Ning Wu

Lin Sun, Chuang Liu, Peng Liu et al.

Direct Preference Optimization (DPO) have emerged as a popular method for aligning Large Language Models (LLMs) with human preferences. While DPO effectively preserves the relative ordering between chosen and rejected responses through pairwise ranking losses, it often neglects absolute reward magnitudes. This oversight can decrease the likelihood of chosen responses and increase the risk of generating out-of-distribution responses, leading to poor performance. We term this issue Degraded Chosen Responses (DCR).To address this issue, we propose Balanced Preference Optimization (BPO), a novel framework that dynamically balances the optimization of chosen and rejected responses through two key components: balanced reward margin and gap adaptor. Unlike previous methods, BPO can fundamentally resolve DPO's DCR issue, without introducing additional constraints to the loss function. Experimental results on multiple mathematical reasoning tasks show that BPO significantly outperforms DPO, improving accuracy by +10.1% with Llama-3.1-8B-Instruct (18.8% to 28.9%) and +11.7% with Qwen2.5-Math-7B (35.0% to 46.7%). It also surpasses DPO variants by +3.6% over IPO (43.1%), +5.0% over SLiC (41.7%), and +3.1% over Cal-DPO (43.6%) on the same model. Remarkably, our algorithm requires only a single line of code modification, making it simple to implement and fully compatible with existing DPO-based frameworks.

Yu Zhang, Shujun Peng, Nengwu Wu et al.

Recently, substantial advancements have been made in training language models to carry out step-by-step reasoning for solving intricate numerical reasoning tasks. Beyond the methods used to solve these problems, the structure and formulation of the problems themselves also play a crucial role in determining the performance of large language models. We observe that even small changes in the surface form of mathematical problems can have a profound impact on both the answer distribution and solve rate. This highlights the vulnerability of LLMs to surface-level variations, revealing its limited robustness when reasoning through complex problems. In this paper, we propose RM-PoT, a three-stage framework that integrates problem reformulation (RM), code-aided reasoning (PoT), and domain-aware few-shot learning to address these limitations. Our approach first reformulates the input problem into diverse surface forms to reduce structural bias, then retrieves five semantically aligned examples from a pre-constructed domain-specific question bank to provide contextual guidance, and finally generates executable Python code for precise computation.

Yaobo Liang, Nan Duan, Yeyun Gong et al.

In this paper, we introduce XGLUE, a new benchmark dataset that can be used to train large-scale cross-lingual pre-trained models using multilingual and bilingual corpora and evaluate their performance across a diverse set of cross-lingual tasks. Comparing to GLUE(Wang et al., 2019), which is labeled in English for natural language understanding tasks only, XGLUE has two main advantages: (1) it provides 11 diversified tasks that cover both natural language understanding and generation scenarios; (2) for each task, it provides labeled data in multiple languages. We extend a recent cross-lingual pre-trained model Unicoder(Huang et al., 2019) to cover both understanding and generation tasks, which is evaluated on XGLUE as a strong baseline. We also evaluate the base versions (12-layer) of Multilingual BERT, XLM and XLM-R for comparison.

Jingyuan Wang, Ning Wu, Wayne Xin Zhao et al.

Personalized Route Recommendation (PRR) aims to generate user-specific route suggestions in response to users' route queries. Early studies cast the PRR task as a pathfinding problem on graphs, and adopt adapted search algorithms by integrating heuristic strategies. Although these methods are effective to some extent, they require setting the cost functions with heuristics. In addition, it is difficult to utilize useful context information in the search procedure. To address these issues, we propose using neural networks to automatically learn the cost functions of a classic heuristic algorithm, namely A* algorithm, for the PRR task. Our model consists of two components. First, we employ attention-based Recurrent Neural Networks (RNN) to model the cost from the source to the candidate location by incorporating useful context information. Instead of learning a single cost value, the RNN component is able to learn a time-varying vectorized representation for the moving state of a user. Second, we propose to use a value network for estimating the cost from a candidate location to the destination. For capturing structural characteristics, the value network is built on top of improved graph attention networks by incorporating the moving state of a user and other context information. The two components are integrated in a principled way for deriving a more accurate cost of a candidate location. Extensive experiment results on three real-world datasets have shown the effectiveness and robustness of the proposed model.