Translating wordplay across languages presents unique challenges that have long confounded both professional human translators and machine translation systems. This research proposes a novel approach for translating puns from English to French by combining state-of-the-art large language models with specialized techniques for wordplay generation. Our methodology employs a three-stage approach. First, we establish a baseline using multiple frontier large language models with feedback based on a new contrastive learning dataset. Second, we implement a guided chain-of-thought pipeline with combined phonetic-semantic embeddings. Third, we implement a multi-agent generator-discriminator framework for evaluating and regenerating puns with feedback. Moving beyond the limitations of literal translation, our methodology's primary objective is to capture the linguistic creativity and humor of the source text wordplay, rather than simply duplicating its vocabulary. Our best runs earned first and second place in the CLEF JOKER 2025 Task 2 competition where they were evaluated manually by expert native French speakers. This research addresses a gap between translation studies and computational linguistics by implementing linguistically-informed techniques for wordplay translation, advancing our understanding of how language models can be leveraged to handle the complex interplay between semantic ambiguity, phonetic similarity, and the implicit cultural and linguistic awareness needed for successful humor.
We present a two-speaker automatic speech recognition (ASR) system that combines DiCoW -- a diarization-conditioned variant of Whisper -- with DiariZen, a diarization pipeline built on top of Pyannote. We first evaluate both systems in out-of-domain (OOD) multilingual scenarios without any fine-tuning. In this scenario, DiariZen consistently outperforms the baseline Pyannote diarization model, demonstrating strong generalization. Despite being fine-tuned on English-only data for target-speaker ASR, DiCoW retains solid multilingual performance, indicating that encoder modifications preserve Whisper's multilingual capabilities. We then fine-tune both DiCoW and DiariZen on the MLC-SLM challenge data. The fine-tuned DiariZen continues to outperform the fine-tuned Pyannote baseline, while DiCoW sees further gains from domain adaptation. Our final system achieves a micro-average tcpWER/CER of 16.75% and ranks second in Task 2 of the MLC-SLM challenge. Lastly, we identify several labeling inconsistencies in the training data -- such as missing speech segments and incorrect silence annotations -- which can hinder diarization fine-tuning. We propose simple mitigation strategies to address these issues and improve system robustness.
Speech emotion recognition (SER) in naturalistic conditions presents a significant challenge for the speech processing community. Challenges include disagreement in labeling among annotators and imbalanced data distributions. This paper presents a reproducible framework that achieves superior (top 1) performance in the Emotion Recognition in Naturalistic Conditions Challenge (IS25-SER Challenge) - Task 2, evaluated on the MSP-Podcast dataset. Our system is designed to tackle the aforementioned challenges through multimodal learning, multi-task learning, and imbalanced data handling. Specifically, our best system is trained by adding text embeddings, predicting gender, and including ``Other'' (O) and ``No Agreement'' (X) samples in the training set. Our system's results secured both first and second places in the IS25-SER Challenge, and the top performance was achieved by a simple two-system ensemble.
True intelligence requires active capability acquisition, yet current LLM agents inject pre-defined tool schemas into prompts, reducing models to passive selectors and falling short of robust general-purpose agency. We introduce MCP-Zero, an active agent framework that restores tool discovery autonomy to LLMs themselves. Instead of overwhelming models with all available tools, MCP-Zero enables agents to actively identify capability gaps, and request specific tools on-demand, transforming them from large-scale retrievers into genuine autonomous agents. The framework operates through three core mechanisms: (1) Active Tool Request, where models autonomously generate structured requests specifying their exact tool requirements; (2) Hierarchical Semantic Routing, a two-stage algorithm that matches requests to relevant servers and tools through improved semantic alignment; (3) Iterative Capability Extension, enabling agents to progressively build cross-domain toolchains while maintaining minimal context footprint. We construct MCP-tools, a comprehensive dataset of 308 MCP servers and 2,797 tools from the official Model-Context-Protocol repository. Experiments demonstrate that MCP-Zero preserves agent autonomy while achieving substantial efficiency gains: (i) accurate tool selection from nearly 3k candidates across 248.1k tokens; (ii) 98\% reduction in token consumption on APIBank while maintaining high accuracy; and (iii) consistent multi-turn performance that scales with tool ecosystem growth. This work establishes active tool discovery as a fundamental design pattern for scalable autonomous agent systems.
Visual Speech Recognition (VSR) transcribes speech by analyzing lip movements. Recently, Large Language Models (LLMs) have been integrated into VSR systems, leading to notable performance improvements. However, the potential of LLMs has not been extensively studied, and how to effectively utilize LLMs in VSR tasks remains unexplored. This paper systematically explores how to better leverage LLMs for VSR tasks and provides three key contributions: (1) Scaling Test: We study how the LLM size affects VSR performance, confirming a scaling law in the VSR task. (2) Context-Aware Decoding: We add contextual text to guide the LLM decoding, improving recognition accuracy. (3) Iterative Polishing: We propose iteratively refining LLM outputs, progressively reducing recognition errors. Extensive experiments demonstrate that by these designs, the great potential of LLMs can be largely harnessed, leading to significant VSR performance improvement.
Recently, Vision-Language foundation models like CLIP and ALIGN, which are pre-trained on large-scale data have shown remarkable zero-shot generalization to diverse datasets with different classes and even domains. In this work, we take a step further and analyze whether these models can be adapted to target datasets having very different distributions and classes compared to what these models have been trained on, using only a few labeled examples from the target dataset. In such scenarios, finetuning large pretrained models is challenging due to problems of overfitting as well as loss of generalization, and has not been well explored in prior literature. Since, the pre-training data of such models are unavailable, it is difficult to comprehend the performance on various downstream datasets. First, we try to answer the question: Given a target dataset with a few labelled examples, can we estimate whether further fine-tuning can enhance the performance compared to zero-shot evaluation? by analyzing the common vision-language embedding space. Based on the analysis, we propose a novel prompt-tuning method, PromptMargin for adapting such large-scale VLMs directly on the few target samples. PromptMargin effectively tunes the text as well as visual prompts for this task, and has two main modules: 1) Firstly, we use a selective augmentation strategy to complement the few training samples in each task; 2) Additionally, to ensure robust training in the presence of unfamiliar class names, we increase the inter-class margin for improved class discrimination using a novel Multimodal Margin Regularizer. Extensive experiments and analysis across fifteen target benchmark datasets, with varying degrees of distribution shifts from natural images, shows the effectiveness of the proposed framework over the existing state-of-the-art approaches applied to this setting. github.com/debarshigit/PromptMargin.
Assessing the capacity of Large Language Models (LLMs) to plan and reason within the constraints of interactive environments is crucial for developing capable AI agents. We introduce $\textbf{LLM-BabyBench}$, a new benchmark suite designed specifically for this purpose. Built upon a textual adaptation of the procedurally generated BabyAI grid world, this suite evaluates LLMs on three fundamental aspects of grounded intelligence: (1) predicting the consequences of actions on the environment state ($\textbf{Predict}$ task), (2) generating sequences of low-level actions to achieve specified objectives ($\textbf{Plan}$ task), and (3) decomposing high-level instructions into coherent subgoal sequences ($\textbf{Decompose}$ task). We detail the methodology for generating the three corresponding datasets ($\texttt{LLM-BabyBench-Predict}$, $\texttt{-Plan}$, $\texttt{-Decompose}$) by extracting structured information from an expert agent operating within the text-based environment. Furthermore, we provide a standardized evaluation harness and metrics, including environment interaction for validating generated plans, to facilitate reproducible assessment of diverse LLMs. Initial baseline results highlight the challenges posed by these grounded reasoning tasks. The benchmark suite, datasets, data generation code, and evaluation code are made publicly available ($\href{https://github.com/choukrani/llm-babybench}{\text{GitHub}}$, $\href{https://huggingface.co/datasets/salem-mbzuai/LLM-BabyBench}{\text{HuggingFace}}$).
Recently, large language models(LLMs) have played an increasingly important role in solving a wide range of NLP tasks, leveraging their capabilities of natural language understanding and generating. Integration with external tools further enhances LLMs' effectiveness, providing more precise, timely, and specialized responses. However, LLMs still encounter difficulties with non-executable actions and improper actions, which are primarily attributed to incorrect parameters. The process of generating parameters by LLMs is confined to the tool level, employing the coarse-grained strategy without considering the different difficulties of various tools. To address this issue, we propose TUMS, a novel framework designed to enhance the tool-use capabilities of LLMs by transforming tool-level processing into parameter-level processing. Specifically, our framework consists of four key components: (1) an intent recognizer that identifies the user's intent to help LLMs better understand the task; (2) a task decomposer that breaks down complex tasks into simpler subtasks, each involving a tool call; (3) a subtask processor equipped with multi-structure handlers to generate accurate parameters; and (4) an executor. Our empirical studies have evidenced the effectiveness and efficiency of the TUMS framework with an average of 19.6\% and 50.6\% improvement separately on easy and hard benchmarks of ToolQA, meanwhile, we demonstrated the key contribution of each part with ablation experiments, offering more insights and stimulating future research on Tool-augmented LLMs.
With the rapid advancement of Multimodal Large Language Models (MLLMs), an increasing number of researchers are exploring their application in recommendation systems. However, the high latency associated with large models presents a significant challenge for such use cases. The EReL@MIR workshop provided a valuable opportunity to experiment with various approaches aimed at improving the efficiency of multimodal representation learning for information retrieval tasks. As part of the competition's requirements, participants were mandated to submit a technical report detailing their methodologies and findings. Our team was honored to receive the award for Task 2 - Winner (Multimodal CTR Prediction). In this technical report, we present our methods and key findings. Additionally, we propose several directions for future work, particularly focusing on how to effectively integrate recommendation signals into multimodal representations. The codebase for our implementation is publicly available at: https://github.com/Lattice-zjj/MMCTR_Code, and the trained model weights can be accessed at: https://huggingface.co/FireFlyCourageous/MMCTR_DIN_MicroLens_1M_x1.
Multimodal click-through rate (CTR) prediction is a key technique in industrial recommender systems. It leverages heterogeneous modalities such as text, images, and behavioral logs to capture high-order feature interactions between users and items, thereby enhancing the system's understanding of user interests and its ability to predict click behavior. The primary challenge in this field lies in effectively utilizing the rich semantic information from multiple modalities while satisfying the low-latency requirements of online inference in real-world applications. To foster progress in this area, the Multimodal CTR Prediction Challenge Track of the WWW 2025 EReL@MIR Workshop formulates the problem into two tasks: (1) Task 1 of Multimodal Item Embedding: this task aims to explore multimodal information extraction and item representation learning methods that enhance recommendation tasks; and (2) Task 2 of Multimodal CTR Prediction: this task aims to explore what multimodal recommendation model can effectively leverage multimodal embedding features and achieve better performance. In this paper, we propose a novel model for Task 2, named Quadratic Interest Network (QIN) for Multimodal CTR Prediction. Specifically, QIN employs adaptive sparse target attention to extract multimodal user behavior features, and leverages Quadratic Neural Networks to capture high-order feature interactions. As a result, QIN achieved an AUC of 0.9798 on the leaderboard and ranked second in the competition. The model code, training logs, hyperparameter configurations, and checkpoints are available at https://github.com/salmon1802/QIN.
The WWW 2025 EReL@MIR Workshop Multimodal CTR Prediction Challenge focuses on effectively applying multimodal embedding features to improve click-through rate (CTR) prediction in recommender systems. This technical report presents our 1$^{st}$ place winning solution for Task 2, combining sequential modeling and feature interaction learning to effectively capture user-item interactions. For multimodal information integration, we simply append the frozen multimodal embeddings to each item embedding. Experiments on the challenge dataset demonstrate the effectiveness of our method, achieving superior performance with a 0.9839 AUC on the leaderboard, much higher than the baseline model. Code and configuration are available in our GitHub repository and the checkpoint of our model can be found in HuggingFace.
Translating knowledge-intensive and entity-rich text between English and Korean requires transcreation to preserve language-specific and cultural nuances beyond literal, phonetic or word-for-word conversion. We evaluate 13 models (LLMs and MT models) using automatic metrics and human assessment by bilingual annotators. Our findings show LLMs outperform traditional MT systems but struggle with entity translation requiring cultural adaptation. By constructing an error taxonomy, we identify incorrect responses and entity name errors as key issues, with performance varying by entity type and popularity level. This work exposes gaps in automatic evaluation metrics and hope to enable future work in completing culturally-nuanced machine translation.
Mixture-of-Experts (MoE) have emerged as a powerful architecture for large language models (LLMs), enabling efficient scaling of model capacity while maintaining manageable computational costs. The key advantage lies in their ability to route different tokens to different ``expert'' networks within the model, enabling specialization and efficient handling of diverse input. However, the vulnerabilities of MoE-based LLMs still have barely been studied, and the potential for backdoor attacks in this context remains largely unexplored. This paper presents the first backdoor attack against MoE-based LLMs where the attackers poison ``dormant experts'' (i.e., underutilized experts) and activate them by optimizing routing triggers, thereby gaining control over the model's output. We first rigorously prove the existence of a few ``dominating experts'' in MoE models, whose outputs can determine the overall MoE's output. We also show that dormant experts can serve as dominating experts to manipulate model predictions. Accordingly, our attack, namely BadMoE, exploits the unique architecture of MoE models by 1) identifying dormant experts unrelated to the target task, 2) constructing a routing-aware loss to optimize the activation triggers of these experts, and 3) promoting dormant experts to dominating roles via poisoned training data. Extensive experiments show that BadMoE successfully enforces malicious prediction on attackers' target tasks while preserving overall model utility, making it a more potent and stealthy attack than existing methods.
Motivated by the emerging demand in the financial industry for the automatic analysis of unstructured and structured data at scale, Question Answering (QA) systems can provide lucrative and competitive advantages to companies by facilitating the decision making of financial advisers. Consequently, we propose a novel financial QA system using the transformer-based pre-trained BERT language model to address the limitations of data scarcity and language specificity in the financial domain. Our system focuses on financial non-factoid answer selection, which retrieves a set of passage-level texts and selects the most relevant as the answer. To increase efficiency, we formulate the answer selection task as a re-ranking problem, in which our system consists of an Answer Retriever using BM25, a simple information retrieval approach, to first return a list of candidate answers, and an Answer Re-ranker built with variants of pre-trained BERT language models to re-rank and select the most relevant answers. We investigate various learning, further pre-training, and fine-tuning approaches for BERT. Our experiments suggest that FinBERT-QA, a model built from applying the Transfer and Adapt further fine-tuning and pointwise learning approach, is the most effective, improving the state-of-the-art results of task 2 of the FiQA dataset by 16% on MRR, 17% on NDCG, and 21% on Precision@1.
This report details MERL's system for room impulse response (RIR) estimation submitted to the Generative Data Augmentation Workshop at ICASSP 2025 for Augmenting RIR Data (Task 1) and Improving Speaker Distance Estimation (Task 2). We first pre-train a neural acoustic field conditioned by room geometry on an external large-scale dataset in which pairs of RIRs and the geometries are provided. The neural acoustic field is then adapted to each target room by using the enrollment data, where we leverage either the provided room geometries or geometries retrieved from the external dataset, depending on availability. Lastly, we predict the RIRs for each pair of source and receiver locations specified by Task 1, and use these RIRs to train the speaker distance estimation model in Task 2.
This paper presents our findings for SemEval 2025 Task 2, a shared task on entity-aware machine translation (EA-MT). The goal of this task is to develop translation models that can accurately translate English sentences into target languages, with a particular focus on handling named entities, which often pose challenges for MT systems. The task covers 10 target languages with English as the source. In this paper, we describe the different systems we employed, detail our results, and discuss insights gained from our experiments.
Cardiovascular magnetic resonance (CMR) imaging offers diverse contrasts for non-invasive assessment of cardiac function and myocardial characterization. However, CMR often requires the acquisition of many contrasts, and each contrast takes a considerable amount of time. The extended acquisition time will further increase the susceptibility to motion artifacts. Existing deep learning-based reconstruction methods have been proven to perform well in image reconstruction tasks, but most of them are designed for specific acquisition modality or dedicated imaging parameter, which limits their ability to generalize across a variety of scan scenarios. To address this issue, the CMRxRecon2024 challenge consists of two specific tasks: Task 1 focuses on a modality-universal setting, evaluating the out-of-distribution generalization of existing learning-based models, while Task 2 follows a k-space sampling-universal setting, assessing the all-in-one adaptability of universal models. Main contributions of this challenge include providing the largest publicly available multi-modality, multi-view cardiac k-space dataset; and developing an open benchmarking platform for algorithm evaluation and shared code library for data processing. In addition, through a detailed analysis of the results submitted to the challenge, we have also made several findings, including: 1) adaptive prompt-learning embedding is an effective means for achieving strong generalization in reconstruction models; 2) enhanced data consistency based on physics-informed networks is also an effective pathway toward a universal model; 3) traditional evaluation metrics have limitations when assessing ground-truth references with moderate or lower image quality, highlighting the need for subjective evaluation methods. This challenge attracted 200 participants from 18 countries, aimed at promoting their translation into clinical practice.
Reliable evaluation of AI models is critical for scientific progress and practical application. While existing VLM benchmarks provide general insights into model capabilities, their heterogeneous designs and limited focus on a few imaging domains pose significant challenges for both cross-domain performance comparison and targeted domain-specific evaluation. To address this, we propose three key contributions: (1) a framework for the resource-efficient creation of domain-specific VLM benchmarks enabled by task augmentation for creating multiple diverse tasks from a single existing task, (2) the release of new VLM benchmarks for seven domains, created according to the same homogeneous protocol and including 162,946 thoroughly human-validated answers, and (3) an extensive benchmarking of 22 state-of-the-art VLMs on a total of 37,171 tasks, revealing performance variances across domains and tasks, thereby supporting the need for tailored VLM benchmarks. Adoption of our methodology will pave the way for the resource-efficient domain-specific selection of models and guide future research efforts toward addressing core open questions.
World models that forecast environmental changes from actions are vital for autonomous driving models with strong generalization. The prevailing driving world model mainly build on video prediction model. Although these models can produce high-fidelity video sequences with advanced diffusion-based generator, they are constrained by their predictive duration and overall generalization capabilities. In this paper, we explore to solve this problem by combining generation loss with MAE-style feature-level context learning. In particular, we instantiate this target with three key design: (1) A more scalable Diffusion Transformer (DiT) structure trained with extra mask construction task. (2) we devise diffusion-related mask tokens to deal with the fuzzy relations between mask reconstruction and generative diffusion process. (3) we extend mask construction task to spatial-temporal domain by utilizing row-wise mask for shifted self-attention rather than masked self-attention in MAE. Then, we adopt a row-wise cross-view module to align with this mask design. Based on above improvement, we propose MaskGWM: a Generalizable driving World Model embodied with Video Mask reconstruction. Our model contains two variants: MaskGWM-long, focusing on long-horizon prediction, and MaskGWM-mview, dedicated to multi-view generation. Comprehensive experiments on standard benchmarks validate the effectiveness of the proposed method, which contain normal validation of Nuscene dataset, long-horizon rollout of OpenDV-2K dataset and zero-shot validation of Waymo dataset. Quantitative metrics on these datasets show our method notably improving state-of-the-art driving world model.
Background: The radiation oncology clinical practice involves many steps relying on the dynamic interplay of abundant text data. Large language models have displayed remarkable capabilities in processing complex text information. But their direct applications in specific fields like radiation oncology remain underexplored. Purpose: This study aims to investigate whether fine-tuning LLMs with domain knowledge can improve the performance on Task (1) treatment regimen generation, Task (2) treatment modality selection (photon, proton, electron, or brachytherapy), and Task (3) ICD-10 code prediction in radiation oncology. Methods: Data for 15,724 patient cases were extracted. Cases where patients had a single diagnostic record, and a clearly identifiable primary treatment plan were selected for preprocessing and manual annotation to have 7,903 cases of the patient diagnosis, treatment plan, treatment modality, and ICD-10 code. Each case was used to construct a pair consisting of patient diagnostics details and an answer (treatment regimen, treatment modality, or ICD-10 code respectively) for the supervised fine-tuning of these three tasks. Open source LLaMA2-7B and Mistral-7B models were utilized for the fine-tuning with the Low-Rank Approximations method. Accuracy and ROUGE-1 score were reported for the fine-tuned models and original models. Clinical evaluation was performed on Task (1) by radiation oncologists, while precision, recall, and F-1 score were evaluated for Task (2) and (3). One-sided Wilcoxon signed-rank tests were used to statistically analyze the results. Results: Fine-tuned LLMs outperformed original LLMs across all tasks with p-value <= 0.001. Clinical evaluation demonstrated that over 60% of the fine-tuned LLMs-generated treatment regimens were clinically acceptable. Precision, recall, and F1-score showed improved performance of fine-tuned LLMs.