How to Guide LLM Generation: Dual-Surrogate Guided Search for Automated Heuristic Design

Large language models (LLMs) have made automated heuristic design (AHD) increasingly practical by generating executable heuristic code from task descriptions and evaluator feedback. Yet under a limited query and evaluation budget, search efficiency depends critically on a pre-generation decision. Before each LLM query and black-box evaluation, the system must choose which archived heuristics to reuse as parents and which generation operator should transform them. Existing methods typically choose such actions with predefined rules, leaving the expected outcome of each concrete operator-parent action only indirectly modeled. Therefore, we propose \emph{\fullmethod{}} (\method{}), a surrogate-guided action-selection module for operator-parent selection in LLM-based AHD. \method{} guides the LLM code-generation process by scoring pre-generation actions with two complementary surrogates. Specifically, a transition surrogate is proposed to predict the latent distribution of the child representation induced by an operator-parent action, while an instance-conditioned utility surrogate is proposed to estimate the expected performance of sampled child latents. Moreover, we propose an uncertainty-aware acquisition rule that combines predicted utility, utility uncertainty, and transition uncertainty to select the next LLM generation action. Across a diverse heuristic-design suite, \method{} is competitive with strong LLM-AHD baselines, and ablation and action-selection analyses suggest that its behavior goes beyond simple archive ranking or fixed operator preferences.

Semantics-Aware Bilevel Co-Evolution: Towards Automated Multicomponent Algorithm Design

LLM-assisted evolutionary search (LES) has emerged as a promising paradigm for automated algorithm design. However, existing methods usually suffer from two inherent limitations when facing the automated design of real-world complex algorithms that usually consist of multiple components. The first limitation is that they either focus on modifying entire algorithms, making it difficult to reuse high-quality components, or concentrate on component refinement within a limited set of predefined multicomponent configurations. The second limitation is the insufficient explicit modeling and exploitation of algorithm semantics. These limitations severely degrade search efficiency and hinder effective exploration of complex design spaces. Therefore, this paper proposes STABLE (Semantics-Aware Bilevel Co-Evolution), an LES method purpose-built for automated multicomponent algorithm design that introduces structural algorithm formulation and semantics-driven evolution. In STABLE, complex algorithms are organized into hierarchical and modular architectures rooted in domain knowledge, aligning the search space with their intrinsic compositional traits. Based on this structured algorithm formulation, STABLE simultaneously optimizes high-level multicomponent configurations and low-level functional components, enabling coordinated cross-level updates while maintaining suitable granularities for design space exploration. At each level, STABLE establishes a multi-faceted semantic model to assist LLMs in capturing structural correlations, functional compatibilities, and inherent rationalities among algorithm components. This semantic model serves as the core guidance for evolutionary search, enabling principled algorithm generation and algorithm evaluation. Extensive experiments demonstrate that STABLE outperform both human-designed baselines and those from advanced LES methods.

Towards Adaptive Continual Model Merging via Manifold-Aware Expert Evolution

Continual Model Merging (CMM) sequentially integrates task-specific models into a unified architecture without intensive retraining. However, existing CMM methods are hindered by a fundamental saturation-redundancy dilemma: backbone-centric approaches face parameter saturation and representation interference within fixed capacities, whereas Mixture-of-Experts (MoE) variants resort to indiscriminate expansion, incurring expert redundancy and a routing bottleneck reliant on additional data-driven optimization. To resolve these challenges, we propose MADE-IT (Manifold-Aware Dynamic Expert Evolution and Implicit rouTing), an adaptive CMM method that orchestrates expert management and activation by grounding intrinsic expert representations in manifold geometry. We introduce a projection-based subspace affinity metric coupled with a distribution-aware adaptive threshold mechanism to guide autonomous expert evolution, harmonizing diversity with architectural parsimony. Furthermore, to bypass parameterized gating networks, we design a data-free and training-free implicit routing mechanism that activates experts via feature-subspace alignment. Extensive experiments demonstrate that MADE-IT consistently outperforms strong baselines in accuracy and robustness across long-horizon and shuffled task sequences, while significantly pruning redundant experts, particularly within generic modules and early layers.

VecFormer: Towards Efficient and Generalizable Graph Transformer with Graph Token Attention

Graph Transformer has demonstrated impressive capabilities in the field of graph representation learning. However, existing approaches face two critical challenges: (1) most models suffer from exponentially increasing computational complexity, making it difficult to scale to large graphs; (2) attention mechanisms based on node-level operations limit the flexibility of the model and result in poor generalization performance in out-of-distribution (OOD) scenarios. To address these issues, we propose \textbf{VecFormer} (the \textbf{Vec}tor Quantized Graph Trans\textbf{former}), an efficient and highly generalizable model for node classification, particularly under OOD settings. VecFormer adopts a two-stage training paradigm. In the first stage, two codebooks are used to reconstruct the node features and the graph structure, aiming to learn the rich semantic \texttt{Graph Codes}. In the second stage, attention mechanisms are performed at the \texttt{Graph Token} level based on the transformed cross codebook, reducing computational complexity while enhancing the model's generalization capability. Extensive experiments on datasets of various sizes demonstrate that VecFormer outperforms the existing Graph Transformer in both performance and speed.

Fine-Grained Model Merging via Modular Expert Recombination

Model merging constructs versatile models by integrating task-specific models without requiring labeled data or expensive joint retraining. Although recent methods improve adaptability to heterogeneous tasks by generating customized merged models for each instance, they face two critical limitations. First, the instance-specific merged models lack reusability, restricting the exploitation of high-quality merging configurations and efficient batch inference. Second, these methods treat each task-specific model as a monolithic whole, overlooking the diverse mergeability of homologous components such as attention and multilayer perceptron layers, and the differing merging sensitivities across components. To address these limitations, we propose MERGE (\underline{M}odular \underline{E}xpert \underline{R}ecombination for fine-\underline{G}rained m\underline{E}rging), a method that enables component-wise model merging and input-aware, on-demand module recombination at inference. MERGE formulates component-wise merging as a bi-objective optimization problem that balances cross-task performance and storage efficiency, and develops a surrogate-assisted evolutionary algorithm to efficiently identify Pareto-optimal merging configurations. These high-quality configurations underpin a reusable modular expert library, from which a lightweight routing network dynamically activates and recombines modular experts to assemble input-specific models and enable efficient inference under storage constraints. Extensive experiments across various model scales, task types, and fine-tuning strategies demonstrate that MERGE consistently outperforms strong baselines and generalizes effectively.

Cooper: Co-Optimizing Policy and Reward Models in Reinforcement Learning for Large Language Models

Large language models (LLMs) have demonstrated remarkable performance in reasoning tasks, where reinforcement learning (RL) serves as a key algorithm for enhancing their reasoning capabilities. Currently, there are two mainstream reward paradigms: model-based rewards and rule-based rewards. However, both approaches suffer from limitations: rule-based rewards lack robustness, while model-based rewards are vulnerable to reward hacking. To address these issues, we propose Cooper(Co-optimizing Policy Model and Reward Model), a RL framework that jointly optimizes both the policy model and the reward model. Cooper leverages the high precision of rule-based rewards when identifying correct responses, and dynamically constructs and selects positive-negative sample pairs for continued training the reward model. This design enhances robustness and mitigates the risk of reward hacking. To further support Cooper, we introduce a hybrid annotation strategy that efficiently and accurately generates training data for the reward model. We also propose a reference-based reward modeling paradigm, where the reward model takes a reference answer as input. Based on this design, we train a reward model named VerifyRM, which achieves higher accuracy on VerifyBench compared to other models of the same size. We conduct reinforcement learning using both VerifyRM and Cooper. Our experiments show that Cooper not only alleviates reward hacking but also improves end-to-end RL performance, for instance, achieving a 0.54% gain in average accuracy on Qwen2.5-1.5B-Instruct. Our findings demonstrate that dynamically updating reward model is an effective way to combat reward hacking, providing a reference for better integrating reward models into RL.

When Large Language Models Meet Law: Dual-Lens Taxonomy, Technical Advances, and Ethical Governance

This paper establishes the first comprehensive review of Large Language Models (LLMs) applied within the legal domain. It pioneers an innovative dual lens taxonomy that integrates legal reasoning frameworks and professional ontologies to systematically unify historical research and contemporary breakthroughs. Transformer-based LLMs, which exhibit emergent capabilities such as contextual reasoning and generative argumentation, surmount traditional limitations by dynamically capturing legal semantics and unifying evidence reasoning. Significant progress is documented in task generalization, reasoning formalization, workflow integration, and addressing core challenges in text processing, knowledge integration, and evaluation rigor via technical innovations like sparse attention mechanisms and mixture-of-experts architectures. However, widespread adoption of LLM introduces critical challenges: hallucination, explainability deficits, jurisdictional adaptation difficulties, and ethical asymmetry. This review proposes a novel taxonomy that maps legal roles to NLP subtasks and computationally implements the Toulmin argumentation framework, thus systematizing advances in reasoning, retrieval, prediction, and dispute resolution. It identifies key frontiers including low-resource systems, multimodal evidence integration, and dynamic rebuttal handling. Ultimately, this work provides both a technical roadmap for researchers and a conceptual framework for practitioners navigating the algorithmic future, laying a robust foundation for the next era of legal artificial intelligence. We have created a GitHub repository to index the relevant papers: https://github.com/Kilimajaro/LLMs_Meet_Law.

Diversity-Aware Policy Optimization for Large Language Model Reasoning

The reasoning capabilities of large language models (LLMs) have advanced rapidly, particularly following the release of DeepSeek R1, which has inspired a surge of research into data quality and reinforcement learning (RL) algorithms. Despite the pivotal role diversity plays in RL, its influence on LLM reasoning remains largely underexplored. To bridge this gap, this work presents a systematic investigation into the impact of diversity in RL-based training for LLM reasoning, and proposes a novel diversity-aware policy optimization method. Across evaluations on 12 LLMs, we observe a strong positive correlation between the solution diversity and Potential at k (a novel metric quantifying an LLM's reasoning potential) in high-performing models. This finding motivates our method to explicitly promote diversity during RL training. Specifically, we design a token-level diversity and reformulate it into a practical objective, then we selectively apply it to positive samples. Integrated into the R1-zero training framework, our method achieves a 3.5 percent average improvement across four mathematical reasoning benchmarks, while generating more diverse and robust solutions.

Semi-Supervised Multi-Label Feature Selection with Consistent Sparse Graph Learning

In practical domains, high-dimensional data are usually associated with diverse semantic labels, whereas traditional feature selection methods are designed for single-label data. Moreover, existing multi-label methods encounter two main challenges in semi-supervised scenarios: (1). Most semi-supervised methods fail to evaluate the label correlations without enough labeled samples, which are the critical information of multi-label feature selection, making label-specific features discarded. (2). The similarity graph structure directly derived from the original feature space is suboptimal for multi-label problems in existing graph-based methods, leading to unreliable soft labels and degraded feature selection performance. To overcome them, we propose a consistent sparse graph learning method for multi-label semi-supervised feature selection (SGMFS), which can enhance the feature selection performance by maintaining space consistency and learning label correlations in semi-supervised scenarios. Specifically, for Challenge (1), SGMFS learns a low-dimensional and independent label subspace from the projected features, which can compatibly cross multiple labels and effectively achieve the label correlations. For Challenge (2), instead of constructing a fixed similarity graph for semi-supervised learning, SGMFS thoroughly explores the intrinsic structure of the data by performing sparse reconstruction of samples in both the label space and the learned subspace simultaneously. In this way, the similarity graph can be adaptively learned to maintain the consistency between label space and the learned subspace, which can promote propagating proper soft labels for unlabeled samples, facilitating the ultimate feature selection. An effective solution with fast convergence is designed to optimize the objective function. Extensive experiments validate the superiority of SGMFS.

Hybrid Local Causal Discovery

Local causal discovery aims to learn and distinguish the direct causes and effects of a target variable from observed data. Existing constraint-based local causal discovery methods use AND or OR rules in constructing the local causal skeleton, but using either rule alone is prone to produce cascading errors in the learned local causal skeleton, and thus impacting the inference of local causal relationships. On the other hand, directly applying score-based global causal discovery methods to local causal discovery may randomly return incorrect results due to the existence of local equivalence classes. To address the above issues, we propose a Hybrid Local Causal Discovery algorithm, called HLCD. Specifically, HLCD initially utilizes a constraint-based approach combined with the OR rule to obtain a candidate skeleton and then employs a score-based method to eliminate redundant portions in the candidate skeleton. Furthermore, during the local causal orientation phase, HLCD distinguishes between V-structures and equivalence classes by comparing the local structure scores between the two, thereby avoiding orientation interference caused by local equivalence classes. We conducted extensive experiments with seven state-of-the-art competitors on 14 benchmark Bayesian network datasets, and the experimental results demonstrate that HLCD significantly outperforms existing local causal discovery algorithms.