This paper presents UltraEdit, a large-scale ( 4M editing samples), automatically generated dataset for instruction-based image editing. Our key idea is to address the drawbacks in existing image editing datasets like InstructPix2Pix and MagicBrush, and provide a systematic approach to producing massive and high-quality image editing samples: 1) UltraEdit includes more diverse editing instructions by combining LLM creativity and in-context editing examples by human raters; 2) UltraEdit is anchored on real images (photographs or artworks), which offers more diversity and less biases than those purely synthesized by text-to-image models; 3) UltraEdit supports region-based editing with high-quality, automatically produced region annotations. Our experiments show that canonical diffusion-based editing baselines trained on UltraEdit set new records on challenging MagicBrush and Emu-Edit benchmarks, respectively. Our analysis further confirms the crucial role of real image anchors and region-based editing data. The dataset, code, and models will be made public.

The scarcity of high-quality and multi-task singing datasets significantly hinders the development of diverse controllable and personalized singing tasks, as existing singing datasets suffer from low quality, limited diversity of languages and singers, absence of multi-technique information and realistic music scores, and poor task suitability.To tackle these problems, we present GTSinger, a large Global, multi-Technique, free-to-use, high-quality singing corpus with realistic music scores, designed for all singing tasks, along with its benchmarks.Particularly,(1) we collect 80.59 hours of high-quality singing voices, forming the largest recorded singing dataset;(2) 20 professional singers across nine widely spoken languages offer diverse timbres and styles;(3) we provide controlled comparison and phoneme-level annotations of six commonly used singing techniques, helping technique modeling and control;(4) GTSinger offers realistic music scores, assisting real-world musical composition;(5) singing voices are accompanied by manual phoneme-to-audio alignments, global style labels, and 16.16 hours of paired speech for various singing tasks.Moreover, to facilitate the use of GTSinger, we conduct four benchmark experiments: technique-controllable singing voice synthesis, technique recognition, style transfer, and speech-to-singing conversion.

Millions of people rely on search functionality to find and explore content on entertainment platforms. Modern search systems use a combination of candidate generation and ranking approaches, with advanced methods leveraging deep learning and LLM-based techniques to retrieve, generate, and categorize search results. Despite these advancements, search algorithms can still surface inappropriate or irrelevant content due to factors like model unpredictability, metadata errors, or overlooked design flaws. Such issues can misalign with product goals and user expectations, potentially harming user trust and business outcomes. In this work, we introduce an additional monitoring layer using Large Language Models (LLMs) to enhance content moderation. This additional layer flags content if the user did not intend to search for it. This approach serves as a baseline for product quality assurance, with collected feedback used to refine the initial retrieval mechanisms of the search model, ensuring a safer and more reliable user experience.

Despite the importance of aligning language models with human preferences, crowd-sourced human feedback is often noisy -- for example, preferring less desirable responses -- posing a fundamental challenge to alignment. A truly robust alignment objective should yield identical model parameters even under severe label noise, a property known as redescending. We prove that no existing alignment methods satisfy this property. To address this, we propose Hölder-DPO, the first principled alignment loss with a provable redescending property, enabling estimation of the clean data distribution from noisy feedback. The aligned model estimates the likelihood of clean data, providing a theoretically grounded metric for dataset valuation that identifies the location and fraction of mislabels. This metric is gradient-free, enabling scalable and automated human feedback valuation without costly manual verification or clean validation dataset. Hölder-DPO achieves state-of-the-art robust alignment performance while accurately detecting mislabels in controlled datasets. Finally, we apply Hölder-DPO to widely used alignment datasets, revealing substantial noise levels and demonstrating that removing these mislabels significantly improves alignment performance across methods.

The growing deployment of large language models (LLMs) has amplified concerns regarding their inherent biases, raising critical questions about their fairness, safety, and societal impact. However, quantifying LLM bias remains a fundamental challenge, complicated by the ambiguity of what "bias" entails. This challenge grows as new models emerge rapidly and gain widespread use, while introducing potential biases that have not been systematically assessed. In this paper, we propose the Relative Bias framework, a method designed to assess how an LLM's behavior deviates from other LLMs within a specified target domain. We introduce two complementary methodologies: (1) Embedding Transformation analysis, which captures relative bias patterns through sentence representations over the embedding space, and (2) LLM-as-a-Judge, which employs a language model to evaluate outputs comparatively. Applying our framework to several case studies on bias and alignment scenarios following by statistical tests for validation, we find strong alignment between the two scoring methods, offering a systematic, scalable, and statistically grounded approach for comparative bias analysis in LLMs.

Recent Multimodal Large Language Models (MLLMs) achieve promising performance on visual and audio benchmarks independently. However, the ability of these models to process cross-modal information synchronously remains largely unexplored. In this paper, we introduce: 1) Daily-Omni, an Audio-Visual Questioning and Answering benchmark comprising 684 videos of daily life scenarios from diverse sources, rich in both audio and visual information, and featuring 1197 multiple-choice QA pairs across 6 major tasks; 2) Daily-Omni QA Generation Pipeline, which includes automatic annotation, QA generation and QA optimization, significantly improves efficiency for human evaluation and scalability of the benchmark; 3) Daily-Omni-Agent, a training-free agent utilizing open-source Visual Language Model (VLM), Audio Language Model (ALM) and Automatic Speech Recognition (ASR) model to establish a baseline for this benchmark. The results show that current MLLMs still struggle significantly with tasks requiring audio-visual integration, but combining VLMs and ALMs with simple temporal alignment techniques can achieve substantially better performance. Codes and benchmark are available at \href{https://github.com/Lliar-liar/Daily-Omni}{https://github.com/Lliar-liar/Daily-Omni}.

Musical (MSS) source separation of western popular music using non-causal deep learning can be very effective. In contrast, MSS for classical music is an unsolved problem. Classical ensembles are harder to separate than popular music because of issues such as the inherent greater variation in the music; the sparsity of recordings with ground truth for supervised training; and greater ambiguity between instruments. The Cadenza project has been exploring MSS for classical music. This is being done so music can be remixed to improve listening experiences for people with hearing loss. To enable the work, a new database of synthesized woodwind ensembles was created to overcome instrumental imbalances in the EnsembleSet. For the MSS, a set of ConvTasNet models was used with each model being trained to extract a string or woodwind instrument. ConvTasNet was chosen because it enabled both causal and non-causal approaches to be tested. Non-causal approaches have dominated MSS work and are useful for recorded music, but for live music or processing on hearing aids, causal signal processing is needed. The MSS performance was evaluated on the two small datasets (Bach10 and URMP) of real instrument recordings where the ground-truth is available. The performances of the causal and non-causal systems were similar. Comparing the average Signal-to-Distortion (SDR) of the synthesized validation set (6.2 dB causal; 6.9 non-causal), to the real recorded evaluation set (0.3 dB causal, 0.4 dB non-causal), shows that mismatch between synthesized and recorded data is a problem. Future work needs to either gather more real recordings that can be used for training, or to improve the realism and diversity of the synthesized recordings to reduce the mismatch...

The majority of inhabitants in Hong Kong are able to read and write in standard Chinese butuse Cantonese as theprimary spoken language in daily life. Spoken Cantonese can be transcribed into Chinese characters, which constitute the so-called writte n Cantonese. Written Cantonese exhibits significant lexical and grammatical differences from standard written Chinese. The riseof written Cantonese is increasingly evident in thecyber world.The growing interaction between Mandarin speakers and Cantonese sp eak-ers is leading to a clear demand for automatic translation between Chinese and Cantonese. This paper describes a transformer-based neural machine translation (NMT) system for written-Chine se-to-written-Cantonese translation. Given that parallel text data of Chinese and Cantonese are extremely scarce, a major focus of thi s study is on the effort of preparing good amount of training dat a for NMT. In addition to collecting 28K parallel sentences from previous linguistic studies and scattered internet resources, we devise an effective approach to obtaining 72K parallel sentences by automatically extracting pairs of semantically similar senten ces from parallel articles on Chinese Wikipedia and Cantonese Wikip edia. We show that leveraging highly similar sentence pairs minedfrom Wikipedia improves translation performance in all test set s. Our system outperforms Baidu Fanyi's Chinese-to-Cantonese tr ansla-tion on 6 out of 8 test sets in BLEU scores. Translation exampl es reveal that our system is able to capture important linguistic transformations between standard Chinese and spoken Cantonese.

Large Language Models (LLMs) augmented with retrieval mechanisms have demonstrated significant potential in fact-checking tasks by integrating external knowledge. However, their reliability decreases when confronted with conflicting evidence from sources of varying credibility. This paper presents the first systematic evaluation of Retrieval-Augmented Generation (RAG) models for fact-checking in the presence of conflicting evidence. To support this study, we introduce \textbf{CONFACT} (\textbf{Con}flicting Evidence for \textbf{Fact}-Checking) (Dataset available at https://github.com/zoeyyes/CONFACT), a novel dataset comprising questions paired with conflicting information from various sources. Extensive experiments reveal critical vulnerabilities in state-of-the-art RAG methods, particularly in resolving conflicts stemming from differences in media source credibility. To address these challenges, we investigate strategies to integrate media background information into both the retrieval and generation stages. Our results show that effectively incorporating source credibility significantly enhances the ability of RAG models to resolve conflicting evidence and improve fact-checking performance.

Automatically evaluating multimodal generation presents a significant challenge, as automated metrics often struggle to align reliably with human evaluation, especially for complex tasks that involve multiple modalities. To address this, we present MMMG, a comprehensive and human-aligned benchmark for multimodal generation across 4 modality combinations (image, audio, interleaved text and image, interleaved text and audio), with a focus on tasks that present significant challenges for generation models, while still enabling reliable automatic evaluation through a combination of models and programs. MMMG encompasses 49 tasks (including 29 newly developed ones), each with a carefully designed evaluation pipeline, and 937 instructions to systematically assess reasoning, controllability, and other key capabilities of multimodal generation models. Extensive validation demonstrates that MMMG is highly aligned with human evaluation, achieving an average agreement of 94.3%. Benchmarking results on 24 multimodal generation models reveal that even though the state-of-the-art model, GPT Image, achieves 78.3% accuracy for image generation, it falls short on multimodal reasoning and interleaved generation. Furthermore, results suggest considerable headroom for improvement in audio generation, highlighting an important direction for future research.