In web-based AI services, providers typically host multiple large language models (LLMs) that exhibit diverse capabilities and incur different API costs. Meanwhile, LLM's performance depends not only on its inherent capacity but also on the reasoning strategy it employs, which together influence both answer quality and computational cost. A key challenge is therefore how to adaptively allocate models and strategies to achieve high-quality responses under constrained costs. To address this challenge, we propose Route-To-Reason (RTR), a unified routing framework that simultaneously selects suitable LLMs and reasoning strategies according to query complexity and user budget. Specifically, RTR learns dense vector representations of models and strategies that capture their behavioral characteristics in handling different queries. Leveraging these embeddings, RTR builds a routing table that estimates the cost and performance of different model–strategy pairs. During inference, RTR consults this routing table to dynamically assign the most appropriate pair, enabling adaptive and cost-efficient reasoning tailored to query difficulty and budget scenarios. Extensive experiments across multiple reasoning benchmarks show that RTR achieves comparable or higher accuracy than the best single LLM while substantially reducing both token usage and API cost (by up to 60%), achieving a superior trade-off between performance and efficiency. By lowering the overhead of large-scale LLM inference, RTR contributes to cost-aware and environmentally sustainable deployment of web-based AI services.

With the rapid advancement of deep learning, text-to-speech and voice conversion systems can now generate synthetic speech, commonly known as deepfake speech, that sounds nearly identical to real human voices. This raises serious threats to public security, driving the urgent need for reliable detection methods. However, most existing approaches rely only on either raw waveform or spectrogram features, overlooking cross-view relationships that could reveal artifacts from unknown spoofing attacks. To address this gap, we propose a tri-view collaborative graph learning framework to enhance detection robustness. Our model integrates three complementary views: 1D waveform features, 2D spectral features, and linguistically derived features from an automatic speech recognition (ASR) system. To improve both discriminability and cross-view independence, we design a Tri-View Contrastive Learning (TVCL) framework, which employs cross-view contrastive loss to emphasize cross-view complementarity and intra-view contrastive loss to strengthen class separation within each view. We further introduce a Dynamic Graph Attention Network (DGAT) that captures temporal dependencies across the multi-view features. Through attention-based aggregation and a learnable weighting mechanism, the DGAT adaptively balances contributions from different views, suppressing noise and promoting complementary cooperation. Finally, the graph-level embeddings produced by the DGAT are used for classification. Extensive experiments on five benchmark datasets demonstrate the superiority of our approach, showing consistent improvements over state-of-the-art methods in cross-method, cross-dataset, and cross-language scenarios.

Instruction-tuned Code Large Language Models (Code LLMs) have excelled in diverse code-related tasks, such as program synthesis, automatic program repair, and code explanation. To collect training datasets for instruction-tuning, a popular method involves having models autonomously generate instructions and corresponding responses. However, the direct generation of responses does not ensure functional correctness, a crucial requirement for generating responses to code instructions. To overcome this, we present Verification-Based Self-Play (VERSE), aiming to enhance model proficiency in generating correct responses. VERSE establishes a robust verification framework that covers various code instructions. Employing VERSE, Code LLMs engage in self-play to generate instructions and corresponding verifications. They evaluate execution results and self-consistency as verification outcomes, using them as scores to rank generated data for self-training. Experiments show that VERSE improves multiple base Code LLMs (average 7.6%) across various languages and tasks on many benchmarks, affirming its effectiveness.

Recent development in Large Language Models (LLMs) and Multi-modal Large Language Models (MLLMs) have achieved superior performance and generalization capabilities, covered extensive areas of traditional tasks. However, existing large model training frameworks support only a limited number of models and techniques, particularly lacking in support for new models, which makes fine-tuning LLMs challenging for most developers. Therefore, we develop SWIFT, a customizable one-stop infrastructure for large models. With support of over 350+ LLMs and 80+ MLLMs, SWIFT stands as the open-source framework that provide the most comprehensive support for fine-tuning large models. In particular, it is the first training framework that provides systematic support for MLLMs. Moreover, SWIFT integrates post-training processes such as inference, evaluation, and quantization, to facilitate fast adoptions of large models in various application scenarios, offering helpful utilities like benchmark comparisons among different training techniques.

Despite existing multimodal language models showing impressive performance on the video understanding task, extremely long videos still pose significant challenges to language model's context length, memory consumption, and computational complexity.To address these issues, we propose a vision-language model named Sophia for long video understanding, which can efficiently handle hour-scale long videos.First, we employ a Shot-adaptive Frame Pruning technique, which naturally segments long videos into multiple camera shots, to more sharply identify and focus on the frames relevant to the query.Additionally, we introduce a Hierarchical Attention mechanism to effectively model the long-term temporal dependencies between video frames, which achieves a time and space complexity of O(N ) w.r.t. the input sequence length N while theoretically maintaining the global modeling efficiency.Experimentally, our Sophia exhibits competitive performance compared to existing video understanding baselines across various benchmarks for long video understanding with reduced time and memory consumption.The model code and weights are available at this repository.

The reasoning abilities are one of the most enigmatic and captivating aspects of large language models (LLMs). Numerous studies are dedicated to exploring and expanding the boundaries of this reasoning capability. However, tasks that embody both reasoning and recall characteristics are often overlooked. In this paper, we introduce such a novel task, code reasoning, to provide a new perspective for the reasoning abilities of LLMs. We summarize three meta-benchmarks based on established forms of logical reasoning, and instantiate these into eight specific benchmark tasks. Our testing on these benchmarks reveals that LLMs continue to struggle with identifying satisfactory reasoning pathways. Additionally, we present a new pathway exploration pipeline inspired by human intricate problem-solving methods. This Reflective Hypothesis Decomposition and Amendment (RHDA) pipeline consists of the following iterative steps: (1) Proposing potential hypotheses based on observations and decomposing them; (2) Utilizing tools to validate hypotheses and reflection outcomes; (3) Revising hypothesis in light of observations. Our approach effectively mitigates logical chain collapses arising from forgetting or hallucination issues in multi-step reasoning, resulting in performance gains of up to $3\times$. Finally, we expanded this pipeline by applying it to simulate complex household tasks in real-world scenarios, specifically in VirtualHome, enhancing the handling of failure cases. We release our code and all of results at https://github.com/TnTWoW/code_reasoning.

Programming Knowledge Tracking (PKT) aims to dynamically diagnose learners' mastery levels of programming knowledge based on their coding activities, facilitating more effective and personalized programming education. However, current PKT studies primarily focus on the implicit relationship between code content and knowledge assessment, often overlooking two types of noise signals in long-term programming activities: unwanted signals from unrelated submissions and weak signals from minor modifications. This practical challenge significantly limits model performance and application. To address this issue, we propose Coda, a Code graph-based tuning adaptor designed to enhance existing PKT models by identifying and mitigating the impact of noise. Specifically, Coda first transforms the loose code sequences submitted by each learner into a compact code graph. By leveraging this code graph, unwanted signals can be identified from a semantic similarity perspective. We then apply a cluster-aware GCN to the code graph, which improves the discrimination of weak signals and enables their clustering for identification. Finally, a lightweight yet effective adaptor is incorporated into the PKT task through optimization with two noise feature-based constraints and a navigational regularization term, to correct knowledge states affected by noise. It is worth mentioning that the Coda framework is model-agnostic and can be adapted to most existing PKT solutions. Extensive experimental results on four real-world datasets demonstrate that Coda effectively performs the PKT task in the presence of noisy programming records, outperforming typical baselines.

Code summarization is designed to generate descriptive natural language for code snippets, facilitating understanding and increasing productivity for developers. Previous research often overlooks the semantic connection between code and its natural language description, resulting in a noticeable gap and suboptimal solution. To address this issue, we introduce a semantic-aligned code summarization framework that leverages crucial data flow information from code for semantic analysis, ensuring alignment between code and summaries. Specifically, we utilize a semantic extraction module (SEM) to decipher the meaning of code and align it with natural language through a semantic alignment module. In the SEM, we construct a code graph that includes data flow edges using static program analysis techniques. Then, on this well-constructed code graph, we innovatively adopt a walking algorithm guided by data flow to extract the semantics of the code. This walking algorithm understands code semantics by analyzing the information transfer between variables during the program execution process. In the semantic alignment module, we integrate a contrastive learning loss mechanism for semantic alignment, which cohesively maps the semantic domains of code and natural language into a unified vector space. We further theoretically analyzed that the data-flow-guided walking algorithm can ensure capturing semantically highly related nodes in shorter paths. Extensive experiments on two benchmark datasets demonstrate the efficacy and broad applicability of the framework.

Multilingual code retrieval aims to find code snippets relevant to a user's query from a multilingual codebase, which plays a crucial role in software development and expands their application scenarios compared to classical monolingual code retrieval. Despite the performance improvements achieved by previous studies, two crucial problems are overlooked in the multilingual scenario. First, certain programming languages face data scarcity in specific domains, resulting in limited representation capabilities within those domains. Second, different programming languages can be used interchangeably within the same domain, making it challenging for multilingual models to accurately identify the intended programming language of a user's query. To address these issues, we propose the CommONalities and SpecIalties Driven Multilingual CodE Retrieval Framework (CONSIDER), which includes two modules. The first module enhances the representation of various programming languages by modeling pairwise and global commonalities among them. The second module introduces a novel contrastive learning negative sampling algorithm that leverages language confusion to automatically extract specific language features. Through our experiments, we confirm the significant benefits of our model in real-world multilingual code retrieval scenarios in various aspects. Furthermore, an evaluation demonstrates the effectiveness of our proposed CONSIDER framework in monolingual scenarios as well. Our source code is available at https://github.com/smsquirrel/consider.

Pediatric chest X-rays (CXRs) are crucial for diagnosing respiratory diseases in children. However, most deep learning models perform poorly on pediatric data due to domain gaps, as they are primarily trained on adult datasets with limited pediatric samples. Recent large models, though effective, still struggle with domain-specific terminology and specialized medical reasoning essential for pediatric diagnoses. To address these challenges, we propose PRIDE: a two-stage Pediatric Radiological Image Diagnosis Engine guided by medical knowledge. PRIDE works consistently with real clinical workflows by first gathering medical evidence from patient data, followed by applying clinical knowledge to enhance diagnostic accuracy. In the Multi-source Radiological Findings Recognition Stage, PRIDE integrates insights from generalized medical models and fine-tuned adult and pediatric radiological models to provide comprehensive findings results. In the Knowledge and Evidence-guided Diagnosis Stage, a Multimodal Large Language Model (MLLM) acts as a pediatric clinician, making diagnostic decisions using CXRs, radiological findings, demographic data, and clinical knowledge. Our evaluation on the VinDr-PCXR pediatric dataset demonstrates that PRIDE outperforms existing methods. Ablation studies further confirm the importance and effectiveness of its key components.