Published in ICML 2026, CCF A, 2026
Adaptive test-time compute for Vision-Language-Action models, using a relative action critic model to allocate more deliberation to difficult embodied decision points.
Recommended citation: Li W, et al. VLA-ATTC: Adaptive Test-Time Compute for VLA Models with Relative Action Critic Model. International Conference on Machine Learning (ICML), 2026. CCF A. First author.
Published in ICML 2026, CCF A, 2026
A metacognitive Vision-Language-Action model with active status monitoring, dynamic reasoning, and error recovery for robust embodied task execution.
Recommended citation: Li W, et al. Sentinel-VLA: A Metacognitive VLA Model with Active Status Monitoring for Dynamic Reasoning and Error Recovery. International Conference on Machine Learning (ICML), 2026. CCF A. First author.
Published in ACM MM 2025, Oral, CCF A, 2025
A self-evolving distillation method for mitigating hallucinations in large vision-language models by identifying and isolating hallucination-related knowledge components.
Recommended citation: Li W, et al. Identify, Isolate, and Purge: Mitigating Hallucinations in LVLMs via Self-Evolving Distillation. ACM International Conference on Multimedia (ACM MM), 2025. Oral, CCF A. First author.
Published in ICME 2025, 2025
Diffusion models have demonstrated remarkable efficacy in various generative tasks with the predictive prowess of denoising model. Currently, these models employ a uniform denoising approach across all timesteps. However, the inherent variations in noisy latents at each timestep lead to conflicts during training, constraining the potential of diffusion models. To address this challenge, we propose a novel two-stage training strategy termed Step-Adaptive Training. In the initial stage, a base denoising model is trained to encompass all timesteps. Subsequently, we partition the timesteps into distinct groups, fine-tuning the model within each group to achieve specialized denoising capabilities. Recognizing that the difficulties of predicting noise at different timesteps vary, we introduce a diverse model size requirement. We dynamically adjust the model size for each timestep by estimating task difficulty based on its signal-to-noise ratio before fine-tuning. This adjustment is facilitated by a proxy-based structural importance assessment mechanism, enabling precise and efficient pruning of the base denoising model. Our experiments validate the effectiveness of the proposed training strategy, demonstrating an improvement in the FID score on CIFAR10 by over 0.3 while utilizing only 80% of the computational resources. This innovative approach not only enhances model performance but also significantly reduces computational costs, opening new avenues for the development and application of diffusion models.
Recommended citation: Li W, Su X, You S, et al. Adaptive Training Meets Progressive Scaling: Elevating Efficiency in Diffusion Models. IEEE International Conference on Multimedia and Expo (ICME), 2025. https://arxiv.org/html/2312.13307v2
Published in ICDM 2023, Oral, Core A*, 2023
Diffusion models have recently exhibited remarkable performance on synthetic data. After a diffusion path is selected, a base model, such as UNet, operates as a denoising autoencoder, primarily predicting noises that need to be eliminated step by step. Consequently, it is crucial to employ a model that aligns with the expected budgets to facilitate superior synthetic performance. In this paper, we meticulously analyze the diffusion model and engineer a base model search approach, denoted “DiffNAS”. Specifically, we leverage GPT-4 as a supernet to expedite the search, supplemented with a search memory to enhance the results. Moreover, we employ RFID as a proxy to promptly rank the experimental outcomes produced by GPT-4. We also adopt a rapid-convergence training strategy to boost search efficiency. Rigorous experimentation corroborates that our algorithm can augment the search efficiency by 2× under GPT-based scenarios, while also attaining a performance of 2.82 with 0.37 improvement in FID on CIFAR10 relative to the benchmark IDDPM algorithm.
Recommended citation: Li W, Su X, You S, et al. DiffNAS: Bootstrapping Diffusion Models by Prompting for Better Architectures. IEEE International Conference on Data Mining (ICDM), 2023. Oral, Core A*. https://ieeexplore.ieee.org/abstract/document/10415684