Large Language Models (LLMs) with long chain-of-thought (CoT) capability, termed Reasoning Models, demonstrate superior intricate problem-solving abilities through multi-step long CoT reasoning. To create a dual-capability model with long CoT capability and domain-specific knowledge without substantial computational and data costs, model merging emerges as a highly resource-efficient method. However, significant challenges lie in merging domain-specific LLMs with long CoT ones since nowadays merging methods suffer from reasoning capability degradation, even gibberish output and output collapse. To overcome this, we introduce RCP-Merging: Merging Long Chain-of-Thought Models with Domain-Specific Models by Considering Reasoning Capability as Prior, a novel merging framework designed to integrate domain-specific LLMs with long CoT capability, meanwhile maintaining model performance in the original domain. Treating reasoning model weights as foundational prior, our method utilizes a reasoning capability indicator to preserve core long CoT capability model weights while selectively merging essential domain-specific weights. We conducted extensive experiments on Qwen2.5-7B, Llama3.1-8B, and Qwen2.5-1.5B models in BioMedicine and Finance domains. Our results show that RCP-Merging successfully merges a reasoning model with domain-specific ones, improving domain task performance by 9.5% and 9.2% over state-of-the-art methods, without significantly harming the original long CoT reasoning capability.

Large Reasoning Models (LRMs) with long chain-of-thought reasoning have recently achieved remarkable success.Yet, equipping domain-specialized models with such reasoning capabilities, referred to as "Reasoning + X", remains a significant challenge.While model merging offers a promising training-free solution, existing methods often suffer from a destructive performance collapse: existing methods tend to both weaken reasoning depth and compromise domain-specific utility.Interestingly, we identify a counter-intuitive phenomenon underlying this failure: reasoning ability predominantly resides in parameter regions with low gradient sensitivity, contrary to the common assumption that domain capabilities correspond to high-magnitude parameters.Motivated by this insight, we propose ReasonAny, a novel merging framework that resolves the reasoning-domain performance collapse through Contrastive Gradient Identification.Experiments across safety, biomedicine, and finance domains show that ReasonAny effectively synthesizes "Reasoning + X" capabilities, significantly outperforming state-of-theart baselines while retaining robust reasoning performance.

Ziqian Zeng, Jianwei Wang, Junyao Yang, Zhengdong Lu, Haoran Li, Huiping Zhuang, Cen Chen. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

The success of large language models (LLMs) has attracted many individuals to fine-tune them for domain-specific tasks by uploading their data.However, in sensitive areas like healthcare and finance, privacy concerns often arise.One promising solution is to generate synthetic data with Differential Privacy (DP) guarantees to replace private data.However, these synthetic data contain significant flawed data, which are considered as noise.Existing solutions typically rely on naive filtering by comparing ROUGE-L scores or embedding similarities, which are ineffective in addressing the noise.To address this issue, we propose RewardDS, a novel privacy-preserving framework that fine-tunes a reward proxy model and uses reward signals to guide the synthetic data generation.Our RewardDS introduces two key modules, Reward Guided Filtering and Self-Optimizing Refinement, to both filter and refine the synthetic data, effectively mitigating the noise.Extensive experiments across medical, financial, and code generation domains demonstrate the effectiveness of our method.

Dual-energy X-ray absorptiometry (DXA) enables accurate bone mineral density but requires specialized equipment and protocols. X-ray-based BMD screening offers opportunistic early detection, though prior methods struggle with X-ray intensity variations and demand large datasets. We introduce a cross-modal knowledge distillation BMD prediction framework (CMKD-BMD) fusing X-ray/CT data to enhance X-ray-only BMD prediction. Each single-modal student network employs a multi-scale visual extractor for hierarchical features, an unsupervised graph-based structural learner for anatomical relationships, and an adaptive fusion module to generate a unified representation. The teacher network integrates single-modal representations from students and transfers multimodal knowledge from the teacher to students. Our model outperforms existing methods on both the collected dataset (comprising 1620 X-rays and 280 CT cases) and the publicly available VerSe2019 dataset, demonstrating superior BMD estimation performance. Furthermore, we developed OrthoSim, an orthopedic surgical simulation platform with CMKD-BMD, which has shown promising clinical effectiveness in trial evaluations. Our code is available at https://github.com/KeyueShi/CMKD-BMD.

In-memory key-value caches are widely used as a performance-critical layer in web applications, disk-based storage, and distributed systems. The Least Recently Used (LRU) replacement policy has become thede factostandard in those systems since it exploits workload locality well. However, the LRU implementation can be costly due to the rigid data structure in maintaining object priority, as well as the locks for object order updating. Redis as one of the most effective and prevalent deployed commercial systems adopts an approximated LRU policy, where the least recently used item from a small, randomly sampled set of items is chosen to evict. This random sampling-based policy is lightweight and shows its flexibility. We observe that there can exist a significant miss ratio gap between exact LRU and random sampling-based LRU under different sampling size$K$s. Therefore existing LRU miss ratio curve (MRC) construction techniques cannot be directly applied without loss of accuracy. In this article, we introduce a new probabilistic stack algorithm namedKRRto accurately model random sampling based-LRU, and extend it to handle both fixed and variable objects in key-value caches. We present an efficient stack update algorithm that reduces the expected running time of KRR significantly. To improve the performance of the in-memory multi-tenant key-value cache that utilizes random sampling-based replacement, we propose kRedis, a reference locality- and latency-aware memory partitioning scheme. kRedis guides the memory allocation among the tenants and dynamically customizes$K$to better exploit the locality of each individual tenant. Evaluation results over diverse workloads show that our model generates accurate miss ratio curves for both fixed and variable object size workloads, and enables practical, low-overhead online MRC prediction. Equipped with KRR, kRedis delivers up to a 50.2% average access latency reduction, and up to a 262.8% throughput improvement compared to Redis. Furthermore, by comparing with pRedis, a state-of-the-art design of memory allocation in Redis, kRedis shows up to 24.8% and 61.8% improvements in average access latency and throughput, respectively.

Unsupervised learning methods have achieved remarkable performance in monocular depth estimation and camera pose, which mostly solve the multi-task learning problem by using their inner geometry consistency as the self-supervision signal. While most existing approaches mostly adopt the generative model to obtain the depth map prediction, so in the resolution of depth map there is room for improvement. To this end, we present our unsupervised learning architecture based on adversarial learning model, which is used for unsupervised learning of high-resolution single view depth and camera pose. Specifically, we present a multi-scale deep convolutional Generative Adversarial Network (GAN) based learning system, which consists of three networks (pose estimation network PCNN, Generator-D and Discriminator-D for depth map prediction). Furthermore, in order to generate high-resolution depth map, we propose a multi-scale GAN model (MSGAN) to decompose the hard high-quality image generation problem into more manageable sub-problems through a coarse-to-fine process. Then, we modify the overall generation architecture of GAN model by changing the down-sampling and up-sampling components to improve the quality and accuracy of the depth map prediction. Finally, in order to improve the rate of convergence, we use the Least Square Error to increase the penalty for outliers. Detailed quantitative and qualitative evaluations of the proposed framework on the KITTI dataset show that the proposed method provides better results for both pose estimation and depth recovery.

The Miss Ratio Curve (MRC) is an important metric and effective tool for caching system performance prediction and optimization. Since the Least Recently Used (LRU) replacement policy is the de facto policy for many existing caching systems, most previous studies on efficient MRC construction are predominantly focused on the LRU replacement policy. Recently, the random sampling-based replacement mechanism, as opposed to replacement relying on the rigid LRU data structure, gains more popularity due to its lightweight and flexibility. To approximate LRU, at replacement times, the system randomly selects K objects and replaces the least recently used object among the sample. Redis implements this approximated LRU policy. We observe that there can exist a significant miss ratio gap between exact LRU and random sampling-based LRU under different sampling size K; therefore existing LRU MRC construction techniques cannot be directly applied to random sampling based LRU cache without loss of accuracy.