Bo-Wen Zhang

Zhongzhi Chen, Guang Liu, Bo-Wen Zhang et al. · meta-ai

In this work, we present a conceptually simple and effective method to train a strong bilingual/multilingual multimodal representation model. Starting from the pre-trained multimodal representation model CLIP released by OpenAI, we altered its text encoder with a pre-trained multilingual text encoder XLM-R, and aligned both languages and image representations by a two-stage training schema consisting of teacher learning and contrastive learning. We validate our method through evaluations of a wide range of tasks. We set new state-of-the-art performances on a bunch of tasks including ImageNet-CN, Flicker30k-CN, COCO-CN and XTD. Further, we obtain very close performances with CLIP on almost all tasks, suggesting that one can simply alter the text encoder in CLIP for extended capabilities such as multilingual understanding. Our models and code are available at https://github.com/FlagAI-Open/FlagAI.

Bo-Wen Zhang, Yan Yan, Lin Li et al.

Recent advancements in Chain-of-Thoughts (CoT) and Program-of-Thoughts (PoT) methods have greatly enhanced language models' mathematical reasoning capabilities, facilitating their integration into instruction tuning datasets with LLMs. However, existing methods for large-scale dataset creation require substantial seed data and high computational costs for data synthesis, posing significant challenges for scalability. We introduce InfinityMATH, a scalable instruction tuning dataset for programmatic mathematical reasoning. The construction pipeline emphasizes decoupling numbers from mathematical problems to synthesize number-independent programs, enabling efficient and flexible scaling while minimizing dependency on specific numerical values. Fine-tuning experiments with open-source language and code models, such as Llama2 and CodeLlama, demonstrate the practical benefits of InfinityMATH. These fine-tuned models, showed significant relative improvements on both in-domain and out-of-domain benchmarks, ranging from 184.7% to 514.3% on average. Additionally, these models exhibited high robustness on the GSM8K+ and MATH+ benchmarks, which are enhanced version of test sets with simply the number variations. InfinityMATH ensures that models are more versatile and effective across a broader range of mathematical problems. The data is available at https://huggingface.co/datasets/flagopen/InfinityMATH.

Bo-Wen Zhang, Liangdong Wang, Jijie Li et al.

This paper introduces the Aquila2 series, which comprises a wide range of bilingual models with parameter sizes of 7, 34, and 70 billion. These models are trained based on an innovative framework named HeuriMentor (HM), which offers real-time insights into model convergence and enhances the training process and data management. The HM System, comprising the Adaptive Training Engine (ATE), Training State Monitor (TSM), and Data Management Unit (DMU), allows for precise monitoring of the model's training progress and enables efficient optimization of data distribution, thereby enhancing training effectiveness. Extensive evaluations show that the Aquila2 model series performs comparably well on both English and Chinese benchmarks. Specifically, Aquila2-34B demonstrates only a slight decrease in performance when quantized to Int4. Furthermore, we have made our training code (https://github.com/FlagOpen/FlagScale) and model weights (https://github.com/FlagAI-Open/Aquila2) publicly available to support ongoing research and the development of applications.

Rongao Li, Jie Fu, Bo-Wen Zhang et al.

We introduce TACO, an open-source, large-scale code generation dataset, with a focus on the optics of algorithms, designed to provide a more challenging training dataset and evaluation benchmark in the field of code generation models. TACO includes competition-level programming questions that are more challenging, to enhance or evaluate problem understanding and reasoning abilities in real-world programming scenarios. There are 25433 and 1000 coding problems in training and test set, as well as up to 1.55 million diverse solution answers. Moreover, each TACO problem includes several fine-grained labels such as task topics, algorithms, programming skills, and difficulty levels, providing a more precise reference for the training and evaluation of code generation models. The dataset and evaluation scripts are available on Hugging Face Hub (https://huggingface.co/datasets/BAAI/TACO) and Github (https://github.com/FlagOpen/TACO).

Shuhao Gu, Jialing Zhang, Siyuan Zhou et al.

Recently, Vision-Language Models (VLMs) have achieved remarkable progress in multimodal tasks, and multimodal instruction data serves as the foundation for enhancing VLM capabilities. Despite the availability of several open-source multimodal datasets, limitations in the scale and quality of open-source instruction data hinder the performance of VLMs trained on these datasets, leading to a significant gap compared to models trained on closed-source data. To address this challenge, we introduce Infinity-MM, a large-scale multimodal instruction dataset. We collected the available multimodal instruction datasets and performed unified preprocessing, resulting in a dataset with over 40 million samples that ensures diversity and accuracy. Furthermore, to enable large-scale expansion of instruction data and support the continuous acquisition of high-quality data, we propose a synthetic instruction generation method based on a tagging system and open-source VLMs. By establishing correspondences between different types of images and associated instruction types, this method can provide essential guidance during data synthesis. Leveraging this high-quality data, we have trained a 2-billion-parameter Vision-Language Model, Aquila-VL-2B, which achieves state-of-the-art (SOTA) performance among models of similar scale. The data is available at: https://huggingface.co/datasets/BAAI/Infinity-MM.

Ruichao Yang, Wei Gao, Xiaobin Zhu et al.

Multimodal misinformation poses an escalating challenge that often evades traditional detectors, which are opaque black boxes and fragile against new manipulation tactics. We present Probabilistic Concept Graph Reasoning (PCGR), an interpretable and evolvable framework that reframes multimodal misinformation detection (MMD) as structured and concept-based reasoning. PCGR follows a build-then-infer paradigm, which first constructs a graph of human-understandable concept nodes, including novel high-level concepts automatically discovered and validated by multimodal large language models (MLLMs), and then applies hierarchical attention over this concept graph to infer claim veracity. This design produces interpretable reasoning chains linking evidence to conclusions. Experiments demonstrate that PCGR achieves state-of-the-art MMD accuracy and robustness to emerging manipulation types, outperforming prior methods in both coarse detection and fine-grained manipulation recognition.

Jijie Li, Li Du, Hanyu Zhao et al.

Large Language Models (LLMs) demonstrate strong performance in real-world applications, yet existing open-source instruction datasets often concentrate on narrow domains, such as mathematics or coding, limiting generalization and widening the gap with proprietary models. To bridge this gap, we introduce Infinity-Instruct, a high-quality instruction dataset designed to enhance both foundational and chat capabilities of LLMs through a two-phase pipeline. In Phase 1, we curate 7.4M high-quality foundational instructions (InfInstruct-F-7.4M) from over 100M samples using hybrid data selection techniques. In Phase 2, we synthesize 1.5M high-quality chat instructions (InfInstruct-G-1.5M) through a two-stage process involving instruction selection, evolution, and diagnostic filtering. We empirically evaluate Infinity-Instruct by fine-tuning several open-source models, including Mistral, LLaMA, Qwen, and Yi, and observe substantial performance gains across both foundational and instruction following benchmarks, consistently surpassing official instruction-tuned counterparts. Notably, InfInstruct-LLaMA3.1-70B outperforms GPT-4-0314 by 8.6\% on instruction following tasks while achieving comparable foundational performance. These results underscore the synergy between foundational and chat training and offer new insights into holistic LLM development. Our dataset\footnote{https://huggingface.co/datasets/BAAI/Infinity-Instruct} and codes\footnote{https://gitee.com/li-touch/infinity-instruct} have been publicly released.

Bo-Wen Zhang, Liangdong Wang, Ye Yuan et al.

In recent years, with the rapid application of large language models across various fields, the scale of these models has gradually increased, and the resources required for their pre-training have grown exponentially. Training an LLM from scratch will cost a lot of computation resources while scaling up from a smaller model is a more efficient approach and has thus attracted significant attention. In this paper, we present AquilaMoE, a cutting-edge bilingual 8*16B Mixture of Experts (MoE) language model that has 8 experts with 16 billion parameters each and is developed using an innovative training methodology called EfficientScale. This approach optimizes performance while minimizing data requirements through a two-stage process. The first stage, termed Scale-Up, initializes the larger model with weights from a pre-trained smaller model, enabling substantial knowledge transfer and continuous pretraining with significantly less data. The second stage, Scale-Out, uses a pre-trained dense model to initialize the MoE experts, further enhancing knowledge transfer and performance. Extensive validation experiments on 1.8B and 7B models compared various initialization schemes, achieving models that maintain and reduce loss during continuous pretraining. Utilizing the optimal scheme, we successfully trained a 16B model and subsequently the 8*16B AquilaMoE model, demonstrating significant improvements in performance and training efficiency.

Liangdong Wang, Bo-Wen Zhang, Chengwei Wu et al.

We present CCI3.0-HQ (https://huggingface.co/datasets/BAAI/CCI3-HQ), a high-quality 500GB subset of the Chinese Corpora Internet 3.0 (CCI3.0)(https://huggingface.co/datasets/BAAI/CCI3-Data), developed using a novel two-stage hybrid filtering pipeline that significantly enhances data quality. To evaluate its effectiveness, we trained a 0.5B parameter model from scratch on 100B tokens across various datasets, achieving superior performance on 10 benchmarks in a zero-shot setting compared to CCI3.0, SkyPile, and WanjuanV1. The high-quality filtering process effectively distills the capabilities of the Qwen2-72B-instruct model into a compact 0.5B model, attaining optimal F1 scores for Chinese web data classification. We believe this open-access dataset will facilitate broader access to high-quality language models.

Xiao-Wen Yang, Jie-Jing Shao, Lan-Zhe Guo et al.

Large Language Models (LLMs) have shown promising results across various tasks, yet their reasoning capabilities remain a fundamental challenge. Developing AI systems with strong reasoning capabilities is regarded as a crucial milestone in the pursuit of Artificial General Intelligence (AGI) and has garnered considerable attention from both academia and industry. Various techniques have been explored to enhance the reasoning capabilities of LLMs, with neuro-symbolic approaches being a particularly promising way. This paper comprehensively reviews recent developments in neuro-symbolic approaches for enhancing LLM reasoning. We first present a formalization of reasoning tasks and give a brief introduction to the neurosymbolic learning paradigm. Then, we discuss neuro-symbolic methods for improving the reasoning capabilities of LLMs from three perspectives: Symbolic->LLM, LLM->Symbolic, and LLM+Symbolic. Finally, we discuss several key challenges and promising future directions. We have also released a GitHub repository including papers and resources related to this survey: https://github.com/LAMDASZ-ML/Awesome-LLM-Reasoning-with-NeSy.

Bo-Wen Zhang, Jin Ye, Peng-Yu Hua et al.

Travel planning serves as a critical task for long-horizon reasoning, exposing significant deficits in LLMs. However, existing benchmarks and evaluations primarily assess final plans in an end-to-end manner, which lacks interpretability and makes it difficult to analyze the root causes of failures. To bridge this gap, we decompose travel planning into five constituent atomic sub-capabilities, including \emph{Constraint Extraction}, \emph{Tool Use}, \emph{Plan Generation}, \emph{Error Identification}, and \emph{Error Correction}. We implement a decoupled evaluation protocol leveraging oracle intermediate contexts to rigorously isolate these components, thereby measuring the atomic performance boundary without the noise of cascading errors. Our results highlight a clear contrast in performance: while LLMs are proficient in extracting explicit constraints, they struggle to infer implicit, open-world requirements. Furthermore, they exhibit structural biases in plan generation and suffer from ineffective self-correction, characterized by excessive sensitivity and erroneous persistence. These findings offer precise directions for improving LLM reasoning and planning abilities.

Jie-Jing Shao, Bo-Wen Zhang, Xiao-Wen Yang et al.

Recent advances in LLMs, particularly in language reasoning and tool integration, have rapidly sparked the \emph{Language Agents} for real-world development. Among these, travel planning represents a prominent domain, combining complex multi-objective planning challenges with practical deployment demands. However, existing benchmarks often oversimplify real-world requirements by focusing on synthetic queries and limited constraints. We address the gap of evaluating language agents in multi-day, multi-POI travel planning scenarios with diverse and open human needs. Specifically, we introduce \emph{ChinaTravel}, the first open-ended benchmark grounded in authentic Chinese travel requirements collected from 1,154 human participants. We design a compositionally generalizable domain-specific language (DSL) for scalable evaluation, covering feasibility, constraint satisfaction, and preference comparison. Empirical studies reveal the potential of neuro-symbolic agents in travel planning, achieving a 37.0\% constraint satisfaction rate on human queries, a 10\times improvement over purely neural models. These findings highlight ChinaTravel as a pivotal milestone for advancing language agents in complex, real-world planning scenarios.

Yunan Wang, Jijie Li, Bo-Wen Zhang et al.

Direct Preference Optimization (DPO) optimizes language models to align with human preferences. Utilizing on-policy samples, generated directly by the policy model, typically results in better performance due to its distribution consistency with the model compared to off-policy samples. This paper identifies the quality of candidate preference samples as another critical factor. While the quality of on-policy data is inherently constrained by the capabilities of the policy model, off-policy data, which can be derived from diverse sources, offers greater potential for quality despite experiencing distribution shifts. However, current research mostly relies on on-policy data and neglects the value of off-policy data in terms of data quality, due to the challenge posed by distribution shift. In this paper, we propose InCo-DPO, an efficient method for synthesizing preference data by integrating on-policy and off-policy data, allowing dynamic adjustments to balance distribution shifts and data quality, thus finding an optimal trade-off. Consequently, InCo-DPO overcomes the limitations of distribution shifts in off-policy data and the quality constraints of on-policy data. We evaluated InCo-DPO with the Alpaca-Eval 2.0 and Arena-Hard benchmarks. Experimental results demonstrate that our approach not only outperforms both on-policy and off-policy data but also achieves a state-of-the-art win rate of 60.8 on Arena-Hard with the vanilla DPO using Gemma-2 model.

Bo-Wen Zhang, Yan Yan, Boxiang Yang et al.

While scaling laws optimize training configurations for large language models (LLMs) through experiments on smaller or early-stage models, they fail to predict emergent abilities due to the absence of such capabilities in these models. To address this, we propose a method that predicts emergent abilities by leveraging proxy tasks. We begin by establishing relevance metrics between the target task and candidate tasks based on performance differences across multiple models. These candidate tasks are then validated for robustness with small model ensembles, leading to the selection of the most appropriate proxy tasks. The predicted performance on the target task is then derived by integrating the evaluation results of these proxies. In a case study on tool utilization capabilities, our method demonstrated a strong correlation between predicted and actual performance, confirming its effectiveness.

Yulin Luo, Chun-Kai Fan, Menghang Dong et al.

Building robots that can perceive, reason, and act in dynamic, unstructured environments remains a core challenge. Recent embodied systems often adopt a dual-system paradigm, where System 2 handles high-level reasoning while System 1 executes low-level control. In this work, we refer to System 2 as the embodied brain, emphasizing its role as the cognitive core for reasoning and decision-making in manipulation tasks. Given this role, systematic evaluation of the embodied brain is essential. Yet existing benchmarks emphasize execution success, or when targeting high-level reasoning, suffer from incomplete dimensions and limited task realism, offering only a partial picture of cognitive capability. To bridge this gap, we introduce RoboBench, a benchmark that systematically evaluates multimodal large language models (MLLMs) as embodied brains. Motivated by the critical roles across the full manipulation pipeline, RoboBench defines five dimensions-instruction comprehension, perception reasoning, generalized planning, affordance prediction, and failure analysis-spanning 14 capabilities, 25 tasks, and 6092 QA pairs. To ensure realism, we curate datasets across diverse embodiments, attribute-rich objects, and multi-view scenes, drawing from large-scale real robotic data. For planning, RoboBench introduces an evaluation framework, MLLM-as-world-simulator. It evaluate embodied feasibility by simulating whether predicted plans can achieve critical object-state changes. Experiments on 14 MLLMs reveal fundamental limitations: difficulties with implicit instruction comprehension, spatiotemporal reasoning, cross-scenario planning, fine-grained affordance understanding, and execution failure diagnosis. RoboBench provides a comprehensive scaffold to quantify high-level cognition, and guide the development of next-generation embodied MLLMs. The project page is in https://robo-bench.github.io.

Yan Yan, Junyuan Liu, Bo-Wen Zhang

Motivation: Despite recent advancements in semantic representation driven by pre-trained and large-scale language models, addressing long tail challenges in multi-label text classification remains a significant issue. Long tail challenges have persistently posed difficulties in accurately classifying less frequent labels. Current approaches often focus on improving text semantics while neglecting the crucial role of label relationships. Results: This paper introduces LabelCoRank, a novel approach inspired by ranking principles. LabelCoRank leverages label co-occurrence relationships to refine initial label classifications through a dual-stage reranking process. The first stage uses initial classification results to form a preliminary ranking. In the second stage, a label co-occurrence matrix is utilized to rerank the preliminary results, enhancing the accuracy and relevance of the final classifications. By integrating the reranked label representations as additional text features, LabelCoRank effectively mitigates long tail issues in multi-labeltext classification. Experimental evaluations on popular datasets including MAG-CS, PubMed, and AAPD demonstrate the effectiveness and robustness of LabelCoRank.

Zhenduo Zhang, Bo-Wen Zhang, Guang Liu

Current text-to-image editing models often encounter challenges with smoothly manipulating multiple attributes using a single instruction. Taking inspiration from the Chain-of-Thought prompting technique utilized in language models, we present an innovative concept known as Chain-of-Instruct Editing (CoIE), which enhances the capabilities of these models through step-by-step editing using a series of instructions. In particular, in the context of face manipulation, we leverage the contextual learning abilities of a pretrained Large Language Model (LLM), such as GPT-4, to generate a sequence of instructions from the original input, utilizing a purpose-designed 1-shot template. To further improve the precision of each editing step, we conduct fine-tuning on the editing models using our self-constructed instruction-guided face editing dataset, Instruct-CelebA. And additionally, we incorporate a super-resolution module to mitigate the adverse effects of editability and quality degradation. Experimental results across various challenging cases confirm the significant boost in multi-attribute facial image manipulation using chain-of-instruct editing. This is evident in enhanced editing success rates, measured by CLIPSim and Coverage metrics, improved by 17.86% and 85.45% respectively, and heightened controllability indicated by Preserve L1 and Quality metrics, improved by 11.58% and 4.93% respectively.