Unified Multimodal Chain-of-Thought Reward Model through Reinforcement Fine-Tuning

Recent advances in multimodal Reward Models (RMs) have shown significant promise in delivering reward signals to align vision models with human preferences. However, current RMs are generally restricted to providing direct responses or engaging in shallow reasoning processes with limited depth, often leading to inaccurate reward signals. We posit that incorporating explicit long chains of thought (CoT) into the reward reasoning process can significantly strengthen their reliability and robustness. Furthermore, we believe that once RMs internalize CoT reasoning, their direct response accuracy can also be improved through implicit reasoning capabilities. To this end, this paper proposes UnifiedReward-Think, the first unified multimodal CoT-based reward model, capable of multi-dimensional, step-by-step long-chain reasoning for both visual understanding and generation reward tasks. Specifically, we adopt an exploration-driven reinforcement fine-tuning approach to elicit and incentivize the model's latent complex reasoning ability: (1) We first use a small amount of image generation preference data to distill the reasoning process of GPT-4o, which is then used for the model's cold start to learn the format and structure of CoT reasoning. (2) Subsequently, by leveraging the model's prior knowledge and generalization capabilities, we prepare large-scale unified multimodal preference data to elicit the model's reasoning process across various vision tasks. During this phase, correct reasoning outputs are retained for rejection sampling to refine the model (3) while incorrect predicted samples are finally used for Group Relative Policy Optimization (GRPO) based reinforcement fine-tuning, enabling the model to explore diverse reasoning paths and optimize for correct and robust solutions. Extensive experiments across various vision reward tasks demonstrate the superiority of our model.

Efficient Fine-Tuning of Quantized Models via Adaptive Rank and Bitwidth

QLoRA effectively combines low-bit quantization and LoRA to achieve memory-friendly fine-tuning for large language models (LLM). Recently, methods based on SVD for continuous update iterations to initialize LoRA matrices to accommodate quantization errors have generally failed to consistently improve performance. Dynamic mixed precision is a natural idea for continuously improving the fine-tuning performance of quantized models, but previous methods often optimize low-rank subspaces or quantization components separately, without considering their synergy. To address this, we propose \textbf{QR-Adaptor}, a unified, gradient-free strategy that uses partial calibration data to jointly search the quantization components and the rank of low-rank spaces for each layer, thereby continuously improving model performance. QR-Adaptor does not minimize quantization error but treats precision and rank allocation as a discrete optimization problem guided by actual downstream performance and memory usage. Compared to state-of-the-art (SOTA) quantized LoRA fine-tuning methods, our approach achieves a 4.89\% accuracy improvement on GSM8K, and in some cases even outperforms the 16-bit fine-tuned model while maintaining the memory footprint of the 4-bit setting.

Large Language Model Compression with Global Rank and Sparsity Optimization

Low-rank and sparse composite approximation is a natural idea to compress Large Language Models (LLMs). However, such an idea faces two primary challenges that adversely affect the performance of existing methods. The first challenge relates to the interaction and cooperation between low-rank and sparse matrices, while the second involves determining weight allocation across different layers, as redundancy varies considerably among them. To address these challenges, we propose a novel two-stage LLM compression method with the capability of global rank and sparsity optimization. It is noteworthy that the overall optimization space is vast, making comprehensive optimization computationally prohibitive. Therefore, to reduce the optimization space, our first stage utilizes robust principal component analysis to decompose the weight matrices of LLMs into low-rank and sparse components, which span the low dimensional and sparse spaces containing the resultant low-rank and sparse matrices, respectively. In the second stage, we propose a probabilistic global optimization technique to jointly identify the low-rank and sparse structures within the above two spaces. The appealing feature of our approach is its ability to automatically detect the redundancy across different layers and to manage the interaction between the sparse and low-rank components. Extensive experimental results indicate that our method significantly surpasses state-of-the-art techniques for sparsification and composite approximation.

Unified Reward Model for Multimodal Understanding and Generation

Recent advances in human preference alignment have significantly enhanced multimodal generation and understanding. A key approach is training reward models to guide preference optimization. However, existing models are often task-specific, limiting their adaptability across diverse visual applications. We also argue that jointly learning to assess multiple tasks may foster a synergistic effect, where improved image understanding enhances image generation assessment, and refined image evaluation benefits video assessment through better frame analysis. To this end, this paper proposes UnifiedReward, the first unified reward model for multimodal understanding and generation assessment, enabling both pairwise ranking and pointwise scoring, which can be employed for vision model preference alignment. Specifically, (1) we first develop UnifiedReward on our constructed large-scale human preference dataset, including both image and video generation/understanding tasks. (2) Then, it is utilized to automatically construct high-quality preference pair data based on the vision models, fine-gradually filtering their outputs through pair ranking and point sifting. (3) Finally, these data are used for their preference alignment through Direct Preference Optimization (DPO). Experimental results demonstrate that joint learning to assess diverse visual tasks can lead to substantial mutual benefits and we apply our pipeline to both image and video understanding/generation tasks, significantly improving the performance in each domain.

HammerBench: Fine-Grained Function-Calling Evaluation in Real Mobile Device Scenarios

Evaluating the capabilities of large language models (LLMs) in human-LLM interactions remains challenging due to the inherent complexity and openness of dialogue processes. This paper introduces HammerBench, a novel benchmarking framework designed to assess the function-calling ability of LLMs more effectively in such interactions. We model a wide range of real-world user scenarios on mobile devices, encompassing imperfect instructions, diverse question-answer trajectories, intent/argument shifts, and the use of external individual information through pronouns. To construct the corresponding datasets, we propose a comprehensive pipeline that involves LLM-generated data and multiple rounds of human validation, ensuring high data quality. Additionally, we decompose the conversations into function-calling snapshots, enabling a fine-grained evaluation of each turn. We evaluate several popular LLMs using HammerBench and highlight different performance aspects. Our empirical findings reveal that errors in parameter naming constitute the primary factor behind conversation failures across different data types.

Optimized Gradient Clipping for Noisy Label Learning

Previous research has shown that constraining the gradient of loss function with respect to model-predicted probabilities can enhance the model robustness against noisy labels. These methods typically specify a fixed optimal threshold for gradient clipping through validation data to obtain the desired robustness against noise. However, this common practice overlooks the dynamic distribution of gradients from both clean and noisy-labeled samples at different stages of training, significantly limiting the model capability to adapt to the variable nature of gradients throughout the training process. To address this issue, we propose a simple yet effective approach called Optimized Gradient Clipping (OGC), which dynamically adjusts the clipping threshold based on the ratio of noise gradients to clean gradients after clipping, estimated by modeling the distributions of clean and noisy samples. This approach allows us to modify the clipping threshold at each training step, effectively controlling the influence of noise gradients. Additionally, we provide statistical analysis to certify the noise-tolerance ability of OGC. Our extensive experiments across various types of label noise, including symmetric, asymmetric, instance-dependent, and real-world noise, demonstrate the effectiveness of our approach. The code and a technical appendix for better digital viewing are included as supplementary materials and scheduled to be open-sourced upon publication.

LiFT: Leveraging Human Feedback for Text-to-Video Model Alignment

Recent advancements in text-to-video (T2V) generative models have shown impressive capabilities. However, these models are still inadequate in aligning synthesized videos with human preferences (e.g., accurately reflecting text descriptions), which is particularly difficult to address, as human preferences are inherently subjective and challenging to formalize as objective functions. Therefore, this paper proposes LiFT, a novel fine-tuning method leveraging human feedback for T2V model alignment. Specifically, we first construct a Human Rating Annotation dataset, LiFT-HRA, consisting of approximately 10k human annotations, each including a score and its corresponding rationale. Based on this, we train a reward model LiFT-Critic to learn reward function effectively, which serves as a proxy for human judgment, measuring the alignment between given videos and human expectations. Lastly, we leverage the learned reward function to align the T2V model by maximizing the reward-weighted likelihood. As a case study, we apply our pipeline to CogVideoX-2B, showing that the fine-tuned model outperforms the CogVideoX-5B across all 16 metrics, highlighting the potential of human feedback in improving the alignment and quality of synthesized videos.

HMIL: Hierarchical Multi-Instance Learning for Fine-Grained Whole Slide Image Classification

Fine-grained classification of whole slide images (WSIs) is essential in precision oncology, enabling precise cancer diagnosis and personalized treatment strategies. The core of this task involves distinguishing subtle morphological variations within the same broad category of gigapixel-resolution images, which presents a significant challenge. While the multi-instance learning (MIL) paradigm alleviates the computational burden of WSIs, existing MIL methods often overlook hierarchical label correlations, treating fine-grained classification as a flat multi-class classification task. To overcome these limitations, we introduce a novel hierarchical multi-instance learning (HMIL) framework. By facilitating on the hierarchical alignment of inherent relationships between different hierarchy of labels at instance and bag level, our approach provides a more structured and informative learning process. Specifically, HMIL incorporates a class-wise attention mechanism that aligns hierarchical information at both the instance and bag levels. Furthermore, we introduce supervised contrastive learning to enhance the discriminative capability for fine-grained classification and a curriculum-based dynamic weighting module to adaptively balance the hierarchical feature during training. Extensive experiments on our large-scale cytology cervical cancer (CCC) dataset and two public histology datasets, BRACS and PANDA, demonstrate the state-of-the-art class-wise and overall performance of our HMIL framework. Our source code is available at https://github.com/ChengJin-git/HMIL.

GameGen-X: Interactive Open-world Game Video Generation

We introduce GameGen-X, the first diffusion transformer model specifically designed for both generating and interactively controlling open-world game videos. This model facilitates high-quality, open-domain generation by simulating an extensive array of game engine features, such as innovative characters, dynamic environments, complex actions, and diverse events. Additionally, it provides interactive controllability, predicting and altering future content based on the current clip, thus allowing for gameplay simulation. To realize this vision, we first collected and built an Open-World Video Game Dataset from scratch. It is the first and largest dataset for open-world game video generation and control, which comprises over a million diverse gameplay video clips sampling from over 150 games with informative captions from GPT-4o. GameGen-X undergoes a two-stage training process, consisting of foundation model pre-training and instruction tuning. Firstly, the model was pre-trained via text-to-video generation and video continuation, endowing it with the capability for long-sequence, high-quality open-domain game video generation. Further, to achieve interactive controllability, we designed InstructNet to incorporate game-related multi-modal control signal experts. This allows the model to adjust latent representations based on user inputs, unifying character interaction and scene content control for the first time in video generation. During instruction tuning, only the InstructNet is updated while the pre-trained foundation model is frozen, enabling the integration of interactive controllability without loss of diversity and quality of generated video content.

MagicFace: Training-free Universal-Style Human Image Customized Synthesis

Existing human image personalized generation methods often require tedious training: either fine-tuning with a few images or retraining on large-scale datasets. In such cases, these methods are prone to overfitting and encounter difficulties when personalizing individuals of diverse styles. Moreover, these training-based approaches also struggle with multi-concept human image customizing. To this end, we propose MagicFace, the first method for universal-style human image personalized synthesis that enables single/multi-concept customization for humans of any style in a training-free manner. MagicFace introduces a coarse-to-fine generation pipeline, involving two sequential stages: semantic scene construction and concept feature injection. This is achieved by our Reference-aware Self-Attention (RSA) and Region-grouped Blend Attention (RBA) mechanisms. Specifically, in the first stage, RSA enables the latent image to query features from reference concepts simultaneously, extracting the coarse-grained overall semantic understanding to facilitate the initial semantic layout establishment. In the second stage, we employ an attention-based semantic segmentation method to pinpoint the generated regions of all concepts in the latent image at each step. Following this, RBA divides the pixels of the latent image into semantic groups, with each group querying fine-grained features from its reference concept, which ensures precise attribute alignment and feature injection. Throughout the two-stage process, a weight mask strategy is employed to ensure the model focuses more on the reference concepts. Extensive experiments demonstrate our superiority in both human-centric subject-to-image synthesis and multi-concept human image customization. Our approach also can be applied to texture transformation, further enhancing its versatility and applicability.