first level, Affordance, outlines the functionalities that objects in the scene can afford, the second level, Primitive Task, describes the minimal interaction units that humans interact with the object to achieve its affordance, and the third level, Complex Task, illustrates how Primitive Tasks are composed and interdependent. OAKINK2 dataset provides multi-view image streams and precise pose annotations for the human body, hands and various interacting objects. This extensive collection supports applications such as interaction reconstruction and motion synthesis. Based on the 3-level abstraction of OAKINK2, we explore a task-oriented framework for Complex Task Completion (CTC). CTC aims to generate a sequence of bimanual manipulation to achieve task objectives. Within the CTC framework, we employ Large Language Models (LLMs) to decompose the complex task objectives into sequences of Primitive Tasks and have developed a Motion Fulfillment Model that generates bimanual hand motion for each Primitive Task. OAKINK2 datasets and models are available at https://oakink.net/v2.

cs.CV↗