Learning to Skip for Language Modeling

Overparameterized large-scale language models have impressive generalization performance of in-context few-shot learning. However, most language models allocate the same amount of parameters or computation to each token, disregarding the complexity or importance of the input data. We argue that in language model pretraining, a variable amount of computation should be assigned to different tokens, and this can be efficiently achieved via a simple routing mechanism. Different from conventional early stopping techniques where tokens can early exit at only early layers, we propose a more general method that dynamically skips the execution of a layer (or module) for any input token with a binary router. In our extensive evaluation across 24 NLP tasks, we demonstrate that the proposed method can significantly improve the 1-shot performance compared to other competitive baselines only at mild extra cost for inference.

Boost Adversarial Transferability by Uniform Scale and Mix Mask Method

Adversarial examples generated from surrogate models often possess the ability to deceive other black-box models, a property known as transferability. Recent research has focused on enhancing adversarial transferability, with input transformation being one of the most effective approaches. However, existing input transformation methods suffer from two issues. Firstly, certain methods, such as the Scale-Invariant Method, employ exponentially decreasing scale invariant parameters that decrease the adaptability in generating effective adversarial examples across multiple scales. Secondly, most mixup methods only linearly combine candidate images with the source image, leading to reduced features blending effectiveness. To address these challenges, we propose a framework called Uniform Scale and Mix Mask Method (US-MM) for adversarial example generation. The Uniform Scale approach explores the upper and lower boundaries of perturbation with a linear factor, minimizing the negative impact of scale copies. The Mix Mask method introduces masks into the mixing process in a nonlinear manner, significantly improving the effectiveness of mixing strategies. Ablation experiments are conducted to validate the effectiveness of each component in US-MM and explore the effect of hyper-parameters. Empirical evaluations on standard ImageNet datasets demonstrate that US-MM achieves an average of 7% better transfer attack success rate compared to state-of-the-art methods.

Pseudo Label-Guided Data Fusion and Output Consistency for Semi-Supervised Medical Image Segmentation

Supervised learning algorithms based on Convolutional Neural Networks have become the benchmark for medical image segmentation tasks, but their effectiveness heavily relies on a large amount of labeled data. However, annotating medical image datasets is a laborious and time-consuming process. Inspired by semi-supervised algorithms that use both labeled and unlabeled data for training, we propose the PLGDF framework, which builds upon the mean teacher network for segmenting medical images with less annotation. We propose a novel pseudo-label utilization scheme, which combines labeled and unlabeled data to augment the dataset effectively. Additionally, we enforce the consistency between different scales in the decoder module of the segmentation network and propose a loss function suitable for evaluating the consistency. Moreover, we incorporate a sharpening operation on the predicted results, further enhancing the accuracy of the segmentation. Extensive experiments on three publicly available datasets demonstrate that the PLGDF framework can largely improve performance by incorporating the unlabeled data. Meanwhile, our framework yields superior performance compared to six state-of-the-art semi-supervised learning methods. The codes of this study are available at https://github.com/ortonwang/PLGDF.

Complementary Advantages of ChatGPTs and Human Readers in Reasoning: Evidence from English Text Reading Comprehension

ChatGPT has shown its great power in text processing, including its reasoning ability from text reading. However, there has not been any direct comparison between human readers and ChatGPT in reasoning ability related to text reading. This study was undertaken to investigate how ChatGPTs (i.e., ChatGPT and ChatGPT Plus) and Chinese senior school students as ESL learners exhibited their reasoning ability from English narrative texts. Additionally, we compared the two ChatGPTs in the reasoning performances when commands were updated elaborately. The whole study was composed of three reasoning tests: Test 1 for commonsense inference, Test 2 for emotional inference, and Test 3 for causal inference. The results showed that in Test 1, the students outdid the two ChatGPT versions in local-culture-related inferences but performed worse than the chatbots in daily-life inferences. In Test 2, ChatGPT Plus excelled whereas ChatGPT lagged behind in accuracy. In association with both accuracy and frequency of correct responses, the students were inferior to the two chatbots. Compared with ChatGPTs' better performance in positive emotions, the students showed their superiority in inferring negative emotions. In Test 3, the students demonstrated better logical analysis, outdoing both chatbots. In updating command condition, ChatGPT Plus displayed good causal reasoning ability while ChatGPT kept unchanged. Our study reveals that human readers and ChatGPTs have their respective advantages and disadvantages in drawing inferences from text reading comprehension, unlocking a complementary relationship in text-based reasoning.

Controlled Decoding from Language Models

We propose controlled decoding (CD), a novel off-policy reinforcement learning method to control the autoregressive generation from language models towards high reward outcomes. CD solves an off-policy reinforcement learning problem through a value function for the reward, which we call a prefix scorer. The prefix scorer is used at inference time to steer the generation towards higher reward outcomes. We show that the prefix scorer may be trained on (possibly) off-policy data to predict the expected reward when decoding is continued from a partially decoded response. We empirically demonstrate that CD is effective as a control mechanism on Reddit conversations corpus. We also show that the modularity of the design of CD makes it possible to control for multiple rewards, effectively solving a multi-objective reinforcement learning problem with no additional complexity. Finally, we show that CD can be applied in a novel blockwise fashion at inference-time, again without the need for any training-time changes, essentially bridging the gap between the popular best-of-$K$ strategy and token-level reinforcement learning. This makes CD a promising approach for alignment of language models.

Fractal Landscapes in Policy Optimization

Policy gradient lies at the core of deep reinforcement learning (RL) in continuous domains. Despite much success, it is often observed in practice that RL training with policy gradient can fail for many reasons, even on standard control problems with known solutions. We propose a framework for understanding one inherent limitation of the policy gradient approach: the optimization landscape in the policy space can be extremely non-smooth or fractal for certain classes of MDPs, such that there does not exist gradient to be estimated in the first place. We draw on techniques from chaos theory and non-smooth analysis, and analyze the maximal Lyapunov exponents and H\"older exponents of the policy optimization objectives. Moreover, we develop a practical method that can estimate the local smoothness of objective function from samples to identify when the training process has encountered fractal landscapes. We show experiments to illustrate how some failure cases of policy optimization can be explained by such fractal landscapes.

VCL Challenges 2023 at ICCV 2023 Technical Report: Bi-level Adaptation Method for Test-time Adaptive Object Detection

This report outlines our team's participation in VCL Challenges B Continual Test_time Adaptation, focusing on the technical details of our approach. Our primary focus is Testtime Adaptation using bi_level adaptations, encompassing image_level and detector_level adaptations. At the image level, we employ adjustable parameterbased image filters, while at the detector level, we leverage adjustable parameterbased mean teacher modules. Ultimately, through the utilization of these bi_level adaptations, we have achieved a remarkable 38.3% mAP on the target domain of the test set within VCL Challenges B. It is worth noting that the minimal drop in mAP, is mearly 4.2%, and the overall performance is 32.5% mAP.

SCB-Dataset3: A Benchmark for Detecting Student Classroom Behavior

The use of deep learning methods to automatically detect students' classroom behavior is a promising approach for analyzing their class performance and improving teaching effectiveness. However, the lack of publicly available datasets on student behavior poses a challenge for researchers in this field. To address this issue, we propose the Student Classroom Behavior dataset (SCB-dataset3), which represents real-life scenarios. Our dataset comprises 5686 images with 45578 labels, focusing on six behaviors: hand-raising, reading, writing, using a phone, bowing the head, and leaning over the table. We evaluated the dataset using the YOLOv5, YOLOv7, and YOLOv8 algorithms, achieving a mean average precision (map) of up to 80.3$\%$. We believe that our dataset can serve as a robust foundation for future research in student behavior detection and contribute to advancements in this field. Our SCB-dataset3 is available for download at: https://github.com/Whiffe/SCB-dataset

Controllable Residual Speaker Representation for Voice Conversion

Recently, there have been significant advancements in voice conversion, resulting in high-quality performance. However, there are still two critical challenges in this field. Firstly, current voice conversion methods have limited robustness when encountering unseen speakers. Secondly, they also have limited ability to control timbre representation. To address these challenges, this paper presents a novel approach leverages tokens of multi-layer residual approximations to enhance robustness when dealing with unseen speakers, called the residual speaker module. The introduction of multi-layer approximations facilitates the separation of information from the timbre, enabling effective control over timbre in voice conversion. The proposed method outperforms baselines in both subjective and objective evaluations, demonstrating superior performance and increased robustness. Our demo page is publicly available.

Base Station Beamforming Design for Near-field XL-IRS Beam Training

Existing research on extremely large-scale intelligent reflecting surface (XL-IRS) beam training has assumed the far-field channel model for base station (BS)-IRS link. However, this approach may cause degraded beam training performance in practice due to the near-field channel model of the BS-IRS link. To address this issue, we propose two efficient schemes to optimize BS beamforming for improving the XL-IRS beam training performance. Specifically, the first scheme aims to maximize total received signal power on the XL-IRS, which generalizes the existing angle based BS beamforming design and can be resolved using the singular value decomposition (SVD) method. The second scheme aims to maximize the $\ell_1$-norm of incident signals on the XL-IRS, which is shown to achieve the maximum received power at the user. To solve the non-convex $\ell_1$-norm maximization problem, we propose an eficient algorithm by using the alternating optimization (AO) technique. Numerical results show that the proposed AO based BS beamforming design outperforms the SVD/angle based BS beamforming in terms of training accuracy and achievable received signal-to-noise ratio (SNR).