SlimGPT: Layer-wise Structured Pruning for Large Language Models

Large language models (LLMs) have garnered significant attention for their remarkable capabilities across various domains, whose vast parameter scales present challenges for practical deployment. Structured pruning is an effective method to balance model performance with efficiency, but performance restoration under computational resource constraints is a principal challenge in pruning LLMs. Therefore, we present a low-cost and fast structured pruning method for LLMs named SlimGPT based on the Optimal Brain Surgeon framework. We propose Batched Greedy Pruning for rapid and near-optimal pruning, which enhances the accuracy of head-wise pruning error estimation through grouped Cholesky decomposition and improves the pruning efficiency of FFN via Dynamic Group Size, thereby achieving approximate local optimal pruning results within one hour. Besides, we explore the limitations of layer-wise pruning from the perspective of error accumulation and propose Incremental Pruning Ratio, a non-uniform pruning strategy to reduce performance degradation. Experimental results on the LLaMA benchmark show that SlimGPT outperforms other methods and achieves state-of-the-art results.

Resonant Beam Multi-Target DOA Estimation

With the increasing demand for internet of things (IoT) applications, especially for location-based services, how to locate passive mobile targets (MTs) with minimal beam control has become a challenge. Resonant beam systems are considered promising IoT technologies with advantages such as beam self-alignment and energy concentration. To establish a resonant system in the radio frequency (RF) band and achieve multi-target localization, this paper designs a multi-target resonant system architecture, allowing a single base station (BS) to independently connect with multiple MTs. By employing a retro-directive array, a multi-channel cyclic model is established to realize one-to-many electromagnetic wave propagation and MT direction-of-arrival (DOA) estimation through echo resonance. Simulation results show that the proposed system supports resonant establishment between the BS and multiple MTs. This helps the BS to still have high DOA estimation accuracy in the face of multiple passive MTs, and can ensure that the DOA error is less than 1 degree within a range of 6 meters at a 50degree field of view, with higher accuracy than active beamforming localization systems.

Resonant Beam Enabled Passive 3D Positioning

With the rapid development of the internet of things (IoT), location-based services are becoming increasingly prominent in various aspects of social life, and accurate location information is crucial. However, RF-based indoor positioning solutions are severely limited in positioning accuracy due to signal transmission losses and directional difficulties, and optical indoor positioning methods require high propagation conditions. To achieve higher accuracy in indoor positioning, we utilize the principle of resonance to design a triangulation-based resonant beam positioning system (TRBPS) in the RF band. The proposed system employs phase-conjugation antenna arrays and resonance mechanism to achieve energy concentration and beam self-alignment, without requiring active signals from the target for positioning and complex beam control algorithms. Numerical evaluations indicate that TRBPS can achieve millimeter-level accuracy within a range of 3.6 m without the need for additional embedded systems.

No More Tuning: Prioritized Multi-Task Learning with Lagrangian Differential Multiplier Methods

Given the ubiquity of multi-task in practical systems, Multi-Task Learning (MTL) has found widespread application across diverse domains. In real-world scenarios, these tasks often have different priorities. For instance, In web search, relevance is often prioritized over other metrics, such as click-through rates or user engagement. Existing frameworks pay insufficient attention to the prioritization among different tasks, which typically adjust task-specific loss function weights to differentiate task priorities. However, this approach encounters challenges as the number of tasks grows, leading to exponential increases in hyper-parameter tuning complexity. Furthermore, the simultaneous optimization of multiple objectives can negatively impact the performance of high-priority tasks due to interference from lower-priority tasks. In this paper, we introduce a novel multi-task learning framework employing Lagrangian Differential Multiplier Methods for step-wise multi-task optimization. It is designed to boost the performance of high-priority tasks without interference from other tasks. Its primary advantage lies in its ability to automatically optimize multiple objectives without requiring balancing hyper-parameters for different tasks, thereby eliminating the need for manual tuning. Additionally, we provide theoretical analysis demonstrating that our method ensures optimization guarantees, enhancing the reliability of the process. We demonstrate its effectiveness through experiments on multiple public datasets and its application in Taobao search, a large-scale industrial search ranking system, resulting in significant improvements across various business metrics.

Can video generation replace cinematographers? Research on the cinematic language of generated video

Recent advancements in text-to-video (T2V) generation have leveraged diffusion models to enhance the visual coherence of videos generated from textual descriptions. However, most research has primarily focused on object motion, with limited attention given to cinematic language in videos, which is crucial for cinematographers to convey emotion and narrative pacing. To address this limitation, we propose a threefold approach to enhance the ability of T2V models to generate controllable cinematic language. Specifically, we introduce a cinematic language dataset that encompasses shot framing, angle, and camera movement, enabling models to learn diverse cinematic styles. Building on this, to facilitate robust cinematic alignment evaluation, we present CameraCLIP, a model fine-tuned on the proposed dataset that excels in understanding complex cinematic language in generated videos and can further provide valuable guidance in the multi-shot composition process. Finally, we propose CLIPLoRA, a cost-guided dynamic LoRA composition method that facilitates smooth transitions and realistic blending of cinematic language by dynamically fusing multiple pre-trained cinematic LoRAs within a single video. Our experiments demonstrate that CameraCLIP outperforms existing models in assessing the alignment between cinematic language and video, achieving an R@1 score of 0.81. Additionally, CLIPLoRA improves the ability for multi-shot composition, potentially bridging the gap between automatically generated videos and those shot by professional cinematographers.

Advancing Myopia To Holism: Fully Contrastive Language-Image Pre-training

In rapidly evolving field of vision-language models (VLMs), contrastive language-image pre-training (CLIP) has made significant strides, becoming foundation for various downstream tasks. However, relying on one-to-one (image, text) contrastive paradigm to learn alignment from large-scale messy web data, CLIP faces a serious myopic dilemma, resulting in biases towards monotonous short texts and shallow visual expressivity. To overcome these issues, this paper advances CLIP into one novel holistic paradigm, by updating both diverse data and alignment optimization. To obtain colorful data with low cost, we use image-to-text captioning to generate multi-texts for each image, from multiple perspectives, granularities, and hierarchies. Two gadgets are proposed to encourage textual diversity. To match such (image, multi-texts) pairs, we modify the CLIP image encoder into multi-branch, and propose multi-to-multi contrastive optimization for image-text part-to-part matching. As a result, diverse visual embeddings are learned for each image, bringing good interpretability and generalization. Extensive experiments and ablations across over ten benchmarks indicate that our holistic CLIP significantly outperforms existing myopic CLIP, including image-text retrieval, open-vocabulary classification, and dense visual tasks.

Branches, Assemble! Multi-Branch Cooperation Network for Large-Scale Click-Through Rate Prediction at Taobao

Existing click-through rate (CTR) prediction works have studied the role of feature interaction through a variety of techniques. Each interaction technique exhibits its own strength, and solely using one type could constrain the model's capability to capture the complex feature relationships, especially for industrial large-scale data with enormous users and items. Recent research shows that effective CTR models often combine an MLP network with a dedicated feature interaction network in a two-parallel structure. However, the interplay and cooperative dynamics between different streams or branches remain under-researched. In this work, we introduce a novel Multi-Branch Cooperation Network (MBCnet) which enables multiple branch networks to collaborate with each other for better complex feature interaction modeling. Specifically, MBCnet consists of three branches: the Expert-based Feature Grouping and Crossing (EFGC) branch that promotes the model's memorization ability of specific feature fields, the low rank Cross Net branch and Deep branch to enhance both explicit and implicit feature crossing for improved generalization. Among branches, a novel cooperation scheme is proposed based on two principles: branch co-teaching and moderate differentiation. Branch co-teaching encourages well-learned branches to support poorly-learned ones on specific training samples. Moderate differentiation advocates branches to maintain a reasonable level of difference in their feature representations. The cooperation strategy improves learning through mutual knowledge sharing via co-teaching and boosts the discovery of diverse feature interactions across branches. Extensive experiments on large-scale industrial datasets and online A/B test demonstrate MBCnet's superior performance, delivering a 0.09 point increase in CTR, 1.49% growth in deals, and 1.62% rise in GMV. Core codes will be released soon.

Field of View Expansion for Resonant Beam Information and Power Transfer

Simultaneous wireless information and power transfer (SWIPT) leverages lightwave as the wireless transmission medium, emerging as a promising technology in the future Internet of Things (IoT) scenarios. The use of retro-reflectors in constructing spatially separated laser resonators (SSLR) enables a self-aligning wireless transmission system with the self-reproducing resonant beam, i.e. resonant beam system (RBS). However, it's effective Field of View (FoV) is physically limited by the size of retroreflectors and still requires significant improvement. This restricts the transmitter from providing seamless wireless connectivity and power supply to receivers within a large dynamic movement range. In this paper, we propose an FoV-enlarged resonant beam system operating at a meter distance by incorporating a telescope. The telescope plays a crucial role in minimizing the extra loss inflicted on the gain medium, which typically arises from the deviation of the resonant beam within the cavity. Further, we construct the proposed telescope-based RBS and experimentally demonstrate that the design could expand the FoV to 28$^\circ$ over 1 m transmission distance is about triple that of the ordinary RBS design.

Resonant Beam Enabled DoA Estimation in Passive Positioning System

The rapid advancement of the next generation of communications and internet of things (IoT) technologies has made the provision of location-based services for diverse devices an increasingly pressing necessity. Localizing devices with/without intelligent computing abilities, including both active and passive devices is essential, especially in indoor scenarios. For traditional RF positioning systems, aligning transmission signals and dealing with signal interference in complex environments are inevitable challenges. Therefore, this paper proposed a new passive positioning system, the RF-band resonant beam positioning system (RF-RBPS), which achieves energy concentration and beam alignment by amplifying echoes between the base station (BS) and the passive target (PT), without the need for complex channel estimation and time-consuming beamforming and provides high-precision direction of arrival (DoA) estimation for battery-free targets using the resonant mechanism. The direction information of the PT is estimated using the multiple signal classification (MUSIC) algorithm at the end of BS. The feasibility of the proposed system is validated through theoretical analysis and simulations. Results indicate that the proposed RF-RBPS surpasses RF-band active positioning system (RF-APS) in precision, achieving millimeter-level precision at 2m within an elevation angle of 35$^\circ$, and an error of less than 3cm at 2.5m within an elevation angle of 35$^\circ$.

NoiseBoost: Alleviating Hallucination with Noise Perturbation for Multimodal Large Language Models

Multimodal large language models (MLLMs) contribute a powerful mechanism to understanding visual information building on large language models. However, MLLMs are notorious for suffering from hallucinations, especially when generating lengthy, detailed descriptions for images. Our analysis reveals that hallucinations stem from the inherent summarization mechanism of large language models, leading to excessive dependence on linguistic tokens while neglecting vision information. In this paper, we propose NoiseBoost, a broadly applicable and simple method for alleviating hallucinations for MLLMs through the integration of noise feature perturbations. Noise perturbation acts as a regularizer, facilitating a balanced distribution of attention weights among visual and linguistic tokens. Despite its simplicity, NoiseBoost consistently enhances the performance of MLLMs across common training strategies, including supervised fine-tuning and reinforcement learning. Further, NoiseBoost pioneerly enables semi-supervised learning for MLLMs, unleashing the power of unlabeled data. Comprehensive experiments demonstrate that NoiseBoost improves dense caption accuracy by 8.1% with human evaluation and achieves comparable results with 50% of the data by mining unlabeled data. Code and models are available at https://kaiwu5.github.io/noiseboost.