Zhihao Luo, Wentao Yan, Jingyu Gong, Min Wang, Zhizhong Zhang, Xuhong Wang, Yuan Xie, Xin Tan. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Large language models with search capabilities frequently exhibit miscalibrated confidence, producing incorrect answers with high certainty.We present Deliberative Searcher, a reasoning-primary framework that integrates search operations into chain-of-thought generation while maintaining explicit confidence calibration.Our method employs constrained reinforcement learning with adaptive Lagrangian multipliers to jointly optimize correctness and reliability.Experiments across five benchmarks demonstrate substantial improvements: our 7B model reduces average false-certain rates from 54% in baselines to 2%, while our 72B variant achieves competitive accuracy with closedsource models and reduces false-certain rates to 9%.The well-calibrated confidence scores also enable more efficient test-time compute: instead of standard majority voting, we use confidence-weighted aggregation and match the performance of 16-sample majority voting with only 4 samples, a 4× reduction in inference compute.These results establish calibrated confidence as a foundation for both trustworthy outputs and adaptive test-time compute, demonstrating the value of the proposed constrained RL framework in search-augmented language models.

Existing Protein Language Models (PLMs) often suffer from limited adaptability to multiple tasks and exhibit poor generalization across diverse biological contexts. In contrast, general-purpose Large Language Models (LLMs) lack the capability to interpret protein sequences and fall short in domain-specific knowledge, limiting their capacity for effective biosemantic reasoning. To combine the advantages of both, we propose BioBridge, a domain-adaptive continual pretraining framework for protein understanding. This framework employs Domain-Incremental Continual Pre-training (DICP) to infuse protein domain knowledge and general reasoning corpus into a LLM simultaneously, effectively mitigating catastrophic forgetting. Cross-modal alignment is achieved via a PLM-ProjectorLLM pipeline, which maps protein sequence embeddings into the semantic space of the language model. Ultimately, an end-toend optimization is adopted to uniformly support various tasks, including protein property prediction and knowledge questionanswering. Our proposed BioBridge demonstrates performance comparable to that of mainstream PLMs on multiple protein benchmarks, such as EC and BindingDB. It also achieves results on par with LLMs on general understanding tasks like MMLU and RACE. This showcases its innovative advantage of combining domain-specific adaptability with general-purpose language competency.

Recent advancements in large language models (LLMs) have shifted the post-training paradigm from traditional instruction tuning and human preference alignment toward reinforcement learning (RL) focused on reasoning capabilities.However, numerous technical reports indicate that purely rule-based reward RL frequently results in poor-quality reasoning chains or inconsistencies between reasoning processes and final answers, particularly when the base model is of smaller scale.During the RL exploration process, models might employ lowquality reasoning chains due to the lack of knowledge, occasionally producing correct answers randomly and receiving rewards based on established rule-based judges.This constrains the potential for resource-limited organizations to conduct direct reinforcement learning training on smaller-scale models.We propose a novel confidence-based reward model tailored for enhancing STEM reasoning capabilities.Unlike conventional approaches, our model penalizes not only incorrect answers but also low-confidence correct responses, thereby promoting more robust and logically consistent reasoning.We validate the effectiveness of our approach through static evaluations, Bestof-N inference tests, and PPO-based RL training.Our method outperforms several stateof-the-art open-source reward models across diverse STEM benchmarks.We release our codes and model in https://github.com/ qianxiHe147/C2RM.

In the digital era, data has become a pivotal asset, advancing technologies such as autonomous driving. Despite this, data trading faces challenges like the absence of robust pricing methods and the lack of trustworthy trading mechanisms. To address these challenges, we introduce a traffic-oriented data trading platform named Data on The Move (DTM), integrating traffic simulation, data trading, and Artificial Intelligent (AI) agents. The DTM platform supports evident-based data value evaluation and AI-based trading mechanisms. Leveraging the common sense capabilities of Large Language Models (LLMs) to assess traffic state and data value, DTM can determine reasonable traffic data pricing through multi-round interaction and simulations. Moreover, DTM provides a pricing method validation by simulating traffic systems, multi-agent interactions, and the heterogeneity and irrational behaviors of individuals in the trading market. Within the DTM platform, entities such as connected vehicles and traffic light controllers could engage in information collecting, data pricing, trading, and decision-making. Simulation results demonstrate that our proposed AI agent-based pricing approach enhances data trading by offering rational prices, as evidenced by the observed improvement in traffic efficiency. This underscores the effectiveness and practical value of DTM, offering new perspectives for the evolution of data markets and smart cities. To the best of our knowledge, this is the first study employing LLMs in data pricing and a pioneering data trading practice in the field of intelligent vehicles and smart cities.

To capture higher-order structural features, most GNN-based algorithms learn node representations incorporating k-hop neighbors' information. Due to the high time complexity of querying k-hop neighbors, most graph algorithms cannot be deployed in a giant dense temporal network to execute millisecond-level inference. This problem dramatically limits the potential of applying graph algorithms in certain areas, especially financial fraud detection. Therefore, we propose Asynchronous Propagation Attention Network, an asynchronous continuous time dynamic graph algorithm for real-time temporal graph embedding. Traditional graph models usually execute two serial operations: first graph querying and then model inference. Different from previous graph algorithms, we decouple model inference and graph computation to alleviate the damage of the heavy graph query operation to the speed of model inference. Extensive experiments demonstrate that the proposed method can achieve competitive performance while greatly improving the inference speed. The source code is published at a Github repository.

Accurate identification of the miRNA-disease associations (MDAs) helps to understand the etiology and mechanisms of various diseases. However, the experimental methods are costly and time-consuming. Thus, it is urgent to develop computational methods towards the prediction of MDAs. Based on the graph theory, the MDA prediction is regarded as a node classification task in the present study. To solve this task, we propose a novel method MDA-GCNFTG, which predicts MDAs based on Graph Convolutional Networks (GCNs) via graph sampling through the Feature and Topology Graph to improve the training efficiency and accuracy. This method models both the potential connections of feature space and the structural relationships of MDA data. The nodes of the graphs are represented by the disease semantic similarity, miRNA functional similarity and Gaussian interaction profile kernel similarity. Moreover, we considered six tasks simultaneously on the MDA prediction problem at the first time, which ensure that under both balanced and unbalanced sample distribution, MDA-GCNFTG can predict not only new MDAs but also new diseases without known related miRNAs and new miRNAs without known related diseases. The results of 5-fold cross-validation show that the MDA-GCNFTG method has achieved satisfactory performance on all six tasks and is significantly superior to the classic machine learning methods and the state-of-the-art MDA prediction methods. Moreover, the effectiveness of GCNs via the graph sampling strategy and the feature and topology graph in MDA-GCNFTG has also been demonstrated. More importantly, case studies for two diseases and three miRNAs are conducted and achieved satisfactory performance.

In order to identify the disturbance signal in power system and reduce the influence on system security, a voltage disturbance signal classifier based on improved local mean decomposition (ILMD) and BP neural network is proposed. ILMD is used to decompose the disturbance signal in three layers, and the product function (PF) component with amplitude-frequency information of voltage signal is obtained. The signal energy value constructed by PF component is used as the input of BP neural network to identify and classify the voltage disturbance signal. Experiments on four typical voltage disturbance signals show that the signal classifiers based on ILMD and BP neural networks have high accuracy and good working efficiency for the recognition and classification of voltage disturbance signals.