Llama SLayer 8B: Shallow Layers Hold the Key to Knowledge Injection

As a manner to augment pre-trained large language models (LLM), knowledge injection is critical to develop vertical domain large models and has been widely studied. Although most current approaches, including parameter-efficient fine-tuning (PEFT) and block expansion methods, uniformly apply knowledge across all LLM layers, it raises the question: are all layers equally crucial for knowledge injection? We begin by evaluating the importance of each layer in finding the optimal layer range for knowledge injection. Intuitively, the more important layers should play a more critical role in knowledge injection and deserve a denser injection. We observe performance dips in question-answering benchmarks after the removal or expansion of the shallow layers, and the degradation shrinks as the layer gets deeper, indicating that the shallow layers hold the key to knowledge injection. This insight leads us to propose the S strategy, a post-pretraining strategy of selectively enhancing shallow layers while pruning the less effective deep ones. Based on this strategy, we introduce Llama Slayer-8B and Llama Slayer-8B-Instruct. We experimented on the corpus of code $\&$ math and demonstrated the effectiveness of our strategy. Further experiments across different LLM, Mistral-7B, and a legal corpus confirmed the general applicability of the approach, underscoring its wide-ranging efficacy. Our code is available at: \https://github.com/txchen-USTC/Llama-Slayer

Mixture-of-Noises Enhanced Forgery-Aware Predictor for Multi-Face Manipulation Detection and Localization

With the advancement of face manipulation technology, forgery images in multi-face scenarios are gradually becoming a more complex and realistic challenge. Despite this, detection and localization methods for such multi-face manipulations remain underdeveloped. Traditional manipulation localization methods either indirectly derive detection results from localization masks, resulting in limited detection performance, or employ a naive two-branch structure to simultaneously obtain detection and localization results, which cannot effectively benefit the localization capability due to limited interaction between two tasks. This paper proposes a new framework, namely MoNFAP, specifically tailored for multi-face manipulation detection and localization. The MoNFAP primarily introduces two novel modules: the Forgery-aware Unified Predictor (FUP) Module and the Mixture-of-Noises Module (MNM). The FUP integrates detection and localization tasks using a token learning strategy and multiple forgery-aware transformers, which facilitates the use of classification information to enhance localization capability. Besides, motivated by the crucial role of noise information in forgery detection, the MNM leverages multiple noise extractors based on the concept of the mixture of experts to enhance the general RGB features, further boosting the performance of our framework. Finally, we establish a comprehensive benchmark for multi-face detection and localization and the proposed \textit{MoNFAP} achieves significant performance. The codes will be made available.

Transformer based Pluralistic Image Completion with Reduced Information Loss

Transformer based methods have achieved great success in image inpainting recently. However, we find that these solutions regard each pixel as a token, thus suffering from an information loss issue from two aspects: 1) They downsample the input image into much lower resolutions for efficiency consideration. 2) They quantize $256^3$ RGB values to a small number (such as 512) of quantized color values. The indices of quantized pixels are used as tokens for the inputs and prediction targets of the transformer. To mitigate these issues, we propose a new transformer based framework called "PUT". Specifically, to avoid input downsampling while maintaining computation efficiency, we design a patch-based auto-encoder P-VQVAE. The encoder converts the masked image into non-overlapped patch tokens and the decoder recovers the masked regions from the inpainted tokens while keeping the unmasked regions unchanged. To eliminate the information loss caused by input quantization, an Un-quantized Transformer is applied. It directly takes features from the P-VQVAE encoder as input without any quantization and only regards the quantized tokens as prediction targets. Furthermore, to make the inpainting process more controllable, we introduce semantic and structural conditions as extra guidance. Extensive experiments show that our method greatly outperforms existing transformer based methods on image fidelity and achieves much higher diversity and better fidelity than state-of-the-art pluralistic inpainting methods on complex large-scale datasets (e.g., ImageNet). Codes are available at https://github.com/liuqk3/PUT.

Leveraging Large Language Models for Relevance Judgments in Legal Case Retrieval

Collecting relevant judgments for legal case retrieval is a challenging and time-consuming task. Accurately judging the relevance between two legal cases requires a considerable effort to read the lengthy text and a high level of domain expertise to extract Legal Facts and make juridical judgments. With the advent of advanced large language models, some recent studies have suggested that it is promising to use LLMs for relevance judgment. Nonetheless, the method of employing a general large language model for reliable relevance judgments in legal case retrieval is yet to be thoroughly explored. To fill this research gap, we devise a novel few-shot workflow tailored to the relevant judgment of legal cases. The proposed workflow breaks down the annotation process into a series of stages, imitating the process employed by human annotators and enabling a flexible integration of expert reasoning to enhance the accuracy of relevance judgments. By comparing the relevance judgments of LLMs and human experts, we empirically show that we can obtain reliable relevance judgments with the proposed workflow. Furthermore, we demonstrate the capacity to augment existing legal case retrieval models through the synthesis of data generated by the large language model.

MiM-ISTD: Mamba-in-Mamba for Efficient Infrared Small Target Detection

Recently, infrared small target detection (ISTD) has made significant progress, thanks to the development of basic models. Specifically, the structures combining convolutional networks with transformers can successfully extract both local and global features. However, the disadvantage of the transformer is also inherited, i.e., the quadratic computational complexity to the length of the sequence. Inspired by the recent basic model with linear complexity for long-distance modeling, called Mamba, we explore the potential of this state space model for ISTD task in terms of effectiveness and efficiency in the paper. However, directly applying Mamba achieves poor performance since local features, which are critical to detecting small targets, cannot be fully exploited. Instead, we tailor a Mamba-in-Mamba (MiM-ISTD) structure for efficient ISTD. Specifically, we treat the local patches as "visual sentences" and use the Outer Mamba to explore the global information. We then decompose each visual sentence into sub-patches as "visual words" and use the Inner Mamba to further explore the local information among words in the visual sentence with negligible computational costs. By aggregating the word and sentence features, the MiM-ISTD can effectively explore both global and local information. Experiments on NUAA-SIRST and IRSTD-1k show the superior accuracy and efficiency of our method. Specifically, MiM-ISTD is $10 \times$ faster than the SOTA method and reduces GPU memory usage by 73.4$\%$ when testing on $2048 \times 2048$ image, overcoming the computation and memory constraints on high-resolution infrared images. Source code is available at https://github.com/txchen-USTC/MiM-ISTD.

Bootstrapping Audio-Visual Segmentation by Strengthening Audio Cues

How to effectively interact audio with vision has garnered considerable interest within the multi-modality research field. Recently, a novel audio-visual segmentation (AVS) task has been proposed, aiming to segment the sounding objects in video frames under the guidance of audio cues. However, most existing AVS methods are hindered by a modality imbalance where the visual features tend to dominate those of the audio modality, due to a unidirectional and insufficient integration of audio cues. This imbalance skews the feature representation towards the visual aspect, impeding the learning of joint audio-visual representations and potentially causing segmentation inaccuracies. To address this issue, we propose AVSAC. Our approach features a Bidirectional Audio-Visual Decoder (BAVD) with integrated bidirectional bridges, enhancing audio cues and fostering continuous interplay between audio and visual modalities. This bidirectional interaction narrows the modality imbalance, facilitating more effective learning of integrated audio-visual representations. Additionally, we present a strategy for audio-visual frame-wise synchrony as fine-grained guidance of BAVD. This strategy enhances the share of auditory components in visual features, contributing to a more balanced audio-visual representation learning. Extensive experiments show that our method attains new benchmarks in AVS performance.

TCI-Former: Thermal Conduction-Inspired Transformer for Infrared Small Target Detection

Infrared small target detection (ISTD) is critical to national security and has been extensively applied in military areas. ISTD aims to segment small target pixels from background. Most ISTD networks focus on designing feature extraction blocks or feature fusion modules, but rarely describe the ISTD process from the feature map evolution perspective. In the ISTD process, the network attention gradually shifts towards target areas. We abstract this process as the directional movement of feature map pixels to target areas through convolution, pooling and interactions with surrounding pixels, which can be analogous to the movement of thermal particles constrained by surrounding variables and particles. In light of this analogy, we propose Thermal Conduction-Inspired Transformer (TCI-Former) based on the theoretical principles of thermal conduction. According to thermal conduction differential equation in heat dynamics, we derive the pixel movement differential equation (PMDE) in the image domain and further develop two modules: Thermal Conduction-Inspired Attention (TCIA) and Thermal Conduction Boundary Module (TCBM). TCIA incorporates finite difference method with PMDE to reach a numerical approximation so that target body features can be extracted. To further remove errors in boundary areas, TCBM is designed and supervised by boundary masks to refine target body features with fine boundary details. Experiments on IRSTD-1k and NUAA-SIRST demonstrate the superiority of our method.

Towards More Unified In-context Visual Understanding

The rapid advancement of large language models (LLMs) has accelerated the emergence of in-context learning (ICL) as a cutting-edge approach in the natural language processing domain. Recently, ICL has been employed in visual understanding tasks, such as semantic segmentation and image captioning, yielding promising results. However, existing visual ICL framework can not enable producing content across multiple modalities, which limits their potential usage scenarios. To address this issue, we present a new ICL framework for visual understanding with multi-modal output enabled. First, we quantize and embed both text and visual prompt into a unified representational space, structured as interleaved in-context sequences. Then a decoder-only sparse transformer architecture is employed to perform generative modeling on them, facilitating in-context learning. Thanks to this design, the model is capable of handling in-context vision understanding tasks with multimodal output in a unified pipeline. Experimental results demonstrate that our model achieves competitive performance compared with specialized models and previous ICL baselines. Overall, our research takes a further step toward unified multimodal in-context learning.

GUPNet++: Geometry Uncertainty Propagation Network for Monocular 3D Object Detection

Geometry plays a significant role in monocular 3D object detection. It can be used to estimate object depth by using the perspective projection between object's physical size and 2D projection in the image plane, which can introduce mathematical priors into deep models. However, this projection process also introduces error amplification, where the error of the estimated height is amplified and reflected into the projected depth. It leads to unreliable depth inferences and also impairs training stability. To tackle this problem, we propose a novel Geometry Uncertainty Propagation Network (GUPNet++) by modeling geometry projection in a probabilistic manner. This ensures depth predictions are well-bounded and associated with a reasonable uncertainty. The significance of introducing such geometric uncertainty is two-fold: (1). It models the uncertainty propagation relationship of the geometry projection during training, improving the stability and efficiency of the end-to-end model learning. (2). It can be derived to a highly reliable confidence to indicate the quality of the 3D detection result, enabling more reliable detection inference. Experiments show that the proposed approach not only obtains (state-of-the-art) SOTA performance in image-based monocular 3D detection but also demonstrates superiority in efficacy with a simplified framework.

Exploiting Modality-Specific Features For Multi-Modal Manipulation Detection And Grounding

AI-synthesized text and images have gained significant attention, particularly due to the widespread dissemination of multi-modal manipulations on the internet, which has resulted in numerous negative impacts on society. Existing methods for multi-modal manipulation detection and grounding primarily focus on fusing vision-language features to make predictions, while overlooking the importance of modality-specific features, leading to sub-optimal results. In this paper, we construct a simple and novel transformer-based framework for multi-modal manipulation detection and grounding tasks. Our framework simultaneously explores modality-specific features while preserving the capability for multi-modal alignment. To achieve this, we introduce visual/language pre-trained encoders and dual-branch cross-attention (DCA) to extract and fuse modality-unique features. Furthermore, we design decoupled fine-grained classifiers (DFC) to enhance modality-specific feature mining and mitigate modality competition. Moreover, we propose an implicit manipulation query (IMQ) that adaptively aggregates global contextual cues within each modality using learnable queries, thereby improving the discovery of forged details. Extensive experiments on the $\rm DGM^4$ dataset demonstrate the superior performance of our proposed model compared to state-of-the-art approaches.