Wenhao Huang

Xiang Yue, Yuansheng Ni, Kai Zhang et al. · microsoft-research, princeton

We introduce MMMU: a new benchmark designed to evaluate multimodal models on massive multi-discipline tasks demanding college-level subject knowledge and deliberate reasoning. MMMU includes 11.5K meticulously collected multimodal questions from college exams, quizzes, and textbooks, covering six core disciplines: Art & Design, Business, Science, Health & Medicine, Humanities & Social Science, and Tech & Engineering. These questions span 30 subjects and 183 subfields, comprising 30 highly heterogeneous image types, such as charts, diagrams, maps, tables, music sheets, and chemical structures. Unlike existing benchmarks, MMMU focuses on advanced perception and reasoning with domain-specific knowledge, challenging models to perform tasks akin to those faced by experts. The evaluation of 14 open-source LMMs as well as the proprietary GPT-4V(ision) and Gemini highlights the substantial challenges posed by MMMU. Even the advanced GPT-4V and Gemini Ultra only achieve accuracies of 56% and 59% respectively, indicating significant room for improvement. We believe MMMU will stimulate the community to build next-generation multimodal foundation models towards expert artificial general intelligence.

Xiang Yue, Xingwei Qu, Ge Zhang et al. · microsoft-research

We introduce MAmmoTH, a series of open-source large language models (LLMs) specifically tailored for general math problem-solving. The MAmmoTH models are trained on MathInstruct, our meticulously curated instruction tuning dataset. MathInstruct is compiled from 13 math datasets with intermediate rationales, six of which have rationales newly curated by us. It presents a unique hybrid of chain-of-thought (CoT) and program-of-thought (PoT) rationales, and also ensures extensive coverage of diverse fields in math. The hybrid of CoT and PoT not only unleashes the potential of tool use but also allows different thought processes for different math problems. As a result, the MAmmoTH series substantially outperform existing open-source models on nine mathematical reasoning datasets across all scales with an average accuracy gain between 16% and 32%. Remarkably, our MAmmoTH-7B model reaches 33% on MATH (a competition-level dataset), which exceeds the best open-source 7B model (WizardMath) by 23%, and the MAmmoTH-34B model achieves 44% accuracy on MATH, even surpassing GPT-4's CoT result. Our work underscores the importance of diverse problem coverage and the use of hybrid rationales in developing superior math generalist models.

Ge Zhang, Yemin Shi, Ruibo Liu et al. · deepmind

Instruction tuning is widely recognized as a key technique for building generalist language models, which has attracted the attention of researchers and the public with the release of InstructGPT~\citep{ouyang2022training} and ChatGPT\footnote{\url{https://chat.openai.com/}}. Despite impressive progress in English-oriented large-scale language models (LLMs), it is still under-explored whether English-based foundation LLMs can perform similarly on multilingual tasks compared to English tasks with well-designed instruction tuning and how we can construct the corpora needed for the tuning. To remedy this gap, we propose the project as an attempt to create a Chinese instruction dataset by various methods adapted to the intrinsic characteristics of 4 sub-tasks. We collect around 200k Chinese instruction tuning samples, which have been manually checked to guarantee high quality. We also summarize the existing English and Chinese instruction corpora and briefly describe some potential applications of the newly constructed Chinese instruction corpora. The resulting \textbf{C}hinese \textbf{O}pen \textbf{I}nstruction \textbf{G}eneralist (\textbf{COIG}) corpora are available in Huggingface\footnote{\url{https://huggingface.co/datasets/BAAI/COIG}} and Github\footnote{\url{https://github.com/BAAI-Zlab/COIG}}, and will be continuously updated.

Dongfu Jiang, Yishan Li, Ge Zhang et al. · allen-ai

We present TIGERScore, a \textbf{T}rained metric that follows \textbf{I}nstruction \textbf{G}uidance to perform \textbf{E}xplainable, and \textbf{R}eference-free evaluation over a wide spectrum of text generation tasks. Different from other automatic evaluation methods that only provide arcane scores, TIGERScore is guided by natural language instruction to provide error analysis to pinpoint the mistakes in the generated text. Our metric is based on LLaMA-2, trained on our meticulously curated instruction-tuning dataset MetricInstruct which covers 6 text generation tasks and 23 text generation datasets. The dataset consists of 42K quadruple in the form of (instruction, input, system output $\rightarrow$ error analysis). We collected the `system outputs' through from a large variety of models to cover different types of errors. To quantitatively assess our metric, we evaluate its correlation with human ratings on 5 held-in datasets, 2 held-out datasets and show that TIGERScore can achieve the open-source SoTA correlation with human ratings across these datasets and almost approaches GPT-4 evaluator. As a reference-free metric, its correlation can even surpass the best existing reference-based metrics. To further qualitatively assess the rationale generated by our metric, we conduct human evaluation on the generated explanations and found that the explanations are 70.8\% accurate. Through these experimental results, we believe TIGERScore demonstrates the possibility of building universal explainable metrics to evaluate any text generation task. All the resourced are released in our project website: \url{https://tiger-ai-lab.github.io/TIGERScore/}.

Qianyu He, Jie Zeng, Wenhao Huang et al.

Large language models (LLMs) can understand human instructions, showing their potential for pragmatic applications beyond traditional NLP tasks. However, they still struggle with complex instructions, which can be either complex task descriptions that require multiple tasks and constraints, or complex input that contains long context, noise, heterogeneous information and multi-turn format. Due to these features, LLMs often ignore semantic constraints from task descriptions, generate incorrect formats, violate length or sample count constraints, and be unfaithful to the input text. Existing benchmarks are insufficient to assess LLMs' ability to understand complex instructions, as they are close-ended and simple. To bridge this gap, we propose CELLO, a benchmark for evaluating LLMs' ability to follow complex instructions systematically. We design eight features for complex instructions and construct a comprehensive evaluation dataset from real-world scenarios. We also establish four criteria and develop corresponding metrics, as current ones are inadequate, biased or too strict and coarse-grained. We compare the performance of representative Chinese-oriented and English-oriented models in following complex instructions through extensive experiments. Resources of CELLO are publicly available at https://github.com/Abbey4799/CELLO.

Zhouhong Gu, Xiaoxuan Zhu, Haoning Ye et al.

New Natural Langauge Process~(NLP) benchmarks are urgently needed to align with the rapid development of large language models (LLMs). We present Xiezhi, the most comprehensive evaluation suite designed to assess holistic domain knowledge. Xiezhi comprises multiple-choice questions across 516 diverse disciplines ranging from 13 different subjects with 249,587 questions and accompanied by Xiezhi-Specialty and Xiezhi-Interdiscipline, both with 15k questions. We conduct evaluation of the 47 cutting-edge LLMs on Xiezhi. Results indicate that LLMs exceed average performance of humans in science, engineering, agronomy, medicine, and art, but fall short in economics, jurisprudence, pedagogy, literature, history, and management. We anticipate Xiezhi will help analyze important strengths and shortcomings of LLMs, and the benchmark is released in~\url{https://github.com/MikeGu721/XiezhiBenchmark}.

King Zhu, Qianbo Zang, Shian Jia et al.

Multimodal Large Language Models (MLLMs) are evaluated on various benchmarks, such as image captioning, visual question answering, and reasoning. However, many of these benchmarks include overly simple or uninformative samples, complicating the effective distinction of different MLLMs' performance. Furthermore, evaluating models across numerous benchmarks incurs a significant computational burden. To address these issues, we propose LIME (Less Is More for MLLM Evaluation), a refined and efficient benchmark curated through a semi-automated pipeline. This pipeline filters out uninformative samples and eliminates answer leakage by focusing on tasks that necessitate image-based understanding. Our experiments indicate that LIME reduces the number of samples by 76% and evaluation time by 77%, while also providing a more effective means of distinguishing the capabilities of different models. Notably, we find that traditional automatic metrics, such as CIDEr, are inadequate for assessing MLLMs' captioning performance; excluding the caption task score yields a more accurate reflection of overall model performance. All code and data are available at https://github.com/kangreen0210/LIME.

Yinghao Ma, Anders Øland, Anton Ragni et al.

In recent years, foundation models (FMs) such as large language models (LLMs) and latent diffusion models (LDMs) have profoundly impacted diverse sectors, including music. This comprehensive review examines state-of-the-art (SOTA) pre-trained models and foundation models in music, spanning from representation learning, generative learning and multimodal learning. We first contextualise the significance of music in various industries and trace the evolution of AI in music. By delineating the modalities targeted by foundation models, we discover many of the music representations are underexplored in FM development. Then, emphasis is placed on the lack of versatility of previous methods on diverse music applications, along with the potential of FMs in music understanding, generation and medical application. By comprehensively exploring the details of the model pre-training paradigm, architectural choices, tokenisation, finetuning methodologies and controllability, we emphasise the important topics that should have been well explored, like instruction tuning and in-context learning, scaling law and emergent ability, as well as long-sequence modelling etc. A dedicated section presents insights into music agents, accompanied by a thorough analysis of datasets and evaluations essential for pre-training and downstream tasks. Finally, by underscoring the vital importance of ethical considerations, we advocate that following research on FM for music should focus more on such issues as interpretability, transparency, human responsibility, and copyright issues. The paper offers insights into future challenges and trends on FMs for music, aiming to shape the trajectory of human-AI collaboration in the music realm.

Zekun Wang, King Zhu, Chunpu Xu et al.

In this paper, we introduce MIO, a novel foundation model built on multimodal tokens, capable of understanding and generating speech, text, images, and videos in an end-to-end, autoregressive manner. While the emergence of large language models (LLMs) and multimodal large language models (MM-LLMs) propels advancements in artificial general intelligence through their versatile capabilities, they still lack true any-to-any understanding and generation. Recently, the release of GPT-4o has showcased the remarkable potential of any-to-any LLMs for complex real-world tasks, enabling omnidirectional input and output across images, speech, and text. However, it is closed-source and does not support the generation of multimodal interleaved sequences. To address this gap, we present MIO, which is trained on a mixture of discrete tokens across four modalities using causal multimodal modeling. MIO undergoes a four-stage training process: (1) alignment pre-training, (2) interleaved pre-training, (3) speech-enhanced pre-training, and (4) comprehensive supervised fine-tuning on diverse textual, visual, and speech tasks. Our experimental results indicate that MIO exhibits competitive, and in some cases superior, performance compared to previous dual-modal baselines, any-to-any model baselines, and even modality-specific baselines. Moreover, MIO demonstrates advanced capabilities inherent to its any-to-any feature, such as interleaved video-text generation, chain-of-visual-thought reasoning, visual guideline generation, instructional image editing, etc.

Siwei Wu, Kang Zhu, Yu Bai et al.

Given the remarkable success that large visual language models (LVLMs) have achieved in image perception tasks, the endeavor to make LVLMs perceive the world like humans is drawing increasing attention. Current multi-modal benchmarks primarily focus on facts or specific topic-related knowledge contained within individual images. However, they often overlook the associative relations between multiple images, which require the identification and analysis of similarities among entities or content present in different images. Therefore, we propose the multi-image relation association task and a meticulously curated Multi-granularity Multi-image Relational Association (MMRA) benchmark, comprising 1,024 samples. In order to systematically and comprehensively evaluate current LVLMs, we establish an associational relation system among images that contain 11 subtasks (e.g, UsageSimilarity, SubEvent) at two granularity levels (i.e., image and entity) according to the relations in ConceptNet. Our experiments reveal that on the MMRA benchmark, current multi-image LVLMs exhibit distinct advantages and disadvantages across various subtasks. Notably, fine-grained, entity-level multi-image perception tasks pose a greater challenge for LVLMs compared to image-level tasks. Moreover, LVLMs perform poorly on spatial-related tasks, indicating that LVLMs still have limited spatial awareness. Additionally, our findings indicate that while LVLMs demonstrate a strong capability to perceive image details, enhancing their ability to associate information across multiple images hinges on improving the reasoning capabilities of their language model component. Moreover, we explored the ability of LVLMs to perceive image sequences within the context of our multi-image association task. Our experiments show that the majority of current LVLMs do not adequately model image sequences during the pre-training process.

Wenhao Huang, Jiaqing Liang, Zhixu Li et al.

Information extraction (IE) has been studied extensively. The existing methods always follow a fixed extraction order for complex IE tasks with multiple elements to be extracted in one instance such as event extraction. However, we conduct experiments on several complex IE datasets and observe that different extraction orders can significantly affect the extraction results for a great portion of instances, and the ratio of sentences that are sensitive to extraction orders increases dramatically with the complexity of the IE task. Therefore, this paper proposes a novel adaptive ordered IE paradigm to find the optimal element extraction order for different instances, so as to achieve the best extraction results. We also propose an reinforcement learning (RL) based framework to generate optimal extraction order for each instance dynamically. Additionally, we propose a co-training framework adapted to RL to mitigate the exposure bias during the extractor training phase. Extensive experiments conducted on several public datasets demonstrate that our proposed method can beat previous methods and effectively improve the performance of various IE tasks, especially for complex ones.

Ruibin Yuan, Hanfeng Lin, Yi Wang et al.

While Large Language Models (LLMs) demonstrate impressive capabilities in text generation, we find that their ability has yet to be generalized to music, humanity's creative language. We introduce ChatMusician, an open-source LLM that integrates intrinsic musical abilities. It is based on continual pre-training and finetuning LLaMA2 on a text-compatible music representation, ABC notation, and the music is treated as a second language. ChatMusician can understand and generate music with a pure text tokenizer without any external multi-modal neural structures or tokenizers. Interestingly, endowing musical abilities does not harm language abilities, even achieving a slightly higher MMLU score. Our model is capable of composing well-structured, full-length music, conditioned on texts, chords, melodies, motifs, musical forms, etc, surpassing GPT-4 baseline. On our meticulously curated college-level music understanding benchmark, MusicTheoryBench, ChatMusician surpasses LLaMA2 and GPT-3.5 on zero-shot setting by a noticeable margin. Our work reveals that LLMs can be an excellent compressor for music, but there remains significant territory to be conquered. We release our 4B token music-language corpora MusicPile, the collected MusicTheoryBench, code, model and demo in GitHub.

Zhouhong Gu, Sihang Jiang, Wenhao Huang et al.

The model's ability to understand synonymous expression is crucial in many kinds of downstream tasks. It will make the model to better understand the similarity between context, and more robust to the synonym substitution attack. However, many Pretrained Language Model (PLM) lack synonym knowledge due to limitation of small-scale synsets and PLM's pretraining objectives. In this paper, we propose a framework called Sem4SAP to mine synsets from Open Knowledge Graph (Open-KG) and using the mined synsets to do synonym-aware pretraining for language models. We propose to coarsly filter the content in Open-KG and use the frequency information to better help the clustering process under low-resource unsupervised conditions. We expand the mined synsets by migrating core semantics between synonymous expressions.We also propose two novel and effective synonym-aware pre-training methods for injecting synonym knowledge into PLMs.Extensive experiments demonstrate that Sem4SAP can dramatically outperform the original PLMs and other baselines on ten different tasks.

Jiaheng Liu, Dawei Zhu, Zhiqi Bai et al. · pku

Efficient processing of long contexts has been a persistent pursuit in Natural Language Processing. With the growing number of long documents, dialogues, and other textual data, it is important to develop Long Context Language Models (LCLMs) that can process and analyze extensive inputs in an effective and efficient way. In this paper, we present a comprehensive survey on recent advances in long-context modeling for large language models. Our survey is structured around three key aspects: how to obtain effective and efficient LCLMs, how to train and deploy LCLMs efficiently, and how to evaluate and analyze LCLMs comprehensively. For the first aspect, we discuss data strategies, architectural designs, and workflow approaches oriented with long context processing. For the second aspect, we provide a detailed examination of the infrastructure required for LCLM training and inference. For the third aspect, we present evaluation paradigms for long-context comprehension and long-form generation, as well as behavioral analysis and mechanism interpretability of LCLMs. Beyond these three key aspects, we thoroughly explore the diverse application scenarios where existing LCLMs have been deployed and outline promising future development directions. This survey provides an up-to-date review of the literature on long-context LLMs, which we wish to serve as a valuable resource for both researchers and engineers. An associated GitHub repository collecting the latest papers and repos is available at: \href{https://github.com/LCLM-Horizon/A-Comprehensive-Survey-For-Long-Context-Language-Modeling}{\color[RGB]{175,36,67}{LCLM-Horizon}}.

Yuelin Bai, Xinrun Du, Yiming Liang et al.

Remarkable progress on English instruction tuning has facilitated the efficacy and reliability of large language models (LLMs). However, there remains a noticeable gap in instruction tuning for Chinese, where the complex linguistic features pose significant challenges. Existing datasets, generally distilled from English-centric LLMs, are not well-aligned with Chinese users' interaction patterns. To bridge this gap, we introduce COIG-CQIA, a new Chinese instruction tuning dataset derived from various real-world resources and undergoing rigorous human verification. We conduct extensive experiments on COIG-CQIA, and compare them with strong baseline models and datasets. The experimental results show that models trained on COIG-CQIA achieve highly competitive performance in diverse benchmarks. Additionally, our findings offer several insights for designing effective Chinese instruction-tuning datasets and data-mixing strategies. Our dataset are available at https://huggingface.co/datasets/m-a-p/COIG-CQIA.

Ge Zhang, Xinrun Du, Bei Chen et al.

As the capabilities of large multimodal models (LMMs) continue to advance, evaluating the performance of LMMs emerges as an increasing need. Additionally, there is an even larger gap in evaluating the advanced knowledge and reasoning abilities of LMMs in non-English contexts such as Chinese. We introduce CMMMU, a new Chinese Massive Multi-discipline Multimodal Understanding benchmark designed to evaluate LMMs on tasks demanding college-level subject knowledge and deliberate reasoning in a Chinese context. CMMMU is inspired by and strictly follows the annotation and analysis pattern of MMMU. CMMMU includes 12k manually collected multimodal questions from college exams, quizzes, and textbooks, covering six core disciplines: Art & Design, Business, Science, Health & Medicine, Humanities & Social Science, and Tech & Engineering, like its companion, MMMU. These questions span 30 subjects and comprise 39 highly heterogeneous image types, such as charts, diagrams, maps, tables, music sheets, and chemical structures. CMMMU focuses on complex perception and reasoning with domain-specific knowledge in the Chinese context. We evaluate 11 open-source LLMs and one proprietary GPT-4V(ision). Even GPT-4V only achieves accuracies of 42%, indicating a large space for improvement. CMMMU will boost the community to build the next-generation LMMs towards expert artificial intelligence and promote the democratization of LMMs by providing diverse language contexts.

Alexander Zhang, Marcus Dong, Jiaheng Liu et al.

The critique capacity of Large Language Models (LLMs) is essential for reasoning abilities, which can provide necessary suggestions (e.g., detailed analysis and constructive feedback). Therefore, how to evaluate the critique capacity of LLMs has drawn great attention and several critique benchmarks have been proposed. However, existing critique benchmarks usually have the following limitations: (1). Focusing on diverse reasoning tasks in general domains and insufficient evaluation on code tasks (e.g., only covering code generation task), where the difficulty of queries is relatively easy (e.g., the code queries of CriticBench are from Humaneval and MBPP). (2). Lacking comprehensive evaluation from different dimensions. To address these limitations, we introduce a holistic code critique benchmark for LLMs called CodeCriticBench. Specifically, our CodeCriticBench includes two mainstream code tasks (i.e., code generation and code QA) with different difficulties. Besides, the evaluation protocols include basic critique evaluation and advanced critique evaluation for different characteristics, where fine-grained evaluation checklists are well-designed for advanced settings. Finally, we conduct extensive experimental results of existing LLMs, which show the effectiveness of CodeCriticBench.

Tianyu Zheng, Shuyue Guo, Xingwei Qu et al.

In this paper, we introduce Kun, a novel approach for creating high-quality instruction-tuning datasets for large language models (LLMs) without relying on manual annotations. Adapting a self-training algorithm based on instruction back-translation and answer polishment, Kun leverages unlabelled data from diverse sources such as Wudao, Wanjuan, and SkyPile to generate a substantial dataset of over a million Chinese instructional data points. This approach significantly deviates from traditional methods by using a self-curation process to refine and select the most effective instruction-output pairs. Our experiments with the 6B-parameter Yi model across various benchmarks demonstrate Kun's robustness and scalability. Our method's core contributions lie in its algorithmic advancement, which enhances data retention and clarity, and its innovative data generation approach that substantially reduces the reliance on costly and time-consuming manual annotations. This methodology presents a scalable and efficient solution for improving the instruction-following capabilities of LLMs, with significant implications for their application across diverse fields. The code and dataset can be found at https://github.com/Zheng0428/COIG-Kun

01. AI, Alex Young, Bei Chen et al.

We introduce the Yi model family, a series of language and multimodal models that demonstrate strong multi-dimensional capabilities. The Yi model family is based on 6B and 34B pretrained language models, then we extend them to chat models, 200K long context models, depth-upscaled models, and vision-language models. Our base models achieve strong performance on a wide range of benchmarks like MMLU, and our finetuned chat models deliver strong human preference rate on major evaluation platforms like AlpacaEval and Chatbot Arena. Building upon our scalable super-computing infrastructure and the classical transformer architecture, we attribute the performance of Yi models primarily to its data quality resulting from our data-engineering efforts. For pretraining, we construct 3.1 trillion tokens of English and Chinese corpora using a cascaded data deduplication and quality filtering pipeline. For finetuning, we polish a small scale (less than 10K) instruction dataset over multiple iterations such that every single instance has been verified directly by our machine learning engineers. For vision-language, we combine the chat language model with a vision transformer encoder and train the model to align visual representations to the semantic space of the language model. We further extend the context length to 200K through lightweight continual pretraining and demonstrate strong needle-in-a-haystack retrieval performance. We show that extending the depth of the pretrained checkpoint through continual pretraining further improves performance. We believe that given our current results, continuing to scale up model parameters using thoroughly optimized data will lead to even stronger frontier models.

Ka Man Lo, Yiming Liang, Wenyu Du et al.

Modular neural architectures are gaining attention for their powerful generalization and efficient adaptation to new domains. However, training these models poses challenges due to optimization difficulties arising from intrinsic sparse connectivity. Leveraging knowledge from monolithic models through techniques like knowledge distillation can facilitate training and enable integration of diverse knowledge. Nevertheless, conventional knowledge distillation approaches are not tailored to modular models and struggle with unique architectures and enormous parameter counts. Motivated by these challenges, we propose module-to-module knowledge distillation (m2mKD) for transferring knowledge between modules. m2mKD combines teacher modules of a pretrained monolithic model and student modules of a modular model with a shared meta model respectively to encourage the student module to mimic the behaviour of the teacher module. We evaluate m2mKD on two modular neural architectures: Neural Attentive Circuits (NACs) and Vision Mixture-of-Experts (V-MoE). Applying m2mKD to NACs yields significant improvements in IID accuracy on Tiny-ImageNet (up to 5.6%) and OOD robustness on Tiny-ImageNet-R (up to 4.2%). Additionally, the V-MoE-Base model trained with m2mKD achieves 3.5% higher accuracy than end-to-end training on ImageNet-1k. Code is available at https://github.com/kamanphoebe/m2mKD.

Zhenpeng Zhang, Yi Wang, Shanglei Chai et al.

Lychee is a high-value subtropical fruit. The adoption of vision-based harvesting robots can significantly improve productivity while reduce reliance on labor. High-quality data are essential for developing such harvesting robots. However, there are currently no consistently and comprehensively annotated open-source lychee datasets featuring fruits in natural growing environments. To address this, we constructed a dataset to facilitate lychee detection and maturity classification. Color (RGB) images were acquired under diverse weather conditions, and at different times of the day, across multiple lychee varieties, such as Nuomici, Feizixiao, Heiye, and Huaizhi. The dataset encompasses three different ripeness stages and contains 11,414 images, consisting of 878 raw RGB images, 8,780 augmented RGB images, and 1,756 depth images. The images are annotated with 9,658 pairs of lables for lychee detection and maturity classification. To improve annotation consistency, three individuals independently labeled the data, and their results were then aggregated and verified by a fourth reviewer. Detailed statistical analyses were done to examine the dataset. Finally, we performed experiments using three representative deep learning models to evaluate the dataset. It is publicly available for academic

Haifan Gong, Wenhao Huang, Huan Zhang et al.

Automatic segmentation of the bronchial tree from CT imaging is important, as it provides structural information for disease diagnosis. Despite the merits of previous automatic bronchus segmentation methods, they have paied less attention to the issue we term as \textit{Intensity Confusion}, wherein the intensity values of certain background voxels approach those of the foreground voxels within bronchi. Conversely, the intensity values of some foreground voxels are nearly identical to those of background voxels. This proximity in intensity values introduces significant challenges to neural network methodologies. To address the issue, we introduce a novel Intensity-Distance Guided loss function, which assigns adaptive weights to different image voxels for mining hard samples that cause the intensity confusion. The proposed loss estimates the voxel-level hardness of samples, on the basis of the following intensity and distance priors. We regard a voxel as a hard sample if it is in: (1) the background and has an intensity value close to the bronchus region; (2) the bronchus region and is of higher intensity than most voxels inside the bronchus; (3) the background region and at a short distance from the bronchus. Extensive experiments not only show the superiority of our method compared with the state-of-the-art methods, but also verify that tackling the intensity confusion issue helps to significantly improve bronchus segmentation. Project page: https://github.com/lhaof/ICM.

Leyan Wang, Yonggang Jin, Tianhao Shen et al.

As large language models (LLMs) continue to develop and gain widespread application, the ability of LLMs to exhibit empathy towards diverse group identities and understand their perspectives is increasingly recognized as critical. Most existing benchmarks for empathy evaluation of LLMs focus primarily on universal human emotions, such as sadness and pain, often overlooking the context of individuals' group identities. To address this gap, we introduce GIEBench, a comprehensive benchmark that includes 11 identity dimensions, covering 97 group identities with a total of 999 single-choice questions related to specific group identities. GIEBench is designed to evaluate the empathy of LLMs when presented with specific group identities such as gender, age, occupation, and race, emphasizing their ability to respond from the standpoint of the identified group. This supports the ongoing development of empathetic LLM applications tailored to users with different identities. Our evaluation of 23 LLMs revealed that while these LLMs understand different identity standpoints, they fail to consistently exhibit equal empathy across these identities without explicit instructions to adopt those perspectives. This highlights the need for improved alignment of LLMs with diverse values to better accommodate the multifaceted nature of human identities. Our datasets are available at https://github.com/GIEBench/GIEBench.

Ziqiang Liu, Feiteng Fang, Xi Feng et al.

The rapid advancements in the development of multimodal large language models (MLLMs) have consistently led to new breakthroughs on various benchmarks. In response, numerous challenging and comprehensive benchmarks have been proposed to more accurately assess the capabilities of MLLMs. However, there is a dearth of exploration of the higher-order perceptual capabilities of MLLMs. To fill this gap, we propose the Image Implication understanding Benchmark, II-Bench, which aims to evaluate the model's higher-order perception of images. Through extensive experiments on II-Bench across multiple MLLMs, we have made significant findings. Initially, a substantial gap is observed between the performance of MLLMs and humans on II-Bench. The pinnacle accuracy of MLLMs attains 74.8%, whereas human accuracy averages 90%, peaking at an impressive 98%. Subsequently, MLLMs perform worse on abstract and complex images, suggesting limitations in their ability to understand high-level semantics and capture image details. Finally, it is observed that most models exhibit enhanced accuracy when image sentiment polarity hints are incorporated into the prompts. This observation underscores a notable deficiency in their inherent understanding of image sentiment. We believe that II-Bench will inspire the community to develop the next generation of MLLMs, advancing the journey towards expert artificial general intelligence (AGI). II-Bench is publicly available at https://huggingface.co/datasets/m-a-p/II-Bench.

Liya Zhu, Peizhuang Cong, Aowei Ji et al.

Large Language Models (LLMs) have made rapid progress in reasoning, question answering, and professional applications; however, their true capabilities remain difficult to evaluate using existing benchmarks. Current datasets often focus on simplified tasks or artificial scenarios, overlooking long-tail knowledge and the complexities of real-world applications. To bridge this gap, we propose LPFQA, a long-tail knowledge-based benchmark derived from authentic professional forums across 20 academic and industrial fields, covering 502 tasks grounded in practical expertise. LPFQA introduces four key innovations: fine-grained evaluation dimensions that target knowledge depth, reasoning, terminology comprehension, and contextual analysis; a hierarchical difficulty structure that ensures semantic clarity and unique answers; authentic professional scenario modeling with realistic user personas; and interdisciplinary knowledge integration across diverse domains. We evaluated 12 mainstream LLMs on LPFQA and observed significant performance disparities, especially in specialized reasoning tasks. LPFQA provides a robust, authentic, and discriminative benchmark for advancing LLM evaluation and guiding future model development.

Ruibin Yuan, Hanfeng Lin, Shuyue Guo et al.

We tackle the task of long-form music generation--particularly the challenging \textbf{lyrics-to-song} problem--by introducing YuE, a family of open foundation models based on the LLaMA2 architecture. Specifically, YuE scales to trillions of tokens and generates up to five minutes of music while maintaining lyrical alignment, coherent musical structure, and engaging vocal melodies with appropriate accompaniment. It achieves this through (1) track-decoupled next-token prediction to overcome dense mixture signals, (2) structural progressive conditioning for long-context lyrical alignment, and (3) a multitask, multiphase pre-training recipe to converge and generalize. In addition, we redesign the in-context learning technique for music generation, enabling versatile style transfer (e.g., converting Japanese city pop into an English rap while preserving the original accompaniment) and bidirectional generation. Through extensive evaluation, we demonstrate that YuE matches or even surpasses some of the proprietary systems in musicality and vocal agility. In addition, fine-tuning YuE enables additional controls and enhanced support for tail languages. Furthermore, beyond generation, we show that YuE's learned representations can perform well on music understanding tasks, where the results of YuE match or exceed state-of-the-art methods on the MARBLE benchmark. Keywords: lyrics2song, song generation, long-form, foundation model, music generation

Zhouliang Yu, Ruotian Peng, Keyi Ding et al.

Formal mathematical reasoning remains a critical challenge for artificial intelligence, hindered by limitations of existing benchmarks in scope and scale. To address this, we present FormalMATH, a large-scale Lean4 benchmark comprising 5,560 formally verified problems spanning from high-school Olympiad challenges to undergraduate-level theorems across diverse domains (e.g., algebra, applied mathematics, calculus, number theory, and discrete mathematics). To mitigate the inefficiency of manual formalization, we introduce a novel human-in-the-loop autoformalization pipeline that integrates: (1) specialized large language models (LLMs) for statement autoformalization, (2) multi-LLM semantic verification, and (3) negation-based disproof filtering strategies using off-the-shelf LLM-based provers. This approach reduces expert annotation costs by retaining 72.09% of statements before manual verification while ensuring fidelity to the original natural-language problems. Our evaluation of state-of-the-art LLM-based theorem provers reveals significant limitations: even the strongest models achieve only 16.46% success rate under practical sampling budgets, exhibiting pronounced domain bias (e.g., excelling in algebra but failing in calculus) and over-reliance on simplified automation tactics. Notably, we identify a counterintuitive inverse relationship between natural-language solution guidance and proof success in chain-of-thought reasoning scenarios, suggesting that human-written informal reasoning introduces noise rather than clarity in the formal reasoning settings. We believe that FormalMATH provides a robust benchmark for benchmarking formal mathematical reasoning.

Zepeng Ding, Wenhao Huang, Jiaqing Liang et al.

Relation triple extraction, which outputs a set of triples from long sentences, plays a vital role in knowledge acquisition. Large language models can accurately extract triples from simple sentences through few-shot learning or fine-tuning when given appropriate instructions. However, they often miss out when extracting from complex sentences. In this paper, we design an evaluation-filtering framework that integrates large language models with small models for relational triple extraction tasks. The framework includes an evaluation model that can extract related entity pairs with high precision. We propose a simple labeling principle and a deep neural network to build the model, embedding the outputs as prompts into the extraction process of the large model. We conduct extensive experiments to demonstrate that the proposed method can assist large language models in obtaining more accurate extraction results, especially from complex sentences containing multiple relational triples. Our evaluation model can also be embedded into traditional extraction models to enhance their extraction precision from complex sentences.

Ryan Wong, Jiawei Wang, Junjie Zhao et al.

From professional research to everyday planning, many tasks are bottlenecked by wide-scale information seeking, which is more repetitive than cognitively complex. With the rapid development of Large Language Models (LLMs), automated search agents powered by LLMs offer a promising solution to liberate humans from this tedious work. However, the capability of these agents to perform such "wide-context" collection reliably and completely remains largely unevaluated due to a lack of suitable benchmarks. To bridge this gap, we introduce WideSearch, a new benchmark engineered to evaluate agent reliability on these large-scale collection tasks. The benchmark features 200 manually curated questions (100 in English, 100 in Chinese) from over 15 diverse domains, grounded in real user queries. Each task requires agents to collect large-scale atomic information, which could be verified one by one objectively, and arrange it into a well-organized output. A rigorous five-stage quality control pipeline ensures the difficulty, completeness, and verifiability of the dataset. We benchmark over 10 state-of-the-art agentic search systems, including single-agent, multi-agent frameworks, and end-to-end commercial systems. Most systems achieve overall success rates near 0\%, with the best performer reaching just 5\%. However, given sufficient time, cross-validation by multiple human testers can achieve a near 100\% success rate. These results demonstrate that present search agents have critical deficiencies in large-scale information seeking, underscoring urgent areas for future research and development in agentic search. Our dataset, evaluation pipeline, and benchmark results have been publicly released at https://widesearch-seed.github.io/

Luxi Chen, Zihan Zhou, Min Zhao et al.

Generating flexible-view 3D scenes, including 360° rotation and zooming, from single images is challenging due to a lack of 3D data. To this end, we introduce FlexWorld, a novel framework consisting of two key components: (1) a strong video-to-video (V2V) diffusion model to generate high-quality novel view images from incomplete input rendered from a coarse scene, and (2) a progressive expansion process to construct a complete 3D scene. In particular, leveraging an advanced pre-trained video model and accurate depth-estimated training pairs, our V2V model can generate novel views under large camera pose variations. Building upon it, FlexWorld progressively generates new 3D content and integrates it into the global scene through geometry-aware scene fusion. Extensive experiments demonstrate the effectiveness of FlexWorld in generating high-quality novel view videos and flexible-view 3D scenes from single images, achieving superior visual quality under multiple popular metrics and datasets compared to existing state-of-the-art methods. Qualitatively, we highlight that FlexWorld can generate high-fidelity scenes with flexible views like 360° rotations and zooming. Project page: https://ml-gsai.github.io/FlexWorld.

Yiming Liang, Tianyu Zheng, Xinrun Du et al. · tsinghua

Instruction tuning enhances large language models (LLMs) to follow human instructions across diverse tasks, relying on high-quality datasets to guide behavior. However, these datasets, whether manually curated or synthetically generated, are often narrowly focused and misaligned with the broad distributions captured during pre-training, limiting LLM generalization and effective use of pre-trained knowledge. We propose Aligning Instruction Tuning with Pre-training (AITP), a method that bridges this gap by identifying coverage shortfalls in instruction-tuning datasets and rewriting underrepresented pre-training data into high-quality instruction-response pairs. This approach enriches dataset diversity while preserving task-specific objectives. Evaluations on three fully open LLMs across eight benchmarks demonstrate consistent performance improvements with AITP. Ablations highlight the benefits of adaptive data selection, controlled rewriting, and balanced integration, emphasizing the importance of aligning instruction tuning with pre-training distributions to unlock the full potential of LLMs.

Xiaoxuan Tang, Xinping Lei, Chaoran Zhu et al.

Music-to-Video (M2V) generation for full-length songs faces significant challenges. Existing methods produce short, disjointed clips, failing to align visuals with musical structure, beats, or lyrics, and lack temporal consistency. We propose AutoMV, a multi-agent system that generates full music videos (MVs) directly from a song. AutoMV first applies music processing tools to extract musical attributes, such as structure, vocal tracks, and time-aligned lyrics, and constructs these features as contextual inputs for following agents. The screenwriter Agent and director Agent then use this information to design short script, define character profiles in a shared external bank, and specify camera instructions. Subsequently, these agents call the image generator for keyframes and different video generators for "story" or "singer" scenes. A Verifier Agent evaluates their output, enabling multi-agent collaboration to produce a coherent longform MV. To evaluate M2V generation, we further propose a benchmark with four high-level categories (Music Content, Technical, Post-production, Art) and twelve ine-grained criteria. This benchmark was applied to compare commercial products, AutoMV, and human-directed MVs with expert human raters: AutoMV outperforms current baselines significantly across all four categories, narrowing the gap to professional MVs. Finally, we investigate using large multimodal models as automatic MV judges; while promising, they still lag behind human expert, highlighting room for future work.

M-A-P Team, Xinrun Du, Yifan Yao et al.

Large language models (LLMs) have demonstrated remarkable proficiency in mainstream academic disciplines such as mathematics, physics, and computer science. However, human knowledge encompasses over 200 specialized disciplines, far exceeding the scope of existing benchmarks. The capabilities of LLMs in many of these specialized fields-particularly in light industry, agriculture, and service-oriented disciplines-remain inadequately evaluated. To address this gap, we present SuperGPQA, a comprehensive benchmark that evaluates graduate-level knowledge and reasoning capabilities across 285 disciplines. Our benchmark employs a novel Human-LLM collaborative filtering mechanism to eliminate trivial or ambiguous questions through iterative refinement based on both LLM responses and expert feedback. Our experimental results reveal significant room for improvement in the performance of current state-of-the-art LLMs across diverse knowledge domains (e.g., the reasoning-focused model DeepSeek-R1 achieved the highest accuracy of 61.82% on SuperGPQA), highlighting the considerable gap between current model capabilities and artificial general intelligence. Additionally, we present comprehensive insights from our management of a large-scale annotation process, involving over 80 expert annotators and an interactive Human-LLM collaborative system, offering valuable methodological guidance for future research initiatives of comparable scope.

Shawn Gavin, Tuney Zheng, Jiaheng Liu et al.

The long-context capabilities of large language models (LLMs) have been a hot topic in recent years. To evaluate the performance of LLMs in different scenarios, various assessment benchmarks have emerged. However, as most of these benchmarks focus on identifying key information to answer questions, which mainly requires the retrieval ability of LLMs, these benchmarks can partially represent the reasoning performance of LLMs from large amounts of information. Meanwhile, although LLMs often claim to have context windows of 32k, 128k, 200k, or even longer, these benchmarks fail to reveal the actual supported length of these LLMs. To address these issues, we propose the LongIns benchmark dataset, a challenging long-context instruction-based exam for LLMs, which is built based on the existing instruction datasets. Specifically, in our LongIns, we introduce three evaluation settings: Global Instruction & Single Task (GIST), Local Instruction & Single Task (LIST), and Local Instruction & Multiple Tasks (LIMT). Based on LongIns, we perform comprehensive evaluations on existing LLMs and have the following important findings: (1). The top-performing GPT-4 with 128k context length performs poorly on the evaluation context window of 16k in our LongIns. (2). For the multi-hop reasoning ability of many existing LLMs, significant efforts are still needed under short context windows (less than 4k).

Shun Wang, Ge Zhang, Han Wu et al.

Machine Translation (MT) has developed rapidly since the release of Large Language Models and current MT evaluation is performed through comparison with reference human translations or by predicting quality scores from human-labeled data. However, these mainstream evaluation methods mainly focus on fluency and factual reliability, whilst paying little attention to figurative quality. In this paper, we investigate the figurative quality of MT and propose a set of human evaluation metrics focused on the translation of figurative language. We additionally present a multilingual parallel metaphor corpus generated by post-editing. Our evaluation protocol is designed to estimate four aspects of MT: Metaphorical Equivalence, Emotion, Authenticity, and Quality. In doing so, we observe that translations of figurative expressions display different traits from literal ones.

Zepeng Ding, Ruiyang Ke, Wenhao Huang et al.

Existing research on large language models (LLMs) shows that they can solve information extraction tasks through multi-step planning. However, their extraction behavior on complex sentences and tasks is unstable, emerging issues such as false positives and missing elements. We observe that decomposing complex extraction tasks and extracting them step by step can effectively improve LLMs' performance, and the extraction orders of entities significantly affect the final results of LLMs. This paper proposes a two-stage multi-step method for LLM-based information extraction and adopts the RL framework to execute the multi-step planning. We regard sequential extraction as a Markov decision process, build an LLM-based extraction environment, design a decision module to adaptively provide the optimal order for sequential entity extraction on different sentences, and utilize the DDQN algorithm to train the decision model. We also design the rewards and evaluation metrics suitable for the extraction results of LLMs. We conduct extensive experiments on multiple public datasets to demonstrate the effectiveness of our method in improving the information extraction capabilities of LLMs.