Xin Geng

Yu Shi, Ning Xu, Hua Yuan et al.

Partial label learning (PLL) is a typical weakly supervised learning problem in which each instance is associated with a candidate label set, and among which only one is true. However, the assumption that the ground-truth label is always among the candidate label set would be unrealistic, as the reliability of the candidate label sets in real-world applications cannot be guaranteed by annotators. Therefore, a generalized PLL named Unreliable Partial Label Learning (UPLL) is proposed, in which the true label may not be in the candidate label set. Due to the challenges posed by unreliable labeling, previous PLL methods will experience a marked decline in performance when applied to UPLL. To address the issue, we propose a two-stage framework named Unreliable Partial Label Learning with Recursive Separation (UPLLRS). In the first stage, the self-adaptive recursive separation strategy is proposed to separate the training set into a reliable subset and an unreliable subset. In the second stage, a disambiguation strategy is employed to progressively identify the ground-truth labels in the reliable subset. Simultaneously, semi-supervised learning methods are adopted to extract valuable information from the unreliable subset. Our method demonstrates state-of-the-art performance as evidenced by experimental results, particularly in situations of high unreliability. Code and supplementary materials are available at https://github.com/dhiyu/UPLLRS.

Yucheng Xie, Fu Feng, Ruixiao Shi et al.

Pre-trained models have become the preferred backbone due to the increasing complexity of model parameters. However, traditional pre-trained models often face deployment challenges due to their fixed sizes, and are prone to negative transfer when discrepancies arise between training tasks and target tasks. To address this, we propose KIND, a novel pre-training method designed to construct decomposable models. KIND integrates knowledge by incorporating Singular Value Decomposition (SVD) as a structural constraint, with each basic component represented as a combination of a column vector, singular value, and row vector from U, Σ, and V^\top matrices. These components are categorized into learngenes for encapsulating class-agnostic knowledge and tailors for capturing class-specific knowledge, with knowledge diversion facilitated by a class gate mechanism during training. Extensive experiments demonstrate that models pre-trained with KIND can be decomposed into learngenes and tailors, which can be adaptively recombined for diverse resource-constrained deployments. Moreover, for tasks with large domain shifts, transferring only learngenes with task-agnostic knowledge, when combined with randomly initialized tailors, effectively mitigates domain shifts. Code will be made available at https://github.com/Te4P0t/KIND.

Dongjia Zhao, Lei Qi, Xiao Shi et al.

Single domain generalization aims to enhance the ability of the model to generalize to unknown domains when trained on a single source domain. However, the limited diversity in the training data hampers the learning of domain-invariant features, resulting in compromised generalization performance. To address this, data perturbation (augmentation) has emerged as a crucial method to increase data diversity. Nevertheless, existing perturbation methods often focus on either image-level or feature-level perturbations independently, neglecting their synergistic effects. To overcome these limitations, we propose CPerb, a simple yet effective cross-perturbation method. Specifically, CPerb utilizes both horizontal and vertical operations. Horizontally, it applies image-level and feature-level perturbations to enhance the diversity of the training data, mitigating the issue of limited diversity in single-source domains. Vertically, it introduces multi-route perturbation to learn domain-invariant features from different perspectives of samples with the same semantic category, thereby enhancing the generalization capability of the model. Additionally, we propose MixPatch, a novel feature-level perturbation method that exploits local image style information to further diversify the training data. Extensive experiments on various benchmark datasets validate the effectiveness of our method.

Jiale Fu, Yuchu Jiang, Junkai Chen et al.

Large Language Model (LLM) collaborative decoding techniques improve output quality by combining the outputs of multiple models at each generation step, but they incur high computational costs. In this paper, we introduce Collaborative decoding via Speculation (CoS), a novel framework that accelerates collaborative decoding without compromising performance. Inspired by Speculative Decoding--where a small proposal model generates tokens sequentially, and a larger target model verifies them in parallel, our approach builds on two key insights: (1) the verification distribution can be the combined distribution of both the proposal and target models, and (2) alternating each model as the proposer and verifier can further enhance efficiency. We generalize this method to collaboration among n models and theoretically prove that CoS is never slower than standard collaborative decoding, typically achieving faster speed. Extensive experiments demonstrate CoS is 1.11x-2.23x faster than standard collaborative decoding without compromising generation quality. Our code is available at https://github.com/Kamichanw/CoS/.

Fu Feng, Yucheng Xie, Xu Yang et al.

``Creative'' remains an inherently abstract concept for both humans and diffusion models. While text-to-image (T2I) diffusion models can easily generate out-of-distribution concepts like ``a blue banana'', they struggle with generating combinatorial objects such as ``a creative mixture that resembles a lettuce and a mantis'', due to difficulties in understanding the semantic depth of ``creative''. Current methods rely heavily on synthesizing reference prompts or images to achieve a creative effect, typically requiring retraining for each unique creative output-a process that is computationally intensive and limits practical applications. To address this, we introduce CreTok, which brings meta-creativity to diffusion models by redefining ``creative'' as a new token, \texttt{<CreTok>}, thus enhancing models' semantic understanding for combinatorial creativity. CreTok achieves such redefinition by iteratively sampling diverse text pairs from our proposed CangJie dataset to form adaptive prompts and restrictive prompts, and then optimizing the similarity between their respective text embeddings. Extensive experiments demonstrate that <CreTok> enables the universal and direct generation of combinatorial creativity across diverse concepts without additional training, achieving state-of-the-art performance with improved text-image alignment and higher human preference ratings. Code will be made available at https://github.com/fu-feng/CreTok.

Huazhong Zhao, Lei Qi, Xin Geng

Recent advancements in pre-trained vision-language models like CLIP have shown promise in person re-identification (ReID) applications. However, their performance in generalizable person re-identification tasks remains suboptimal. The large-scale and diverse image-text pairs used in CLIP's pre-training may lead to a lack or insufficiency of certain fine-grained features. In light of these challenges, we propose a hard sample mining method called DFGS (Depth-First Graph Sampler), based on depth-first search, designed to offer sufficiently challenging samples to enhance CLIP's ability to extract fine-grained features. DFGS can be applied to both the image encoder and the text encoder in CLIP. By leveraging the powerful cross-modal learning capabilities of CLIP, we aim to apply our DFGS method to extract challenging samples and form mini-batches with high discriminative difficulty, providing the image model with more efficient and challenging samples that are difficult to distinguish, thereby enhancing the model's ability to differentiate between individuals. Our results demonstrate significant improvements over other methods, confirming the effectiveness of DFGS in providing challenging samples that enhance CLIP's performance in generalizable person re-identification.

Xiaofeng Tan, Hongsong Wang, Xin Geng et al.

Text-to-motion generation is essential for advancing the creative industry but often presents challenges in producing consistent, realistic motions. To address this, we focus on fine-tuning text-to-motion models to consistently favor high-quality, human-preferred motions, a critical yet largely unexplored problem. In this work, we theoretically investigate the DPO under both online and offline settings, and reveal their respective limitation: overfitting in offline DPO, and biased sampling in online DPO. Building on our theoretical insights, we introduce Semi-online Preference Optimization (SoPo), a DPO-based method for training text-to-motion models using "semi-online" data pair, consisting of unpreferred motion from online distribution and preferred motion in offline datasets. This method leverages both online and offline DPO, allowing each to compensate for the other's limitations. Extensive experiments demonstrate that SoPo outperforms other preference alignment methods, with an MM-Dist of 3.25% (vs e.g. 0.76% of MoDiPO) on the MLD model, 2.91% (vs e.g. 0.66% of MoDiPO) on MDM model, respectively. Additionally, the MLD model fine-tuned by our SoPo surpasses the SoTA model in terms of R-precision and MM Dist. Visualization results also show the efficacy of our SoPo in preference alignment. Project page: https://xiaofeng-tan.github.io/projects/SoPo/ .

Junming Yang, Ning Xu, Biao Liu et al.

Preference optimization is crucial for aligning large language models (LLMs) with human values and intentions. A significant challenge in this process is the distribution mismatch between pre-collected offline preference data and the evolving model policy. Existing methods attempt to reduce this gap using static heuristics or decoupled online sampling strategies, but they often fail to adapt to the model's dynamic learning state. To bridge this gap, we propose Meta-Weighted Adaptive Preference Optimization (MetaAPO), a novel framework that dynamically couples data generation with model training. MetaAPO employs a lightweight meta-learner, as an "alignment gap estimator", to evaluate the potential benefits of on-policy sampling in relation to offline data. This guides targeted online generation and assigns sample-wise meta-weights to the optimization objective, dynamically balancing the quality and distribution of online and offline data. Experiments on AlpacaEval 2, Arena-Hard and MT-Bench demonstrate that MetaAPO consistently outperforms existing preference optimization approaches across various settings, while reducing 42% in online annotation costs.

Congyu Qiao, Ning Xu, Yihao Hu et al.

Instance-dependent Partial Label Learning (ID-PLL) aims to learn a multi-class predictive model given training instances annotated with candidate labels related to features, among which correct labels are hidden fixed but unknown. The previous works involve leveraging the identification capability of the training model itself to iteratively refine supervision information. However, these methods overlook a critical aspect of ID-PLL: the training model is prone to overfitting on incorrect candidate labels, thereby providing poor supervision information and creating a bottleneck in training. In this paper, we propose to leverage reduction-based pseudo-labels to alleviate the influence of incorrect candidate labels and train our predictive model to overcome this bottleneck. Specifically, reduction-based pseudo-labels are generated by performing weighted aggregation on the outputs of a multi-branch auxiliary model, with each branch trained in a label subspace that excludes certain labels. This approach ensures that each branch explicitly avoids the disturbance of the excluded labels, allowing the pseudo-labels provided for instances troubled by these excluded labels to benefit from the unaffected branches. Theoretically, we demonstrate that reduction-based pseudo-labels exhibit greater consistency with the Bayes optimal classifier compared to pseudo-labels directly generated from the predictive model.

Shiqi Qiao, Ning Xv, Biao Liu et al.

Large language models have achieved remarkable capabilities, but aligning their outputs with human values and preferences remains a significant challenge. Existing alignment methods primarily focus on positive examples while overlooking the importance of negative responses in guiding models away from undesirable behaviors. For instance, the widely-used alignment datasets reveals a scarcity of explicit negative examples that contradict human values, hindering its ability to discourage harmful or biased outputs during training. To address this limitation, we propose NEAT, i.e., NEgative-prompt-driven AlignmenT, to introduce negative prompts to generate undesirable responses alongside positive examples during the optimization process. NEAT explicitly penalizes the model for producing harmful outputs, guiding it not only toward desirable behaviors but also steering it away from generating undesirable, biased responses. This dual feedback mechanism enables better alignment with human preferences, crucial in contexts where avoiding harm is paramount. Starting from a pre-trained language model, NEAT performs online alignment by incorporating a ranking loss derived from an expanded preference dataset containing both positive and negative examples. Extensive experiments validate NEAT's effectiveness in significantly enhancing language models' alignment with human values and preferences.