This paper presents our team's solution to Shared Task 7 of NLPCC-2025, which focuses on sentence-level gender bias detection and mitigation in Chinese. The task aims to promote fairness and con-trollability in natural language generation by automatically detecting, classifying, and mitigating gender bias. To address this challenge, we adopt a fine-tuning approach based on large language models (LLMs), efficiently adapt to the bias detection task via Low-Rank Adaptation (LoRA). In terms of data processing, we construct a more balanced training set to alleviate class imbalance and introduce heterogeneous samples from multiple sources to enhance model generalization. For the detection and classification sub-tasks, we employ a majority voting strategy that integrates outputs from multiple expert models to boost performance. Additionally, to improve bias generation detection and mitigation, we design a multi-temperature sampling mechanism to capture potential variations in bias expression styles. Experimental results demonstrate the effectiveness of our approach in bias detection, classification, and mitigation. Our method ultimately achieves an average score of 47.90%, ranking fourth in the shared task.

Large language models like ChatGPT are increasingly used in classrooms, but they often provide outdated or fabricated information that can mislead students. Retrieval Augmented Generation (RAG) improves reliability of LLMs by grounding responses in external resources. We investigate two accessible RAG paradigms, vector-based retrieval and graph-based retrieval to identify best practices for classroom question answering (QA). Existing comparative studies fail to account for pedagogical factors such as educational disciplines, question types, and practical deployment costs. Using a novel dataset, EduScopeQA, of 3,176 questions across academic subjects, we measure performance on various educational query types, from specific facts to broad thematic discussions. We also evaluate system alignment with a dataset of systematically altered textbooks that contradict the LLM's latent knowledge. We find that OpenAI Vector Search RAG (representing vector-based RAG) performs well as a low-cost generalist, especially for quick fact retrieval. On the other hand, GraphRAG Global excels at providing pedagogically rich answers to thematic queries, and GraphRAG Local achieves the highest accuracy with the dense, altered textbooks when corpus integrity is critical. Accounting for the 10-20x higher resource usage of GraphRAG (representing graph-based RAG), we show that a dynamic branching framework that routes queries to the optimal retrieval method boosts fidelity and efficiency. These insights provide actionable guidelines for educators and system designers to integrate RAG-augmented LLMs into learning environments effectively.

Textual response generation is pivotal for multimodal \mbox{task-oriented} dialog systems, which aims to generate proper textual responses based on the multimodal context. While existing efforts have demonstrated remarkable progress, there still exist the following limitations: 1) \textit{neglect of unstructured review knowledge} and 2) \textit{underutilization of large language models (LLMs)}. Inspired by this, we aim to fully utilize dual knowledge (\textit{i.e., } structured attribute and unstructured review knowledge) with LLMs to promote textual response generation in multimodal task-oriented dialog systems. However, this task is non-trivial due to two key challenges: 1) \textit{dynamic knowledge type selection} and 2) \textit{intention-response decoupling}. To address these challenges, we propose a novel dual knowledge-enhanced two-stage reasoner by adapting LLMs for multimodal dialog systems (named DK2R). To be specific, DK2R first extracts both structured attribute and unstructured review knowledge from external knowledge base given the dialog context. Thereafter, DK2R uses an LLM to evaluate each knowledge type's utility by analyzing LLM-generated provisional probe responses. Moreover, DK2R separately summarizes the intention-oriented key clues via dedicated reasoning, which are further used as auxiliary signals to enhance LLM-based textual response generation. Extensive experiments conducted on a public dataset verify the superiority of DK2R. We have released the codes and parameters.

The spread of fake news, polarizing, politically biased, and harmful content on online platforms has been a serious concern. With large language models becoming a promising approach, however, no study has properly benchmarked their performance across different models, usage methods, and languages. This study presents a comprehensive overview of different Large Language Models adaptation paradigms for the detection of hyperpartisan and fake news, harmful tweets, and political bias. Our experiments spanned 10 datasets and 5 different languages (English, Spanish, Portuguese, Arabic and Bulgarian), covering both binary and multiclass classification scenarios. We tested different strategies ranging from parameter efficient Fine-Tuning of language models to a variety of different In-Context Learning strategies and prompts. These included zero-shot prompts, codebooks, few-shot (with both randomly-selected and diversely-selected examples using Determinantal Point Processes), and Chain-of-Thought. We discovered that In-Context Learning often underperforms when compared to Fine-Tuning a model. This main finding highlights the importance of Fine-Tuning even smaller models on task-specific settings even when compared to the largest models evaluated in an In-Context Learning setup - in our case LlaMA3.1-8b-Instruct,

Context. Code refactoring improves software quality without changing external behavior. Despite its advantages, its benefits are hindered by the considerable cost of time, resources, and continuous effort it demands. Aim. Understanding why developers refactor, and which metrics capture these motivations, may support wider and more effective use of refactoring in practice. Method. We performed a large-scale empirical study to analyze developers refactoring activity, leveraging Large Language Models (LLMs) to identify underlying motivations from version control data, comparing our findings with previous motivations reported in the literature. Results. LLMs matched human judgment in 80% of cases, but aligned with literature-based motivations in only 47%. They enriched 22% of motivations with more detailed rationale, often highlighting readability, clarity, and structural improvements. Most motivations were pragmatic, focused on simplification and maintainability. While metrics related to developer experience and code readability ranked highest, their correlation with motivation categories was weak. Conclusions. We conclude that LLMs effectively capture surface-level motivations but struggle with architectural reasoning. Their value lies in providing localized explanations, which, when combined with software metrics, can form hybrid approaches. Such integration offers a promising path toward prioritizing refactoring more systematically and balancing short-term improvements with long-term architectural goals.

With the widespread application of Mixture of Experts (MoE) reasoning models in the field of LLM learning, efficiently serving MoE models under limited GPU memory constraints has emerged as a significant challenge. Offloading the non-activated experts to main memory has been identified as an efficient approach to address such a problem, while it brings the challenges of transferring the expert between the GPU memory and main memory. We need to explore an efficient approach to compress the expert and analyze how the compression error affects the inference performance. To bridge this gap, we propose employing error-bounded lossy compression algorithms (such as SZ3 and CuSZp) to compress non-activated experts, thereby reducing data transfer overhead during MoE inference. We conduct extensive experiments across various benchmarks and present a comprehensive analysis of how compression-induced errors in different experts affect overall inference accuracy. The results indicate that experts in the shallow layers, which are primarily responsible for the attention mechanism and the transformation of input tokens into vector representations, exhibit minimal degradation in inference accuracy when subjected to bounded errors. In contrast, errors in the middle-layer experts, which are central to model reasoning, significantly impair inference accuracy. Interestingly, introducing bounded errors in the deep-layer experts, which are mainly responsible for instruction following and output integration, can sometimes lead to improvements in inference accuracy.

As large language models (LLMs) reach high scores on established mathematical benchmarks, such as GSM8K and MATH, the research community has turned to International Mathematical Olympiad (IMO) problems to push the evaluation frontier. However, existing Olympiad-level benchmarks suffer from practical constraints that introduce grading noise and potential bias, such as heterogeneous answer formats requiring model-based judges and a reliance on potentially flawed solutions. We introduce RIMO, a two-track benchmark designed to preserve peak Olympiad difficulty while eliminating this evaluation noise. The first track, RIMO-N, rewrites 335 IMO problems to admit a single, unique integer answer, allowing for deterministic correctness checking. The second track, RIMO-P, features 456 proof problems with expert-checked solutions, which are decomposed into a sequence of sub-problems to evaluate the step-by-step reasoning process via an automated grading system. Our benchmarking of ten frontier LLMs, including GPT-4o and Gemini 2.5 Flash, reveals that while these systems excel on older benchmarks, their performance drops sharply on RIMO. These results highlight a substantial gap between current LLM capabilities and actual Olympiad-level reasoning. By providing a challenging yet easy-to-evaluate suite, RIMO offers a high-resolution yardstick for future research, presenting a clear target for closing the profound reasoning gap our findings expose.

In recent years, the field of medical informatics has seen significant advancements with the introduction of medical large language models (LLMs). These models, powered by artificial intelligence, have demonstrated remarkable capabilities in understanding and generating medical text, providing valuable assistance in clinical decision - making, diagnostics, and patient care. Prom inent examples include models such as Meditron [1], BioMistral [2] and OpenBioLLM [3], which have shown considerable promise in various medical applications. However, despite these advancements, the inherent limitations of medical LLMs highlight the need for more robust solutions.

Video understanding has seen significant progress in recent years, with models' performance on perception from short clips continuing to rise. Yet, multiple recent benchmarks, such as LVBench, Neptune, and ActivityNet-RTL, show performance wanes for tasks requiring complex reasoning on videos as queries grow more complex and videos grow longer. In this work, we ask: can existing perception capabilities be leveraged to successfully perform more complex video reasoning? In particular, we develop a large language model agent given access to video modules as subagents or tools. Rather than following a fixed procedure to solve queries as in previous work such as Visual Programming, ViperGPT, and MoReVQA, the agent uses the results of each call to a module to determine subsequent steps. Inspired by work in the textual reasoning domain, we introduce a critic to distinguish between instances of successful and unsuccessful sequences from the agent. We show that the combination of our agent and critic achieve strong performance on the previously-mentioned datasets.

Document Understanding is a foundational AI capability with broad applications, and Document Question Answering (DocQA) is a key evaluation task. Traditional methods convert the document into text for processing by Large Language Models (LLMs), but this process strips away critical multi-modal information like figures. While Large Vision-Language Models (LVLMs) address this limitation, their constrained input size makes multi-page document comprehension infeasible. Retrieval-augmented generation (RAG) methods mitigate this by selecting relevant pages, but they rely solely on semantic relevance, ignoring logical connections between pages and the query, which is essential for reasoning. To this end, we propose MoLoRAG, a logic-aware retrieval framework for multi-modal, multi-page document understanding. By constructing a page graph that captures contextual relationships between pages, a lightweight VLM performs graph traversal to retrieve relevant pages, including those with logical connections often overlooked. This approach combines semantic and logical relevance to deliver more accurate retrieval. After retrieval, the top-$K$ pages are fed into arbitrary LVLMs for question answering. To enhance flexibility, MoLoRAG offers two variants: a training-free solution for easy deployment and a fine-tuned version to improve logical relevance checking. Experiments on four DocQA datasets demonstrate average improvements of 9.68% in accuracy over LVLM direct inference and 7.44% in retrieval precision over baselines. Codes and datasets are released at https://github.com/WxxShirley/MoLoRAG.