Tianhao Liang

Haodong Duan, Xinyu Fang, Junming Yang et al. · pku

We present VLMEvalKit: an open-source toolkit for evaluating large multi-modality models based on PyTorch. The toolkit aims to provide a user-friendly and comprehensive framework for researchers and developers to evaluate existing multi-modality models and publish reproducible evaluation results. In VLMEvalKit, we implement over 200+ different large multi-modality models, including both proprietary APIs and open-source models, as well as more than 80 different multi-modal benchmarks. By implementing a single interface, new models can be easily added to the toolkit, while the toolkit automatically handles the remaining workloads, including data preparation, distributed inference, prediction post-processing, and metric calculation. Although the toolkit is currently mainly used for evaluating large vision-language models, its design is compatible with future updates that incorporate additional modalities, such as audio and video. Based on the evaluation results obtained with the toolkit, we host OpenVLM Leaderboard, a comprehensive leaderboard to track the progress of multi-modality learning research. The toolkit is released on https://github.com/open-compass/VLMEvalKit and is actively maintained.

Xiangyu Zhao, Peiyuan Zhang, Junming Lin et al.

Reinforcement learning (RL) has emerged as a promising paradigm for enhancing image editing and text-to-image (T2I) generation. However, current reward models, which act as critics during RL, often suffer from hallucinations and assign noisy scores, inherently misguiding the optimization process. In this paper, we present FIRM (Faithful Image Reward Modeling), a comprehensive framework that develops robust reward models to provide accurate and reliable guidance for faithful image generation and editing. First, we design tailored data curation pipelines to construct high-quality scoring datasets. Specifically, we evaluate editing using both execution and consistency, while generation is primarily assessed via instruction following. Using these pipelines, we collect the FIRM-Edit-370K and FIRM-Gen-293K datasets, and train specialized reward models (FIRM-Edit-8B and FIRM-Gen-8B) that accurately reflect these criteria. Second, we introduce FIRM-Bench, a comprehensive benchmark specifically designed for editing and generation critics. Evaluations demonstrate that our models achieve superior alignment with human judgment compared to existing metrics. Furthermore, to seamlessly integrate these critics into the RL pipeline, we formulate a novel "Base-and-Bonus" reward strategy that balances competing objectives: Consistency-Modulated Execution (CME) for editing and Quality-Modulated Alignment (QMA) for generation. Empowered by this framework, our resulting models FIRM-Qwen-Edit and FIRM-SD3.5 achieve substantial performance breakthroughs. Comprehensive experiments demonstrate that FIRM mitigates hallucinations, establishing a new standard for fidelity and instruction adherence over existing general models. All of our datasets, models, and code have been publicly available at https://firm-reward.github.io.

Yuqi Ping, Huahao Ding, Tianhao Liang et al.

Natural language (NL) navigation for low-altitude unmanned aerial vehicles (UAVs) offers an intelligent and convenient solution for low-altitude aerial services by enabling an intuitive interface for non-expert operators. However, deploying this capability in urban environments necessitates the precise grounding of underspecified instructions into safety-critical, dynamically feasible motion plans subject to spatiotemporal constraints. To address this challenge, we propose a unified framework that translates NL instructions into Signal Temporal Logic (STL) specifications and subsequently synthesizes trajectories via mixed-integer linear programming (MILP). Specifically, to generate executable STL formulas from free-form NL, we develop a reasoning-enhanced large language model (LLM) leveraging chain-of-thought (CoT) supervision and group-relative policy optimization (GRPO), which ensures high syntactic validity and semantic consistency. Furthermore, to resolve infeasibilities induced by stringent logical or spatial requirements, we introduce a specification repair mechanism. This module combines MILP-based diagnosis with LLM-guided semantic reasoning to selectively relax task constraints while strictly enforcing safety guarantees. Extensive simulations and real-world flight experiments demonstrate that the proposed closed-loop framework significantly improves NL-to-STL translation robustness, enabling safe, interpretable, and adaptable UAV navigation in complex scenarios.

Longyu Zhou, Supeng Leng, Tianhao Liang et al.

The development of Artificial Intelligence (AI) has enabled agentic robots an appealing paradigm for various applications, such as research and rescue in complex environment. In this context, the next wireless communication technology facilitates robot cooperation for efficient environment sensing and exploration. However, traditional AI solutions cannot always provide reasonable resource utilization decisions, which makes it challenging to achieve both accurate and low-latency research and rescue. To address this issue, we propose a, LSAI, a large small AI model codesign framework to achieve highly accurate and real-time robot cooperation with deep interaction between large AI model and small AI model. We first propose an attention-based model aggregation for LAI construction. It can assist agentic robots in accurately sensing physical environments. Next, we design an adaptive model splitting and update algorithm to enable the robots to perform accurate path planning for high-efficiency environment sensing with low energy consumption. Finally, we demonstrate the effectiveness of our proposed LSAI framework. The simulation results indicate that our solution achieves sensing accuracy of up to 20.4% while reducing sensing cooperation latency by an average of 17.9% compared to traditional AI solutions.

Jiacheng Chen, Tianhao Liang, Sherman Siu et al. · amazon-science

We present MEGA-Bench, an evaluation suite that scales multimodal evaluation to over 500 real-world tasks, to address the highly heterogeneous daily use cases of end users. Our objective is to optimize for a set of high-quality data samples that cover a highly diverse and rich set of multimodal tasks, while enabling cost-effective and accurate model evaluation. In particular, we collected 505 realistic tasks encompassing over 8,000 samples from 16 expert annotators to extensively cover the multimodal task space. Instead of unifying these problems into standard multi-choice questions (like MMMU, MMBench, and MMT-Bench), we embrace a wide range of output formats like numbers, phrases, code, \LaTeX, coordinates, JSON, free-form, etc. To accommodate these formats, we developed over 40 metrics to evaluate these tasks. Unlike existing benchmarks, MEGA-Bench offers a fine-grained capability report across multiple dimensions (e.g., application, input type, output format, skill), allowing users to interact with and visualize model capabilities in depth. We evaluate a wide variety of frontier vision-language models on MEGA-Bench to understand their capabilities across these dimensions.

Tianhao Liang, Nanchi Su, Yuqi Ping et al.

The emerging low-altitude economy has catalyzed the large-scale deployment of unmanned aerial vehicles (UAVs), driving a paradigm shift in environment monitoring, logistics, and emergency response. However, operating within these environments presents notable challenges as pervasive coverage holes, unpredictable interference, and spectrum scarcity. To this end, this article present a communication and control co-design framework to enable a resilient architecture for cellular-connected UAVs. Specifically, we first characterize typical service applications and their stringent performance requirements, followed by a comprehensive analysis of the unique challenges. To bridge the gap between volatile wireless links and rigid flight stability, a three layered architecture is proposed, integrating pre-flight strategic planning, in-flight adaptive action, and system-level resource orchestration. Furthermore, we detail the key enabling technologies for communication and control co-design. Preliminary case studies are proposed to validate that the co-design framework significantly improve the resilience of cellular-connected UAV systems, providing a robust foundation for the evolution of intelligent low-altitude networks.

Guangyu Lei, Tianhao Liang, Yuqi Ping et al.

The rapid development of the low-altitude economy emphasizes the critical need for effective perception and intent recognition of non-cooperative unmanned aerial vehicles (UAVs). The advanced generative reasoning capabilities of multimodal large language models (MLLMs) present a promising approach in such tasks. In this paper, we focus on the combination of UAV intent recognition and the MLLMs. Specifically, we first present an MLLM-enabled UAV intent recognition architecture, where the multimodal perception system is utilized to obtain real-time payload and motion information of UAVs, generating structured input information, and MLLM outputs intent recognition results by incorporating environmental information, prior knowledge, and tactical preferences. Subsequently, we review the related work and demonstrate their progress within the proposed architecture. Then, a use case for low-altitude confrontation is conducted to demonstrate the feasibility of our architecture and offer valuable insights for practical system design. Finally, the future challenges are discussed, followed by corresponding strategic recommendations for further applications.

Xiangyu Zhao, Junming Lin, Tianhao Liang et al. · pku

While current Multimodal Large Language Models (MLLMs) have demonstrated proficiency in reasoning tasks such as mathematics and logic, their capacity for long-chain reflective reasoning, a prerequisite for solving complex real-world problems, remains largely underexplored. In this work, we first conduct an extensive empirical investigation to evaluate this capability. Leveraging a carefully designed data synthesis engine, we construct MM-HELIX, a multimodal benchmark consisting 1,260 samples of 42 challenging synthetic tasks that require iterative thinking and backtracking. Empirical results on this benchmark reveal that existing MLLMs exhibit significant performance deficits in long-chain reflective reasoning. To address this limitation, we generate post-training data and further explore learning paradigms for exploiting such data. We first develop the Step-Elicited Response Generation pipeline to create MM-HELIX-100K, a large-scale dataset of 100k high-quality, reflective reasoning traces for instruction-tuning stage. Given that standard Reinforcement Learning fails on complex tasks due to sparse reward signals and catastrophic forgetting after Supervised Fine-Tuning, we propose Adaptive Hybrid Policy Optimization (AHPO), a novel training strategy that dynamically unifies offline supervision and online optimization into a single stage. This strategy enables the model to learn from expert data when rewards are sparse and conduct independent exploration once proficient. When applied to the Qwen2.5-VL-7B baseline, our method achieves a +18.6\% accuracy improvement on MM-HELIX benchmark and demonstrates strong generalization with a +5.7\% average performance gain on general mathematic and logic tasks. Our work demonstrate that reflective reasoning in MLLMs can be effectively learned and generalized, paving the way for developing more capable MLLMs.

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.