Time's Up! An Empirical Study of LLM Reasoning Ability Under Output Length Constraint

Recent work has demonstrated the remarkable potential of Large Language Models (LLMs) in test-time scaling. By making the models think before answering, they are able to achieve much higher accuracy with extra inference computation. However, in many real-world scenarios, models are used under time constraints, where an answer should be given to the user within a certain output length. It is unclear whether and how the reasoning abilities of LLMs remain effective under such constraints. We take a first look at this problem by conducting an in-depth empirical study. Specifically, we test more than 25 LLMs on common reasoning datasets under a wide range of output length budgets, and we analyze the correlation between the inference accuracy and various properties including model type, model size, prompt style, etc. We also consider the mappings between the token budgets and the actual on-device latency budgets. The results have demonstrated several interesting findings regarding the budget-aware LLM reasoning that differ from the unconstrained situation, e.g. the optimal choices of model sizes and prompts change under different budgets. These findings offer practical guidance for users to deploy LLMs under real-world latency constraints.

Adaptive Wavelet Filters as Practical Texture Feature Amplifiers for Parkinson's Disease Screening in OCT

Parkinson's disease (PD) is a prevalent neurodegenerative disorder globally. The eye's retina is an extension of the brain and has great potential in PD screening. Recent studies have suggested that texture features extracted from retinal layers can be adopted as biomarkers for PD diagnosis under optical coherence tomography (OCT) images. Frequency domain learning techniques can enhance the feature representations of deep neural networks (DNNs) by decomposing frequency components involving rich texture features. Additionally, previous works have not exploited texture features for automated PD screening in OCT. Motivated by the above analysis, we propose a novel Adaptive Wavelet Filter (AWF) that serves as the Practical Texture Feature Amplifier to fully leverage the merits of texture features to boost the PD screening performance of DNNs with the aid of frequency domain learning. Specifically, AWF first enhances texture feature representation diversities via channel mixer, then emphasizes informative texture feature representations with the well-designed adaptive wavelet filtering token mixer. By combining the AWFs with the DNN stem, AWFNet is constructed for automated PD screening. Additionally, we introduce a novel Balanced Confidence (BC) Loss by mining the potential of sample-wise predicted probabilities of all classes and class frequency prior, to further boost the PD screening performance and trustworthiness of AWFNet. The extensive experiments manifest the superiority of our AWFNet and BC over state-of-the-art methods in terms of PD screening performance and trustworthiness.

Iterative Optimal Attention and Local Model for Single Image Rain Streak Removal

High-fidelity imaging is crucial for the successful safety supervision and intelligent deployment of vision-based measurement systems (VBMS). It ensures high-quality imaging in VBMS, which is fundamental for reliable visual measurement and analysis. However, imaging quality can be significantly impaired by adverse weather conditions, particularly rain, leading to blurred images and reduced contrast. Such impairments increase the risk of inaccurate evaluations and misinterpretations in VBMS. To address these limitations, we propose an Expectation Maximization Reconstruction Transformer (EMResformer) for single image rain streak removal. The EMResformer retains the key self-attention values for feature aggregation, enhancing local features to produce superior image reconstruction. Specifically, we propose an Expectation Maximization Block seamlessly integrated into the single image rain streak removal network, enhancing its ability to eliminate superfluous information and restore a cleaner background image. Additionally, to further enhance local information for improved detail rendition, we introduce a Local Model Residual Block, which integrates two local model blocks along with a sequence of convolutions and activation functions. This integration synergistically facilitates the extraction of more pertinent features for enhanced single image rain streak removal. Extensive experiments validate that our proposed EMResformer surpasses current state-of-the-art single image rain streak removal methods on both synthetic and real-world datasets, achieving an improved balance between model complexity and single image deraining performance. Furthermore, we evaluate the effectiveness of our method in VBMS scenarios, demonstrating that high-quality imaging significantly improves the accuracy and reliability of VBMS tasks.

RAD: Training an End-to-End Driving Policy via Large-Scale 3DGS-based Reinforcement Learning

Existing end-to-end autonomous driving (AD) algorithms typically follow the Imitation Learning (IL) paradigm, which faces challenges such as causal confusion and the open-loop gap. In this work, we establish a 3DGS-based closed-loop Reinforcement Learning (RL) training paradigm. By leveraging 3DGS techniques, we construct a photorealistic digital replica of the real physical world, enabling the AD policy to extensively explore the state space and learn to handle out-of-distribution scenarios through large-scale trial and error. To enhance safety, we design specialized rewards that guide the policy to effectively respond to safety-critical events and understand real-world causal relationships. For better alignment with human driving behavior, IL is incorporated into RL training as a regularization term. We introduce a closed-loop evaluation benchmark consisting of diverse, previously unseen 3DGS environments. Compared to IL-based methods, RAD achieves stronger performance in most closed-loop metrics, especially 3x lower collision rate. Abundant closed-loop results are presented at https://hgao-cv.github.io/RAD.

HedgeAgents: A Balanced-aware Multi-agent Financial Trading System

As automated trading gains traction in the financial market, algorithmic investment strategies are increasingly prominent. While Large Language Models (LLMs) and Agent-based models exhibit promising potential in real-time market analysis and trading decisions, they still experience a significant -20% loss when confronted with rapid declines or frequent fluctuations, impeding their practical application. Hence, there is an imperative to explore a more robust and resilient framework. This paper introduces an innovative multi-agent system, HedgeAgents, aimed at bolstering system robustness via ``hedging'' strategies. In this well-balanced system, an array of hedging agents has been tailored, where HedgeAgents consist of a central fund manager and multiple hedging experts specializing in various financial asset classes. These agents leverage LLMs' cognitive capabilities to make decisions and coordinate through three types of conferences. Benefiting from the powerful understanding of LLMs, our HedgeAgents attained a 70% annualized return and a 400% total return over a period of 3 years. Moreover, we have observed with delight that HedgeAgents can even formulate investment experience comparable to those of human experts (https://hedgeagents.github.io/).

MedFILIP: Medical Fine-grained Language-Image Pre-training

Medical vision-language pretraining (VLP) that leverages naturally-paired medical image-report data is crucial for medical image analysis. However, existing methods struggle to accurately characterize associations between images and diseases, leading to inaccurate or incomplete diagnostic results. In this work, we propose MedFILIP, a fine-grained VLP model, introduces medical image-specific knowledge through contrastive learning, specifically: 1) An information extractor based on a large language model is proposed to decouple comprehensive disease details from reports, which excels in extracting disease deals through flexible prompt engineering, thereby effectively reducing text complexity while retaining rich information at a tiny cost. 2) A knowledge injector is proposed to construct relationships between categories and visual attributes, which help the model to make judgments based on image features, and fosters knowledge extrapolation to unfamiliar disease categories. 3) A semantic similarity matrix based on fine-grained annotations is proposed, providing smoother, information-richer labels, thus allowing fine-grained image-text alignment. 4) We validate MedFILIP on numerous datasets, e.g., RSNA-Pneumonia, NIH ChestX-ray14, VinBigData, and COVID-19. For single-label, multi-label, and fine-grained classification, our model achieves state-of-the-art performance, the classification accuracy has increased by a maximum of 6.69\%. The code is available in https://github.com/PerceptionComputingLab/MedFILIP.

Coverage and Spectral Efficiency of NOMA-Enabled LEO Satellite Networks with Ordering Schemes

This paper investigates an analytical model for low-earth orbit (LEO) multi-satellite downlink non-orthogonal multiple access (NOMA) networks. The satellites transmit data to multiple NOMA user terminals (UTs), each employing successive interference cancellation (SIC) for decoding. Two ordering schemes are adopted for NOMA-enabled LEO satellite networks, i.e., mean signal power (MSP)-based ordering and instantaneous-signal-to-inter-satellite-interference-plus-noise ratio (ISINR)-based ordering. For each ordering scheme, we derive the coverage probabilities of UTs under different channel conditions. Moreover, we discuss how coverage is influenced by SIC, main-lobe gain, and tradeoffs between the number of satellites and their altitudes. Additionally, two user fairness-based power allocation (PA) schemes are considered, and PA coefficients with the optimal number of UTs that maximize their sum spectral efficiency (SE) are studied. Simulation results show that there exists a maximum signal-to-inter-satellite-interference-plus-noise ratio (SINR) threshold for each PA scheme that ensures the operation of NOMA in LEO satellite networks, and the benefit of NOMA only exists when the target SINR is below a certain threshold. Compared with orthogonal multiple access (OMA), NOMA increases UTs' sum SE by as much as 35\%. Furthermore, for most SINR thresholds, the sum SE increases with the number of UTs to the highest value, whilst the maximum sum SE is obtained when there are two UTs.

Downlink Performance of Cell-Free Massive MIMO for LEO Satellite Mega-Constellation

Low-earth orbit (LEO) satellite communication (SatCom) has emerged as a promising technology for improving wireless connectivity in global areas. Cell-free massive multiple-input multiple-output (CF-mMIMO), an architecture recently proposed for next-generation networks, has yet to be fully explored for LEO satellites. In this paper, we investigate the downlink performance of a CF-mMIMO LEO SatCom network, where many satellite access points (SAPs) simultaneously serve the corresponding ground user terminals (UTs). Using tools from stochastic geometry, we model the locations of SAPs and UTs on surfaces of concentric spheres using Poisson point processes (PPPs) and present expressions based on linear minimum-mean-square-error (LMMSE) channel estimation and conjugate beamforming. Then, we derive the coverage probabilities in both fading and non-fading scenarios, with significant system parameters such as the Nakagami fading parameter, number of UTs, number of SAPs, orbital altitude, and service range brought by the dome angle. Finally, the analytical model is verified by extensive Monte Carlo simulations. Simulation results show that stronger line-of-sight (LoS) effects and a more comprehensive service range of the UT bring higher coverage probability despite existing multi-user interference. Moreover, we found that there exist optimal numbers of UTs for different orbital altitudes and dome angles, which provides valuable system design insights.

Spectrum Sharing in Satellite-Terrestrial Integrated Networks: Frameworks, Approaches, and Opportunities

To accommodate the increasing communication needs in non-terrestrial networks (NTNs), wireless users in remote areas may require access to more spectrum than is currently allocated. Terrestrial networks (TNs), such as cellular networks, are deployed in specific areas, but many underused licensed spectrum bands remain in remote areas. Therefore, bringing NTNs to a shared spectrum with TNs can improve network capacity under reasonable interference management. However, in satellite-terrestrial integrated networks (STINs), the comprehensive coverage of a satellite and the unbalanced communication resources of STINs make it challenging to effectively manage mutual interference between NTN and TN. This article presents the fundamentals and prospects of spectrum sharing (SS) in STINs by introducing four SS frameworks, their potential application scenarios, and technical challenges. Furthermore, advanced SS approaches related to interference management in STINs and performance metrics of SS in STINs are introduced. Moreover, a preliminary performance evaluation showcases the potential for sharing the spectrum between NTN and TN. Finally, future research opportunities for SS in STINs are discussed.

Spectrum Sharing in 6G Space-Ground Integrated Networks: A Ground Protection Zone-Based Design

Space-ground integrated network (SGIN) has been envisioned as a competitive solution for large scale and wide coverage of future wireless networks. By integrating both the non-terrestrial network (NTN) and the terrestrial network (TN), SGIN can provide high speed and omnipresent wireless network access for the users using the predefined licensed spectrums. Considering the scarcity of the spectrum resource and the low spectrum efficiency of the SGIN, we enable the NTN and TN to share the spectrum to improve overall system performance, i.e., weighted-sum area data rate (WS-ADR). However, mutual interference between NTN and TN is often inevitable and thus causes SGIN performance degradation. In this work, we consider a ground protection zone for the TN base stations, in which the NTN users are only allowed to use the NTN reserved spectrum to mitigate the NTN and TN mutual interference. We analytically derive the coverage probability and area data rate (ADR) of the typical users and study the performance under various protection zone sizes and spectrum allocation parameter settings. Simulation and numerical results demonstrate that the WS-ADR could be maximized by selecting the appropriate radius of protection zone and bandwidth allocation factor in the SGIN.