• 学术搜索
  • 科研智能体
    • Research Labs
    • AI 阅读
    • AI 文库
    • 深度研究
    • 学者亮点
  • 学术资源
    • AI2000
    • 期刊/会议
    • 学者库
    • 学术API
      MCP&Skill
    • 溯源树
    • 数据集
  • 知识沉淀
    • 学术空间
订阅小程序
旧版功能
aminer vip
开通会员低至0.73元/天
一次搞定AI科研
立即登录
  • English
  • 联系方式
    惠普

    惠普

    Hewlett-Packard Inc.
    企业
    7,882论文总数
    23.4万引用总数

    惠普(Hewlett-Packard,简称HP)是信息科技(IT)公司之一,成立于1939年,总部位于美国加利福尼亚州帕洛阿尔托市。惠普下设三大业务集团:信息产品集团、打印及成像系统集团和企业计算机专业服务集团。 中国惠普有限公司总部位于北京,在上海、广州、沈阳、南京、西安、武汉、成都、深圳等都设有分公司。中国惠普在大连设有惠普全球呼叫中心,在重庆设有生产工厂,在天津设有数据中心。

    论文量&引用量时间轴

    机构学者

    论文(7882)

    年份
    起
    –
    止
    排序
    1Why DPO is a Misspecified Estimator and How to Fix It
    Aditya Gopalan,Sayak Ray Chowdhury,Debangshu Banerjee

    Direct alignment algorithms such as Direct Preference Optimization (DPO) fine-tune models based on preference data, using only supervised learning instead of two-stage reinforcement learning with human feedback (RLHF). We show that DPO encodes a statistical estimation problem over reward functions induced by a parametric policy class. When the true reward function that generates preferences cannot be realized via the policy class, DPO becomes misspecified, resulting in failure modes such as preference order reversal, worsening of policy reward, and high sensitivity to the input preference data distribution. On the other hand, we study the local behavior of two-stage RLHF for a parametric class and relate it to a natural gradient step in policy space. Our fine-grained geometric characterization allows us to propose AuxDPO, which introduces additional auxiliary variables in the DPO loss function to help move towards the RLHF solution in a principled manner and mitigate the misspecification in DPO. We empirically demonstrate the superior performance of AuxDPO on didactic bandit settings as well as LLM alignment tasks.

    ICLR 2026引用:5
    引用
    AI阅读
    加入学术空间
    2Quantization-Robust LLM Unlearning Via Low-Rank Adaptation
    João Vitor Boer Abitante, Joana Meneguzzo Pasquali, Luan Fonseca Garcia, Ewerton de Oliveira, Thomas da Silva Paula, Rodrigo C. Barros, Lucas S. Kupssinskü

    Large Language Model (LLM) unlearning aims to remove targeted knowledge from a trained model, but practical deployments often require post-training quantization (PTQ) for efficient inference. However, aggressive low-bit PTQ can mask or erase unlearning updates, causing quantized models to revert to pre-unlearning behavior. We show that standard full-parameter fine-tuning often induce parameter changes that are too small to survive 4-bit quantization. We propose quantization-robust unlearning via low-rank adaptation (LoRA): we freeze the base model and concentrate unlearning into trainable adapters so that the effective update is preserved after quantization. On Llama-2-7B evaluated with MUSE dataset (BOOKS and NEWS), LoRA improves 4-bit utility by up to 7.93 points (NPO+GDR on BOOKS: 50.17 to 58.10) and yields higher 4-bit utility on NEWS for GA+GDR (40.06 to 44.82, increase of 4.76). LoRA also substantially reduces privacy leakage under 4-bit PTQ, e.g., for GA+KLR on BOOKS, PrivLeak moves from -25.68 to -5.86 (closer to ideal 0), while maintaining strong forgetting (VerMem and KnowMem near 0). Thus, using LoRA for Machine Unlearning is beneficial for scenarios where quantization is necessary for model deployment.

    20262026 International Joint Conference on Neural Networks (IJCNN)(2026)引用:4
    引用
    AI阅读
    加入学术空间
    3Belief-Space Control for Personalized Cancer Treatment Via Active Inference
    Deniz Sargun, H. Bugra Tulay,C. Emre Koksal

    Cancer treatment is at the core a sequential decision-making problem with partial observability, latent patient heterogeneity, and explicit constraints on the budget for medical measurements. Unlike standard Reinforcement Learning (RL) approaches that control state trajectories, cancer treatments permanently modify patients' transition dynamics, changing how states evolve over time. We model cancer treatment as a belief-space planning problem using active inference, deriving an expected free-energy objective that unifies goal-directed control and information acquisition under measurement budgets without. We implement this framework using real clinical cancer data from the AACR Project GENIE Biopharma Collaborative dataset. Results on clinical data demonstrate a simultaneous patient categorization and high treatment efficacy, under real measurement and treatment constraints.

    2026引用:1
    引用
    AI阅读
    加入学术空间
    4Blending Neural Control Density Functions for Stabilization and Safety
    Sahil Chaudhary,Chaitanya Murti,Chiranjib Bhattacharyya

    Recent work on Neural Network-based methods for nonlinear control use Lyapunov Functions to obtain controllers with guarantees of stability. However, Lyapunov-based methods are fundamentally limited: they cannot be used for smooth blending with formal Region of Attraction (RoA) expansion guarantees, and also fail to certify stability when unstable equilibria or saddle points are present. Density functions provide an alternate stability certificate, and address these limitations by certifying almost everywhere stability, and enable smooth blending of controllers. Learning valid density certificates is challenging due to integrability constraints, and the effect of density-based blending controllers on RoAs is not well understood. In this work, we provide the first guarantee that controllers blended with density functions yield RoAs containing the union of the RoAs achieved by the constituent controllers. Then, we propose a novel exponential characterization of density functions that provably satisfies the integrability condition, and introduce Neural Control Density Functions (NCDFs), that leverage this new parameterization. We also extend NCDFs for synthesizing safe-stable controllers by combining NCDFs with control barrier functions (NCDF-CBFs). Our experiments show that blended controllers obtain superior RoAs to state-of-the-art methods like Neural Lyapunov Control and Sum-of-Squares based techniques.

    2026ICML 2026(2026)
    引用
    AI阅读
    加入学术空间
    5Evaluating Casing Wear Risk in a Legacy North Sea Well Repurposed for CCUS: A Case Study
    Douglas Gouraly, Gergana Vanova, Lingfen Pan

    Abstract Reusing legacy oil and gas wells as carbon dioxide (CO2) injectors can reduce the cost, schedule and embodied emissions of a carbon capture, utilization and storage (CCUS) development compared with drilling purpose-built injection wells. Legacy wells were, however, designed and constructed for hydrocarbon production, and any casing wear sustained during the original drilling campaign was assessed, if it was assessed at all, against a production design envelope. CO2 injection subjects the same casing string to a different combination of pressure, temperature and load cases, and a wear profile that was acceptable for production cannot be assumed acceptable for injection service. This paper presents a case study of a legacy offshore gas well in the North Sea, referred to here as Well A and its sidetrack Well A-ST1, that is being evaluated for conversion to a CO2 injector. The drilling operations of both wellbores were reconstructed from incomplete 1990s-era records, and a stiff-string torque, drag and buckling (T&D&B) model, combined with a wear factor relationship calibrated against North Sea field data, was used to predict the wear distribution along the 10-3/4 in. × 9-5/8 in. production casing for a Best Estimate and a Conservative case. The predicted wear distribution was then compared against a depth-dependent wear allowance derived by the Operator from CO2-injection-specific collapse and triaxial load cases. The maximum predicted wear, 10.7 to 11.6% of nominal wall thickness, occurs at approximately 1,700 m measured depth (MD) and is consistent with conventional drilling-induced wear concentrated across high-dogleg intervals shallow in the well. At that depth the wear allowance exceeds 20% of wall thickness and a substantial margin remains. In the interval between 3,500 and 4,000 m MD the predicted wear is far lower, 3.0 to 3.4%, but the allowance reduces to 2.4%, producing the only exceedance along the string. This reversal of the critical interval, from the depth of maximum wear to the depth of minimum allowance, is the principal finding of the study. It demonstrates that screening a legacy well for CCUS service requires the full depth-dependent wear distribution to be compared against the full application-specific allowance profile; a single maximum-wear check against a single allowable value can miss the interval that governs the reuse decision. The paper describes the reconstruction and modelling workflow, examines the main uncertainties, and discusses mitigation options including a scab liner and a revised injection strategy. The workflow is directly applicable to other legacy wells being screened as CCUS injection candidates, particularly where historical data are sparse.

    2026SPE Annual Technical Conference and Exhibition(2026)
    引用
    AI阅读
    加入学术空间
    立即登录,查看全部 7882 篇论文

    合作机构(100)

    巴黎医院公共援助合作论文 226
    惠普实验室合作论文 149
    斯坦福大学合作论文 126
    法国国家健康与医学研究院合作论文 126
    俄勒冈州立大学合作论文 87
    普渡大学合作论文 67
    巴黎大学合作论文 62
    加州大学合作论文 54
    索邦大学合作论文 54
    国际商业机器公司合作论文 46

    机构统计