Xin Liu

Xin Liu, Aoyang Zhou, Aoyang Zhou

Visual-Language Pre-training (VLP) models have achieved significant performance across various downstream tasks. However, they remain vulnerable to adversarial examples. While prior efforts focus on improving the adversarial transferability of multimodal adversarial examples through cross-modal interactions, these approaches suffer from overfitting issues, due to a lack of input diversity by relying excessively on information from adversarial examples in one modality when crafting attacks in another. To address this issue, we draw inspiration from strategies in some adversarial training methods and propose a novel attack called Local Shuffle and Sample-based Attack (LSSA). LSSA randomly shuffles one of the local image blocks, thus expanding the original image-text pairs, generating adversarial images, and sampling around them. Then, it utilizes both the original and sampled images to generate the adversarial texts. Extensive experiments on multiple models and datasets demonstrate that LSSA significantly enhances the transferability of multimodal adversarial examples across diverse VLP models and downstream tasks. Moreover, LSSA outperforms other advanced attacks on Large Vision-Language Models.

Tiantian Fan, Lingjun Liu, Yu Yue et al.

Recently, test-time scaling Large Language Models (LLMs) have demonstrated exceptional reasoning capabilities across scientific and professional tasks by generating long chains-of-thought (CoT). As a crucial component for developing these reasoning models, reinforcement learning (RL), exemplified by Proximal Policy Optimization (PPO) and its variants, allows models to learn through trial and error. However, PPO can be time-consuming due to its inherent on-policy nature, which is further exacerbated by increasing response lengths. In this work, we propose Truncated Proximal Policy Optimization (T-PPO), a novel extension to PPO that improves training efficiency by streamlining policy update and length-restricted response generation. T-PPO mitigates the issue of low hardware utilization, an inherent drawback of fully synchronized long-generation procedures, where resources often sit idle during the waiting periods for complete rollouts. Our contributions are two-folds. First, we propose Extended Generalized Advantage Estimation (EGAE) for advantage estimation derived from incomplete responses while maintaining the integrity of policy learning. Second, we devise a computationally optimized mechanism that allows for the independent optimization of the policy and value models. By selectively filtering prompt and truncated tokens, this mechanism reduces redundant computations and accelerates the training process without sacrificing convergence performance. We demonstrate the effectiveness and efficacy of T-PPO on AIME 2024 with a 32B base model. The experimental results show that T-PPO improves the training efficiency of reasoning LLMs by up to 2.5x and outperforms its existing competitors.

Xin Liu, Hao Wang, Shibei Xue et al.

In computer vision, estimating the six-degree-of-freedom pose from an RGB image is a fundamental task. However, this task becomes highly challenging in multi-object scenes. Currently, the best methods typically employ an indirect strategy, which identifies 2D and 3D correspondences, and then solves with the Perspective-n-Points method. Yet, this approach cannot be trained end-to-end. Direct methods, on the other hand, suffer from lower accuracy due to challenges such as varying object sizes and occlusions. To address these issues, we propose SEMPose, an end-to-end multi-object pose estimation network. SEMPose utilizes a well-designed texture-shape guided feature pyramid network, effectively tackling the challenge of object size variations. Additionally, it employs an iterative refinement head structure, progressively regressing rotation and translation separately to enhance estimation accuracy. During training, we alleviate the impact of occlusion by selecting positive samples from visible parts. Experimental results demonstrate that SEMPose can perform inference at 32 FPS without requiring inputs other than the RGB image. It can accurately estimate the poses of multiple objects in real time, with inference time unaffected by the number of target objects. On the LM-O and YCB-V datasets, our method outperforms other RGB-based single-model methods, achieving higher accuracy. Even when compared with multi-model methods and approaches that use additional refinement, our results remain competitive.