Large Language Models (LLMs) are extensively used at biomedical text processing but often fail to capture the complex, functional relationships encoded in expert knowledge graphs like the Human Phenotype Ontology (HPO). This "semantic gap'" limits their utility in precision medicine tasks such as rare disease diagnosis, where distinguishing overlapping clinical presentations requires understanding underlying pathophysiological connections rather than just surface-level textual similarity. In this work, we propose a Neuro-Symbolic Alignment Framework that bridges this separation by integrating literature-mined specialized phenotypical descriptions with the ontological structure used as reference. Specifically, we augment phenotype representations with automatically selected text fragments from massive corpus of descriptions mined from scientific literature (PubMed), overcoming the typical data scarcity of standard ontology definitions. We define a new embedding adaptation procedure whose fine-tuning approach is guided by a novel "Disease-Overlap" similarity measure, which prioritizes clinical co-occurrence of phenotypes over taxonomic distance, and optimizes the embedding space using AnglE Loss to mitigate gradient saturation. Extensive evaluations show that our approach significantly outperforms state-of-the-art baselines, including SapBERT, on both intrinsic semantic correlation and practical downstream tasks, including synthetic patient disease ranking and solving real cases stored in Phenopacket, where our model achieves x4 top-1 accuracy than the previous best model.
论文检索
输入标题、作者或关键词，从 8,216 篇学术成果中精准定位
Federated learning (FL) enables privacy-preserving collaboration for medical image analysis across decentralized institutions but faces major challenges from non-IID data distributions, high communication overhead in multi-round protocols, and catastrophic forgetting when models must adapt to sequentially arriving tasks. These issues are particularly critical in healthcare, where repeated client participation is often infeasible. We address this setting by proposing a novel class-incremental continual learning (CL) model for a one-shot FL paradigm, in which each task introduces new classes, clients observe heterogeneous and evolving class distributions, and communication with the server occurs only once. Clients estimate class-conditional feature distributions via Variational Inference (VI) from private data and transmit compact statistics to the server, which synthesizes features to train a global classifier in a single communication round. The server aggregates these distributional summaries, synthesizes feature embedding, and learns a global classifier without revisiting real past data or contacting clients again. Our major novelty is continual adaptation at the distribution level via synthetic replay from stored class mixtures, complemented by lightweight distillation. This approach substantially mitigates catastrophic forgetting while consistently enhancing recognition of newly introduced classes. Extensive experiments on multiple medical imaging benchmarks demonstrate that our method outperforms both state-of-the-art one-shot FL and class-incremental CL approaches. Compared with existing CL models, it achieves 90–97% average accuracy with near-zero forgetting (≤ 0.10%) across different tasks and evaluation settings.
Real-world deployment of large language models (LLMs) necessitates a robust and low-latency approach to detect prompt injections; existing low-overhead methods fail to simultaneously boost robustness and reduce latency. Current defenses for prompt injection either rely on brittle heuristics or invoke costly auxiliary models, imposing a significant runtime burden. We introduce BERM, a lightweight framework that performs in-situ detection by modeling a host LLM’s internal representations extracted during prefill, adding negligible overhead. Our approach trains a lightweight classifier atop the LLM by learning a compact manifold of benign representations via joint contrastive learning to maximize the separation from malicious representations. At inference, this pre-trained classifier enables in-situ detection without invoking auxiliary guard models. On a diverse landscape of prompt injection attacks, our framework establishes a new state-of-the-art, achieving an F1-score 5.2 percentage points (pp) higher than the best prior work. Critically, BERM achieves this while being over 12x faster, reducing inference overhead to near-zero.
Integration of Answer Set Programming (ASP) with neural networks has emerged as a promising tool in Neuro-symbolic AI. While existing approaches extend the capabilities of ASP to real world domains, their reasoning pipelines depend on classical solvers, which is a bottleneck for scalability. To tackle this problem, we propose a new method to compute stable models, called decision-propagation (DProp), which alternates falsity decisions and truth propagations. Successful DProp computations are shown to capture the stable model semantics. We then develop Neural DProp (NDProp), a differentiable extension of DProp with neural computation for decisions and fuzzy evaluation for propagations. We evaluate the capabilities of NDProp for learning decision heuristics as well as neuro-symbolic integration, and compare it with existing neuro-symbolic approaches. The results show that NDProp can learn to efficiently compute stable models, and it improves accuracy and scalability on neuro-symbolic benchmarks.
Offline safe reinforcement learning learns high-return policies that satisfy hard safety constraints using only a pre-collected dataset. This setting is challenging due to the inability to explore, and the risk of propagating value errors through unsafe state-space regions. To address this, first, we characterize the safe state region by developing a framework for learning control barrier functions (CBFs) using a novel generalized Bellman operator, yielding a persistent safety set, from which the agent can remain safe indefinitely. Second, we show that several existing safety set estimation methods (e.g., reachability-constrained RL) can be formulated within our CBF learning framework, highlighting its generality. We further propose a new CBF that ensures safety under environment dynamics uncertainty, unlike standard CBFs designed for deterministic settings. Third, we propose a new reward maximization algorithm that effectively exploits our learned persistent safety set for reward critic estimation. Empirical results on standard benchmarks show that our approach achieves state-of-the-art safety with fewer constraint violations while maintaining competitive returns.
Current approaches to evaluating personality in large language models (LLMs) typically prompt them to self-report on psychological questionnaires such as the Big Five Inventory. However, these methods assess introspective labels rather than observable behavior, despite the fact that LLMs are deployed to act in realistic contexts, not to reflect on their own traits. To bridge this gap, we introduce BehaviorBench, a new benchmark for evaluating LLM personality through concrete behaviors in everyday scenarios. Grounded in established personality psychology, BehaviorBench links each Big Five trait to validated behavioral manifestations and embeds them in contextually plausible situations that naturally elicit trait-relevant actions. We evaluate a range of models and personality shaping strategies using BehaviorBench and find a substantial mismatch between self-reported personalities and actual behaviors. By grounding evaluation in observable behavior rather than introspection, our work reveals critical gaps in current LLM personality modeling and control mechanisms. Our code and data are available at https://github.com/butra1n/BehaviorBench
AI for materials science is a critical topic within AI for science, aiming to accelerate materials discovery and produce accurate property predictions. Bilayer 2D material stacking is essential for exploring new materials with novel functions and inherent phenomena, enabling the creation of new 2D bilayers for diverse real-world applications. Research on bilayer vdWs materials has made significant progress from experimental and computational perspectives. Various bilayer materials have been successfully synthesized experimentally and the increasing utilization of high-throughput computing technology has constructed several computational two-dimensional materials databases. However, the use of AI to model bilayer stacking and predict new properties remains underexplored, necessitating further research studies. In this work, we propose a novel multimodal learning approach to study the interfaces between dissimilar materials that jointly enable new or multiple functions, and to predict new properties arising from the vertical integration (stacking) of different functional material layers under given configurations. Comprehensive experiments demonstrate the effectiveness and efficiency of our approach compared to baseline methods. Our code is available at https://github.com/AnVuong123/bimat_ml.
Convex clustering is a well-regarded clustering method, resembling the similar centroid-based approach of Lloyd's $k$-means, without requiring a predefined cluster count. It starts with each data point as its centroid and iteratively merges them. Despite its advantages, this method can fail when dealing with data exhibiting linearly non-separable or non-convex structures. To mitigate the limitations, we propose a kernelized extension of the convex clustering method. This approach projects the data points into a Reproducing Kernel Hilbert Space (RKHS) using a feature map, enabling convex clustering in this transformed space. This kernelization not only allows for better handling of complex data distributions but also produces an embedding in a finite-dimensional vector space. We provide a comprehensive theoretical underpinning for our kernelized approach, proving algorithmic convergence and establishing finite sample bounds for our estimates. The effectiveness of our method is demonstrated through extensive experiments on both synthetic and real-world datasets, showing superior performance compared to state-of-the-art clustering techniques. This work marks a significant advancement in the field, offering an effective solution for clustering in non-linear and non-convex data scenarios.
Zero-shot transfer for offline reinforcement learning involves generalizing to a wide range of tasks that are often risk-sensitive, without new interaction. We show here that although Successor Features (SFs) provide a principled framework for transfer through abstracting dynamics from rewards, their typical form is restricted to expected cumulative returns and is thus indifferent to variance in returns. We contend that such a mean-centric approach introduces bias by conflating risky and safe trajectories with identical expected outcomes, as it ignores outcome dispersion critical for risk-aware evaluation. Additionally, we note that distributional methods can be employed to model outcome distributions, but frequently do so at a high computational cost due to the need for discretization or elaborate sampling. To overcome this problem, we introduce a form of successor representation that incorporates an uncertainty envelope over states: Gaussian Distributional Successor Features (GDSF). Rather than inferring an implicit generative process, GDSF learns to approximate the distribution of cumulative outcomes with a Gaussian through recursive moment-matching. We present this framework for efficient representation of aleatoric uncertainty, without resorting to computationally expensive existing distributional methods. We also introduce Latent Outcome Optimization, an inference procedure which finds high-reward outcomes in the data support and drives a goal-conditioned policy. This dual-mode design allows the agent to transfer to non-linear objectives in a zero-shot way while preserving closed-form efficiency for linear tasks. Experiments on the D4RL benchmark show that GDSF yields better performance in risk-sensitive tasks while achieving at least competitive results in standard linear regimes.
In-context learning (ICL) enables large language models (LLMs) to adapt to new tasks with considerable performance gains, yet its effectiveness is highly sensitive to the choice of demonstrations. Most existing selection methods rely on heuristic or proxy signals, such as similarity, diversity, or uncertainty, and select demonstrations independently, which may misalign with downstream performance and overlook set-level composition effects. Therefore, we propose Performance-Driven Demonstration Selection (PDDS), which directly aligns demonstration selection with ICL performance. PDDS formulates selection as predicting the target LLM’s downstream task performance for a given query–in-context pair, replacing proxy heuristics with a performance-aware objective. Unlike prior approaches that rank demonstrations independently before composing the prompt, PDDS evaluates the entire in-context set, capturing inter-demonstration interactions and composition effects. PDDS trains an end-to-end scorer using supervision from the target LLM’s actual task outcomes. At inference time, it selects high-scoring in-context sets without additional target LLM calls or validation-time feedback. Across 6 NLP tasks, 8 datasets, and 8 target LLMs ranging from 1B to 70B parameters, PDDS achieves state-of-the-art results and generalizes well across LLMs and datasets. PDDS remains effective with as few as 30 training instances and can be integrated as a plug-and-play component to enhance existing ICL methods.
Large Reasoning Models (LRMs) often suffer from overthinking due to redundant verification steps. Existing approaches for mitigating overthinking, such as fast-slow thinking switching and reasoning trajectory compression, fail to make a fine-grained distinction between beneficial and redundant steps within the LRM's reasoning process, and may thus impair reasoning capability in their pursuit of efficiency. To simultaneously improve reasoning efficiency and capability, we propose EvoThink, a framework that reduces redundant verification and encourages the exploration of new reasoning paths. EvoThink comprises two key components: Self-Pruning Training (SPT), an unsupervised method that iteratively prunes redundant reasoning steps and self-trains on the concise trajectories; and Aha-Moment Preference Optimization (AMPO), which, inspired by genetic algorithms, identifies valuable failed reasoning attempts, synthesizes from-wrong-to-right aha-moment data, and optimizes the model to internalize this reasoning pattern. Extensive evaluations across mathematical reasoning and code generation benchmarks demonstrate that EvoThink not only substantially reduces inference-time token usage but also improves the reasoning capability of LRMs.
Multi-Agent Path Finding (MAPF) seeks collision-free paths for teams of agents and has a wide range of practical applications. LaCAM*, an anytime configuration-based solver, currently represents the state-of-the-art. Recent work has explored using guidance paths to steer LaCAM* toward configurations that avoid traffic congestion, thereby improving solution quality. However, existing approaches rely on Frank–Wolfe–style optimisation to repeatedly invoke single-agent search before executing LaCAM*, which creates a large computational overhead in large-scale problems. The guide path is also static, which is only helpful for finding the first solution in LaCAM*. To overcome this problem, we propose a new approach that exploits LaCAM*’s ability to construct a dynamic, lightweight traffic map during LaCAM*'s search. Experiments show that our method achieves higher solution quality than state-of-the-art guidance-path approaches in two variants of MAPF problems.
Federated Reinforcement Learning from Human Feedback (RLHF) enables the collaborative alignment of Large Language Models (LLMs) while preserving privacy, yet it faces critical bottlenecks arising from data heterogeneity. Existing approaches typically rely on rigid client-level clustering, which overlooks intra-client heterogeneity and fails to adapt to the multifaceted needs of individual users. To address this, we propose FedPrism, a novel framework that shifts alignment granularity from the coarse client level to the precise instance level. Similar to an optical prism dispersing mixed light into distinct spectral components, FedPrism decomposes complex, intra-client heterogeneous data streams by dynamically routing individual samples to specialized experts based on semantic features. Crucially, we devise a Posterior-Guided Performance Distillation (PGPD) mechanism that leverages experts' actual training loss as self-supervision to autonomously refine routing policies without explicit labels. Extensive experiments show that FedPrism not only establishes new state-of-the-art results but also effectively mitigates the negative transfer prevalent in non-IID settings, ensuring superior alignment fidelity.
Traditional multi-view learning models are primarily designed for static datasets with fixed views. However, in dynamic incremental view environments, this approach inevitably leads to view forgetting, where the introduction of new views weakens previously acquired knowledge. In contrast, the human brain exhibits remarkable memory retention and knowledge transfer capabilities when receiving objects described from different perspectives, with past experience further supplementing new insights. Inspired by underlying neural processing mechanisms, we propose a novel view incremental learning framework named Hebbian View Orthogonal Projection (HVOP). HVOP constructs a knowledge transfer space, where gradient updates are projected onto the orthogonal complement of historical representations, thereby mitigating interference between old and new views. By further incorporating recursive lateral connections and Hebbian learning rules, the proposed model imparts brain-like dynamic adaptability to the learning process, enhancing knowledge transfer and integration, thereby enabling stable knowledge transfer under evolving views. We validate HVOP on node classification tasks, demonstrating its superior performance in both knowledge retention and transfer compared to traditional methods. The results highlight the efficacy of biologically inspired mechanisms in mitigating the view forgetting phenomenon.
Predicting the response of anticancer drugs for specific cancer cell lines remains an important challenge in cancer treatment, largely due to the diversity among cell lines. Existing machine learning methods usually perform well on cell lines seen during training, but their performance often drops when applied to new cell lines. Many approaches rely heavily on cell-line specific features, which limits their ability to generalize. At the same time, methods that focus mainly on generalization often fail to capture differences in drug response across distinct cellular environments. Some studies improve prediction accuracy by introducing additional and more detailed cell line features. However, these features are difficult to collect and are frequently missing, especially for newly studied cell lines. To address these issues, we propose a generalizable bias-disentangled framework for drug response prediction. Our method separates drug-related patterns that are shared across cell lines from representations that reflect cell-line specific environments. We further isolate real but non-generalizable biases that arise from noise or unobserved factors. This design helps preserve accurate predictions for specific cell lines while improving performance on unseen ones. In addition, the bias-related branch supports counterfactual consistency learning when some modalities are missing, allowing the model to handle incomplete data. Experimental results show that GBD-DRP achieves state-of-the-art performance, and additional tests on unseen cell lines and missing-modality settings indicate good generalization in practice.
Decision-focused learning has shown great promise for addressing predict-then-optimize problems, particularly in the presence of under-specified models. However, its practical deployment is often hindered by high computational costs and limited scalability, as it requires solving a constrained optimization problem for each training instance at every iteration. To address these challenges, we propose a novel framework that incorporates Lagrangian decomposition into the decision-focused learning paradigm. Specifically, we introduce a new surrogate objective along with two loss functions for evaluating and training the underlying prediction model. We further propose two variants of our approach, which offer different trade-offs between computational efficiency and solution quality. Our framework can be seamlessly integrated with standard decision-focused learning methods, including Smart Predict-then-Optimize (SPO+) and Implicit Maximum Likelihood Estimation (IMLE). Through experiments on two standard benchmarks, the multi-dimensional knapsack problem and quadratic portfolio optimization, we demonstrate that our approach achieves competitive performance while remaining amenable to parallelization. In particular, it consistently outperforms traditional decision-focused learning methods on large-scale instances, involving up to eight times more variables than those typically considered in related work. The implementation is available at https://github.com/corail-research/DFL-LD.
Inverse reinforcement learning (IRL) has made significant progress in recovering reward functions from expert demonstrations. However, a key challenge remains: how to extract reward functions that generalize across related but distinct tasks. In this paper, we address this by focusing on transferable IRL, learning intrinsic rewards that can drive effective behavior in unseen but structurally aligned environments. Our method leverages a variational autoencoder to learn an abstract representation of the state space shared across multiple source tasks. This abstracted space captures high-level features that are invariant across tasks, enabling the learning of a unified abstract reward function. The learned reward is then used to train policies in a separate, previously unseen target task without requiring new demonstrations in the target task. We evaluate our approach on multiple environments from Gymnasium and AssistiveGym, demonstrating that the learned abstract rewards consistently support successful policy learning in novel task settings.
Class-Incremental Learning (CIL) aims to enable models to sequentially learn new tasks while retaining knowledge from previous ones. Recently, merging-based pre-trained CIL methods have gained significant attention due to their competitive performance and high inference efficiency. However, most existing approaches decouple training from merging, neglecting the compatibility among task-specific adapters. This incompatibility introduces severe conflicts during integration, resulting in catastrophic forgetting and notable performance degradation. To address this limitation, we propose MOBO, a Merging-Oriented Bi-Level Optimization framework that synergizes the optimization of the current task model with the performance of the final merged model. At the upper level, we introduce a global loss that anticipates the merged model's behavior, guiding the current task parameters toward a solution space that facilitates effective merging. At the lower level, we employ a dynamic weighted merging strategy to optimize merging coefficients and update the merged model. By alternately optimizing the task-specific and merged models, MOBO effectively mitigates the performance loss caused by incompatible adapter integration. Comprehensive experiments on CIFAR-100, CUB-200, ImageNet-R, ImageNet-A, and VTAB demonstrate the superiority of our approach, highlighting its robustness and scalability, particularly on long task sequences.
Large Language Models (LLMs) powerful generative capabilities also pose significant risks, underscoring the need for effective detoxification methods to ensure safer deployment. Due to the polysemantic nature of LLM neurons, recent neuron intervention methods inevitably entangle unrelated concepts, compromising generation quality and interpretability. Sparse Autoencoders (SAEs) have opened new horizons for decomposing model activations into monosemantic features, offering interpretability and targeted feature-level steering. Empirical findings reveal that, despite capturing interpretable features, indiscriminate interventions on toxicity-related features expose the fragility of LLMs, achieving toxicity mitigation at the cost of degraded fluency. Building upon this finding, we propose DeLFE, a lightweight controlled detoxification approach that identifies specific toxic features across model layers and performs targeted interventions on them. DeLFE learns toxicity subspaces from label-guided SAE feature subsets to characterize toxic v.s. non-toxic activation patterns. When auto-completing a response token-by-token, DeLFE tracks the toxicity-triggering risks and steers toxic features away from the subspace via a flow-matching feature transformation. We further design three feature-level strategies that adjust intervention timing and strength to reconstruct the target model’s original activations. Extensive experiments demonstrate that our method achieves strong detoxification effectiveness while maintaining high generation quality across models of varying sizes and diverse base LLMs.
Semi-supervised clustering (SSC) enables personalized clustering under limited user supervision. Given the sparsity of initial user intents, constraint propagation has been proposed as a powerful approach to explore and generate new performance-enhancing constraints. However, existing methods struggle to handle a large number of ``gray samples'' that deviate from initial supervision signals and exhibit ambiguous semantics with indistinct boundaries. Compared with ``distinct samples'' that closely match the supervision, gray samples typically contain richer latent semantics, and accurately identifying their relational types can significantly improve clustering performance. To address this challenge, we propose Adversarially Enhanced Propagation-driven Intent-aware Clustering (AEPIC). Specifically, we design an Adversarially Enhanced Constraint Propagation (AECP) mechanism that leverages global adversarial learning over dual relational links to identify gray samples and expand them into meaningful pseudo-user intents. In addition, an intent-aware regularization strategy integrates these pseudo-user intents into representation learning and clustering optimization, further improving clustering performance. Experiments on 5 benchmark datasets demonstrate that, under sparse supervision, AEPIC consistently outperforms state-of-the-art semi-supervised clustering methods.