The performance of large language model (LLM) inference engines heavily relies on deployment-specific parameters such as batch size, parallelism, CUDA graphs, and GPU and network topology, however, with every new LLM released, the optimal setting of these parameters becomes a problem of interest. Additionally, as developers find new applications for LLMs, the parameter space only increases, with large scale inference deployments that span multiple nodes (i.e. llm-d [5]). In this work, we demonstrate ''llm-tuna'', an open-source framework that automates parameter search for vLLM [4] based inference workloads across a diverse set of hardware setups. Our framework utilizes Bayesian optimization via Optuna to select the best configuration while parallelizing trials across multiple nodes to enable large scale studies. We evaluate our system on a mixture-of-experts (MoE) model family (Qwen3-30B-A3B, Kimi-K2-INT4, GPT-OSS-120B, Deepseek-R1-671B), configuring each experiment to avoid artifacts of KV-cache pre-emption effects. Our method achieves output throughput gains of up to 32.9% while never suggesting configurations that are worse than vLLMs' baseline performance. Our results demonstrate the potential of systematic auto-tuning to expose non-obvious performance regimes and guide operator defaults.
论文检索
输入标题、作者或关键词，从 901 篇学术成果中精准定位
Recommender systems have been shown to exhibit popularity bias by over-recommending popular items and under-recommending relevant niche items. We seek to understand niche users in benchmark recommendation datasets as a step toward mitigating popularity bias. We find that, compared to mainstream users, niche-preferring users exhibit a longer-tailed activity-level distribution, indicating the existence of users who both prefer niche items and exhibit high activity levels on platforms. We partition users along two axes: (1) activity level (''power'' vs. ''light'') and (2) item-popularity preference (''mainstream'' vs. ''niche''), and show that in three benchmark datasets, the number of power-niche users (high activity and niche preference) is statistically significantly larger than expected. We also find that interaction data from power-niche users is especially valuable for improving recommendations for not only niche but also mainstream users. In contrast, many existing popularity bias mitigation methods have focused on upweighting niche users regardless of activity level. Motivated by the value of power-niche user data, we propose PAIR (Popularity-and-Activity-Informed Reweighting), a framework for reweighting the Bayesian Personalized Ranking (BPR) loss that simultaneously reweights based on user activity level and item popularity, upweighting power-niche users the most. We instantiate the framework on both deep and shallow collaborative filtering models, and experiments on benchmark datasets show that PAIR reduces popularity bias and can increase overall performance. Although existing popularity-bias mitigation methods yield a trade-off between performance and bias, our results suggest that considering both user activity level and popularity preference leads to Pareto-dominant performance.
As social media and the World Wide Web become hubs for information dissemination, effectively organizing and understanding the vast amounts of dynamically evolving Web content is crucial. Knowledge graphs (KGs) provide a powerful framework for structuring this information. However, the rapid emergence of new hot topics, user relationships, and events in social media renders traditional static knowledge graph embedding (KGE) models rapidly outdated. Continual Knowledge Graph Embedding (CKGE) aims to address this issue, but existing methods commonly suffer from catastrophic forgetting, whereby older, but still valuable, information is lost when learning new knowledge (such as new memes or trending events). This means the model cannot effectively learn the evolution of the data. We propose a novel CKGE framework, BAKE. Unlike existing methods, BAKE formulates CKGE as a sequential Bayesian inference problem and utilizes the Bayesian posterior update principle as a natural continual learning strategy. This principle is insensitive to data order and provides theoretical guarantees to preserve prior knowledge as much as possible. Specifically, we treat each batch of new data as a Bayesian update to the model's prior. By maintaining the posterior distribution, the model effectively preserves earlier knowledge even as it evolves over multiple snapshots. Furthermore, to constrain the evolution of knowledge across snapshots, we introduce a continual clustering method that maintains the compact cluster structure of entity embeddings through a regularization term, ensuring semantic consistency while allowing controlled adaptation to new knowledge. We conduct extensive experiments on multiple CKGE benchmarks, which demonstrate that BAKE achieves the top performance in the vast majority of cases compared to existing approaches.
Large Language Models (LLMs) possess strong representation and reasoning capabilities, but their application to structure-based drug design (SBDD) is limited by insufficient understanding of protein structures and unpredictable molecular generation. To address these challenges, we propose Exploration-Augmented Latent Inference for LLMs (ELILLM), a framework that reinterprets the LLM generation process as an encoding, latent space exploration, and decoding workflow. ELILLM explicitly explores portions of the design problem beyond the model's current knowledge while using a decoding module to handle familiar regions, generating chemically valid and synthetically reasonable molecules. In our implementation, Bayesian optimization guides the systematic exploration of latent embeddings, and a position-aware surrogate model efficiently predicts binding affinity distributions to inform the search. Knowledge-guided decoding further reduces randomness and effectively imposes chemical validity constraints. We demonstrate ELILLM on the CrossDocked2020 benchmark, showing strong controlled exploration and high binding affinity scores compared with seven baseline methods. These results demonstrate that ELILLM can effectively enhance LLMs' capabilities for SBDD. Our code is available at https://github.com/hxnhxn/ELILLM.
In subjective tasks, different individuals can have different correct answers for the same input—the ground truth is not fixed but rather determined by each user's personal perspective. Standard foundation models suppress this individual variation by producing population-averaged predictions; conversely, few-shot in-context learning (ICL) struggles to capture user-specific preferences from limited demonstrations. To address this limitation, we propose Population-Guided Bayesian Calibration (PGBC), a test-time personalization method that applies Bayesian inference to foundation model outputs for user-specific alignment. Unlike ICL methods that rely on the model to implicitly infer preferences from examples, PGBC explicitly quantifies each individual's deviation from the population distribution and incorporates this into prediction with minimal feedback (K < 10). Experiments on three subjective benchmark datasets across four foundation models demonstrate that PGBC significantly outperforms zero-shot baselines with a large effect size (Cohen's d = 1.21), and the improvement over ICL approaches is even more pronounced (d = 1.40). Furthermore, user-level distribution analysis reveals that PGBC aligns predictions with individual preferences when zero-shot baselines exhibit substantial bias, while ICL methods increase distributional mismatch. Our results show that PGBC enables practical deployment in cold-start scenarios where minimal user feedback must yield maximal personalization.
Inertial Confinement Fusion (ICF) holds transformative promise for sustainable, near-limitless clean energy, yet remains constrained by prohibitively high costs and limited experimental opportunities. This paper presents Human-in-the-Loop Meta Bayesian Optimization (HL-MBO), a framework that integrates expert knowledge with few-shot, uncertainty-aware machine learning to accelerate discovery in data-scarce, high-stakes scientific domains. HL-MBO introduces a meta-learned surrogate model with an expert-informed acquisition function to recommend candidate experiments. To foster trust and enable informed decisions, HL-MBO also provides interpretable explanations of its suggestions. We show that HL-MBO outperforms current BO methods for ICF energy yield optimization, as well as benchmarks in molecular optimization and critical-temperature maximization for superconducting materials.
Mechanical ventilation (MV) is essential in intensive care units (ICUs), yet conventional protocols lack personalization and risk harmful over- or under-ventilation. Offline reinforcement learning (ORL) enables policy optimization from retrospective clinical data without unsafe online interaction, but existing methods are highly sensitive to distributional shift and out-of-distribution (OOD) actions, limiting their reliability in complex clinical settings. To address these challenges, We propose UBER-CQL (Uncertainty-Balanced Exploration and Robust Conservative Q-Learning), a robust ORL algorithm for safe decision-making under dataset shift. UBER-CQL integrates heteroscedastic Bayesian neural networks with conservative Q-learning to model posterior Q-value uncertainty, which is used to adaptively penalize unreliable high-risk actions while maintaining performance within the data support. We further design numerically stable objectives for conservative Bayesian value estimation. Experiments on in-distribution and OOD subsets of MIMIC-III and eICU demonstrate that UBER-CQL outperforms state-of-the-art ORL and clinician baselines, producing safer and more effective MV strategies.
Automatic radiology report generation (RRG) aims to simulate the workflow of radiologists, assisting them in clinical diagnosis. However, existing methods often fall short in utilizing all information relevant to the examination, as is typically done in clinical practice. Although some works attempt to incorporate multi-view images and historical data, these additional inputs may sometimes lead to avoidable diagnostic errors on the contrary. To address these challenges, we introduce a decision-making stage after report generation for the first time and propose a Posterior Diagnostic Decision framework (PDD-RRG) to integrate potentially conflicting diagnoses. Specifically, we create various subsets of input data and utilize an existing RRG model to generate reports from different perspectives. Then the Bayesian posterior probability and the learned thresholds for each clinical observation are calculated to obtain an aggregated diagnostic conclusion, which is subsequently used to refine the generated report. Experiments on MIMIC-CXR demonstrate that our proposed PDD-RRG can effectively enhance the clinical efficacy of existing RRG models without any retraining.
Many scientific and engineering applications require estimating unknown parameters from experimentally observable data -- an inverse problem that is inherently challenging due to nonlinearity, noise, and ill-posedness. In this paper, we propose an active diffusion-based inverse problem solver. A diffusion model is trained to learn the mapping between the parameter space and the observable space. By iteratively detecting and correcting model misspecification through posterior uncertainty, the method discovers and learns the correct region of parameter space, even when initial training bounds exclude the true parameters. This provides a principled, Bayesian justification for adaptive domain augmentation and ensures robust inference for inverse problems under incomplete prior knowledge. We demonstrate the effectiveness of our inverse solver for a toy inverse problem with infinite solutions, and for the parameterization of the quantum correlation functions to event observables in a Quantum Chromodynamics analysis of nucleon structure.
Survival models on networks describe the time-evolving dynamics of complex systems, but existing frameworks suffer from two key limitations: First, many past approaches implicitly assume independent censoring, which can introduce substantial bias. Second, point-process-based models often emphasize only sending behavior and overlook shared latent structure between interacting nodes. Recent work has incorporated copulas into deep survival models to mitigate censoring bias, but these methods do not capture topological role structure. Another line of research combines survival analysis with the Mixed-Membership Stochastic Blockmodel (MMSB) to model heterogeneity via role-specific hazard rates; however, it typically assumes structural independence and therefore misses important correlations between roles. To bridge these gaps, we apply an improved Copula–MMSB mechanism and propose the first Survival Fully-Collapsed Copula Mixed Membership Blockmodel. Our framework jointly models time-to-event dynamics and complex dependencies among latent roles. On the Enron benchmark, it surpasses state-of-the-art survival blockmodels in both predictive likelihood and structural interpretability. Additionally, the model admits an exact closed-form discrete distribution over role pairs, which enables a fully collapsed Gibbs sampler and avoids the slow mixing and poor convergence common in non-conjugate inference. In simulations, this yields a 100× improvement in sampling efficiency over semi-collapsed baselines.
Unsupervised 3D semantic segmentation is vital for label-free open-world perception. However, current methods typically struggle with two core limitations: fixed category assumptions that restrict adaptability in complex scenes, and geometric incompleteness caused by irregular point sampling, which degrade the integrity of local geometric descriptors for fine-grained or long-tail objects. To address these, we propose DyG-Seg, a Dynamic Geometric-Semantic Alignment framework. First, targeting the challenge of incomplete geometry, the Geometric Manifold Rectification (GMR) leverages a diffusion model to reconstruct integral object geometry and recover the underlying manifold topology, thereby enhancing feature discriminability. Utilizing this geometric prior, the Dynamic Prototype Contrastive Clustering (DPCC) applies non-parametric Bayesian inference to automatically estimate the optimal number of clusters and generate pseudo-labels. Finally, our Iterative Optimization with Dynamic Constraints integrates Dynamic Class Balancing (DCB) to mitigate long-tail bias. Capitalizing on the recovered manifold topology, we enforce Spatial Geometric Consistency to rectify structural incompleteness and ensure spatially coherent semantic predictions. Extensive experiments on ScanNet and S3DIS demonstrate that DyG-Seg discovers latent semantic structures and significantly outperforms state-of-the-art unsupervised methods.
Brain effective connectivity (EC) characterizes directional causal interactions among brain regions. However, learning stable and directionally explicit EC networks from multimodal data remains challenging. In practice, functional magnetic resonance imaging (fMRI) and electroencephalography (EEG) differ substantially in spatio-temporal resolution and noise characteristics. Existing methods often rely on manual spatio-temporal alignment and tend to recover only partial causal structures under high noise and Bayesian equivalence classes. To address these challenges, we propose a spatio-temporal constrained Bayesian causal network for multimodal brain effective connectivity learning (STCBN-EC). First, STCBN-EC constructs an anatomically guided EEG–fMRI spatial mapping and derives a unified spatio-temporal representation through slice-level alignment and adaptive modality fusion. Then, a Bayesian causal network is employed to model nonlinear inter-regional dependencies, where uncertainty-driven surrogate scoring is used to evaluate candidate structures. Finally, multimodal representation learning and EC structure estimation are jointly optimized via a gradient-free global optimization strategy. Experiments on simulated and real EEG–fMRI datasets demonstrate that STCBN-EC outperforms state-of-the-art methods and effectively captures state-dependent directional interactions among brain regions.
Many real-world applications require solving families of expensive multi-objective optimization problems~(EMOPs) under varying operational conditions. This can be formulated as parametric expensive multi-objective optimization problems (P-EMOPs) where each task parameter defines a distinct optimization instance. Current multi-objective Bayesian optimization methods have been widely used for finding finite sets of Pareto optimal solutions for each task. However, P-EMOPs present a fundamental challenge: the continuous task parameter space can contain infinite distinct problems, each requiring separate expensive evaluations. To address this, we propose learning an inverse model to amortize the multi-objective optimization cost across the continuous task-preference space, enabling direct solution prediction for any query without the need for expensive re-evaluation. This paper introduces a novel parametric multi-objective Bayesian optimizer that learns this inverse model by alternating between (1) generative solution sampling via conditional generative models and (2) acquisition-driven search leveraging inter-task synergies. This approach enables effective optimization across multiple tasks and finally achieves direct solution prediction for unseen parameterized EMOPs without re-evaluations. We theoretically justify the faster convergence by leveraging inter-task synergies through task-aware Gaussian processes. Based on that, empirical studies in synthetic and real-world benchmarks further verify the effectiveness of the proposed parametric optimizer.
Optimization in industrial systems often involves calibrating from a semi-optimized state, where global exploration methods like Reinforcement Learning (RL) or Bayesian Optimization (BO) are inefficient or unsafe. We propose Causal Newton Optimization (CNO), an online algorithm that iteratively calibrates inputs under a known causal graph but unknown structural equations. CNO estimates local linear causal effects via additive interventions and employs a log-linear variance regression to robustly guide Newton-based updates. Evaluations on synthetic systems and a chemical plant simulator demonstrate that CNO achieves the best balance between objective improvement and robustness. While traditional PID control suits standard dynamical systems, CNO significantly outperforms RL and BO in complex structural causal models, providing the robust stability vital for safety-critical real-world applications.
Model counting of Disjunctive Normal Form (DNF) formulas is a critical problem in applications such as probabilistic inference and network reliability. For example, it is often used for query evaluation in probabilistic databases. Due to the computational intractability of exact DNF counting, there has been a line of research into a variety of approximation algorithms. We develop a new Monte Carlo approach with an adaptive stopping rule and short-circuit formula evaluation. We prove it achieves Probably Approximately Correct (PAC) learning bounds and has asymptotically improved time and randomness complexity compared to previous methods. We also show experimentally that it out-performs prior algorithms by at least three orders of magnitude in running time and scalability.
Learning to Defer operates by deferring AI-uncertain samples to humans, enabling the system to outperform either alone. Some works extend it by introducing human-AI combination to deferral. However, they remain limited to binary choices. Given that AI, humans, and human-AI combination are each indispensable, we extend the binary deferral paradigm to a ternary one. There are two challenges: i) how to effectively select among the three modes, especially distinguishing humans from combination? ii) how to design an enhanced combination without being degraded by unreliable model predictions? To address the challenges, we propose TriHAI, a tri-mode deferral method that routes samples to suitable modes based on normalized confidence scores obtained through a discriminative gating network, and enhances combination mode by fusing model predictions refined by human-guided Conformal Prediction and human predictions via Bayesian theory. Experiments on three datasets show that TriHAI surpasses other human-AI baselines by up to 9.57% in accuracy.
Life, Machine Learning, and the Search for Habitability: Predicting Biosignature Fluxes for the Habitable Worlds Observatory
PDF ↗Future direct-imaging flagship missions, such as NASA's Habitable Worlds Observatory (HWO), face critical decisions in prioritizing observations due to extremely stringent time and resource constraints. In this paper, we introduce two advanced machine-learning architectures tailored for predicting biosignature species fluxes from exoplanetary reflected-light spectra: a Bayesian Convolutional Neural Network (BCNN) and our novel model architecture, the Spectral Query Adaptive Transformer (SQuAT). The BCNN robustly quantifies both epistemic and aleatoric uncertainties, offering reliable predictions under diverse observational conditions, whereas SQuAT employs query-driven attention mechanisms to enhance interpretability by explicitly associating spectral features with specific biosignature species. We demonstrate that both models achieve comparably high predictive accuracy on an augmented dataset spanning a wide range of exoplanetary conditions, while highlighting their distinct advantages in uncertainty quantification and spectral interpretability. These capabilities position our methods as promising tools for accelerating target triage, optimizing observation schedules, and maximizing scientific return for upcoming flagship missions such as HWO.
Hyperparameter Optimization (HPO) is crucial in machine learning, aiming to optimize hyperparameters to enhance model performance. Although existing methods that leverage prior knowledge—drawn from either previous experiments or expert insights—can accelerate optimization, acquiring a correct prior for a specific HPO task is non-trivial. In this work, we propose to relieve the reliance on external knowledge by learning a reliable prior {directly} from low-fidelity (LF) problems. We introduce {Lamda}, an algorithm-agnostic framework designed to boost any baseline HPO algorithm. Specifically, {Lamda} operates in two phases: (1) it learns a reliable prior by exploring the LF landscape under limited computational budgets, and (2) it leverages this learned prior to guide the HPO process. We showcase how the {Lamda} framework can be integrated with various HPO algorithms to boost their performance, and further conduct theoretical analysis towards the integrated Bayesian optimization and bandit-based Hyperband. We conduct experiments on 56 HPO problems spanning diverse domains and model scales. Results show that {Lamda} consistently enhances its baseline algorithms. Compared to nine state-of-the-art HPO algorithms, our {Lamda} variant achieves the best performance in 51 out of 56 HPO tasks while it is the second best algorithm in the other 5 cases.
Parametric multi-objective optimization (PMO) addresses the challenge of solving an infinite family of multi-objective optimization problems, where optimal solutions must adapt to varying parameters. Traditional methods require re-execution for each parameter configuration, leading to prohibitive costs when objective evaluations are computationally expensive. To address this issue, we propose Parametric Pareto Set Learning with multi-objective Bayesian Optimization (PPSL-MOBO), a novel framework that learns a unified mapping from both preferences and parameters to Pareto-optimal solutions. PPSL-MOBO leverages a hypernetwork with Low-Rank Adaptation (LoRA) to efficiently capture parametric variations, while integrating Gaussian process surrogates and hypervolume-based acquisition to minimize expensive function evaluations. We demonstrate PPSL-MOBO's effectiveness on two challenging applications: multi-objective optimization with shared components, where certain design variables must be identical across solution families due to modular constraints, and dynamic multi-objective optimization, where objectives evolve over time. Unlike existing methods that cannot directly solve PMO problems in a unified manner, PPSL-MOBO learns a single model that generalizes across the entire parameter space. By enabling instant inference of Pareto sets for new parameter values without retraining, PPSL-BO provides an efficient solution for expensive PMO problems.
This paper presents the Min-Cut Bayesian Network Consensus (MCBNC) algorithm, a greedy method for structural consensus of Bayesian Networks (BNs), with applications in federated learning and model aggregation. MCBNC prunes weak edges from an initial unrestricted fusion using a structural score based on min-cut analysis, integrated into a modified Backward Equivalence Search (BES) phase of the Greedy Equivalence Search (GES) algorithm. The score quantifies edge support across input networks and is computed using max-flow. Unlike methods with fixed treewidth bounds, MCBNC introduces a pruning threshold θ that can be selected post hoc using only structural information. Experiments on real-world BNs show that MCBNC yields sparser, more accurate consensus structures than both canonical fusion and the input networks. The method is scalable, data-agnostic, and well-suited for distributed or federated structural learning of BNs or causal discovery.