HIPPO: Accelerating Video Large Language Models Inference via Holistic-aware Parallel Speculative Decoding

Speculative decoding (SD) has emerged as a promising approach to accelerate LLM inference without sacrificing output quality. Existing SD methods tailored for video-LLMs primarily focus on pruning redundant visual tokens to mitigate the computational burden of massive visual inputs. However, existing methods do not achieve inference acceleration comparable to text-only LLMs. We observe from extensive experiments that this phenomenon mainly stems from two limitations: (i) their pruning strategies inadequately preserve visual semantic tokens, degrading draft quality and acceptance rates; (ii) even with aggressive pruning (e.g., 90% visual tokens removed), the draft model's remaining inference cost limits overall speedup. To address these limitations, we propose HIPPO, a general holistic-aware parallel speculative decoding framework. Specifically, HIPPO proposes (i) a semantic-aware token preservation method, which fuses global attention scores with local visual semantics to retain semantic information at high pruning ratios; (ii) a video parallel SD algorithm that decouples and overlaps draft generation and target verification phases. Experiments on four video-LLMs across six benchmarks demonstrate HIPPO's effectiveness, yielding up to 3.51x speedup compared to vanilla auto-regressive decoding.

LeJOT: An Intelligent Job Cost Orchestration Solution for Databricks Platform

With the rapid advancements in big data technologies, the Databricks platform has become a cornerstone for enterprises and research institutions, offering high computational efficiency and a robust ecosystem. However, managing the escalating operational costs associated with job execution remains a critical challenge. Existing solutions rely on static configurations or reactive adjustments, which fail to adapt to the dynamic nature of workloads. To address this, we introduce LeJOT, an intelligent job cost orchestration framework that leverages machine learning for execution time prediction and a solver-based optimization model for real-time resource allocation. Unlike conventional scheduling techniques, LeJOT proactively predicts workload demands, dynamically allocates computing resources, and minimizes costs while ensuring performance requirements are met. Experimental results on real-world Databricks workloads demonstrate that LeJOT achieves an average 20% reduction in cloud computing costs within a minute-level scheduling timeframe, outperforming traditional static allocation strategies. Our approach provides a scalable and adaptive solution for cost-efficient job scheduling in Data Lakehouse environments.

CLO: Efficient LLM Inference System with CPU-Light KVCache Offloading via Algorithm-System Co-Design

The growth of million-token LLMs exposes the scalability limits of inference systems, where the KVCache dominates memory usage and data transfer overhead. Recent offloading systems migrate the KVCache to CPU memory and incorporate top-k attention to reduce the volume of data transferred from the CPU, while further applying system-level optimizations such as on-GPU caching and prefetching to lower transfer overhead. However, they overlook the CPU bottleneck in three aspects: (1) substantial overhead of fine-grained dynamic cache management performed on the CPU side, (2) significant transfer overhead from poor PCIe bandwidth utilization caused by heavy gathering operations at the CPU side, and (3) GPU runtime bubbles introduced by coarse-grained CPU-centric synchronization. To address these challenges, we propose CLO, a CPU-light KVCache offloading system via algorithm-system co-design. CLO features: (1) a coarse-grained head-wise approximate on-GPU caching strategy with negligible cache management cost, (2) seamless combination of data prefetching and on-GPU persistent caching for lower transfer overhead, (3) a zero-copy transfer engine to fully exploit PCIe bandwidth, and a GPU-centric synchronization method to eliminate GPU stalls. Evaluation on two widely-used LLMs demonstrates that CLO achieves comparable accuracy to state-of-the-art systems, while substantially minimizing CPU overhead, fully utilizing PCIe bandwidth, thus improving decoding throughput by 9.3%-66.6%. Our results highlight that algorithm-system co-design is essential for memory-constrained LLM inference on modern GPU platforms. We open source CLO at https://github.com/CommediaJW/CLO.

Scalable Mixed-Integer Optimization with Neural Constraints via Dual Decomposition

Embedding deep neural networks (NNs) into mixed-integer programs (MIPs) is attractive for decision making with learned constraints, yet state-of-the-art monolithic linearisations blow up in size and quickly become intractable. In this paper, we introduce a novel dual-decomposition framework that relaxes the single coupling equality u=x with an augmented Lagrange multiplier and splits the problem into a vanilla MIP and a constrained NN block. Each part is tackled by the solver that suits it best-branch and cut for the MIP subproblem, first-order optimisation for the NN subproblem-so the model remains modular, the number of integer variables never grows with network depth, and the per-iteration cost scales only linearly with the NN size. On the public \textsc{SurrogateLIB} benchmark, our method proves \textbf{scalable}, \textbf{modular}, and \textbf{adaptable}: it runs \(120\times\) faster than an exact Big-M formulation on the largest test case; the NN sub-solver can be swapped from a log-barrier interior step to a projected-gradient routine with no code changes and identical objective value; and swapping the MLP for an LSTM backbone still completes the full optimisation in 47s without any bespoke adaptation.

Amplifying Prominent Representations in Multimodal Learning via Variational Dirichlet Process

Developing effective multimodal fusion approaches has become increasingly essential in many real-world scenarios, such as health care and finance. The key challenge is how to preserve the feature expressiveness in each modality while learning cross-modal interactions. Previous approaches primarily focus on the cross-modal alignment, while over-emphasis on the alignment of marginal distributions of modalities may impose excess regularization and obstruct meaningful representations within each modality. The Dirichlet process (DP) mixture model is a powerful Bayesian non-parametric method that can amplify the most prominent features by its richer-gets-richer property, which allocates increasing weights to them. Inspired by this unique characteristic of DP, we propose a new DP-driven multimodal learning framework that automatically achieves an optimal balance between prominent intra-modal representation learning and cross-modal alignment. Specifically, we assume that each modality follows a mixture of multivariate Gaussian distributions and further adopt DP to calculate the mixture weights for all the components. This paradigm allows DP to dynamically allocate the contributions of features and select the most prominent ones, leveraging its richer-gets-richer property, thus facilitating multimodal feature fusion. Extensive experiments on several multimodal datasets demonstrate the superior performance of our model over other competitors. Ablation analysis further validates the effectiveness of DP in aligning modality distributions and its robustness to changes in key hyperparameters. Code is anonymously available at https://github.com/HKU-MedAI/DPMM.git

CapRL: Stimulating Dense Image Caption Capabilities via Reinforcement Learning

Image captioning is a fundamental task that bridges the visual and linguistic domains, playing a critical role in pre-training Large Vision-Language Models (LVLMs). Current state-of-the-art captioning models are typically trained with Supervised Fine-Tuning (SFT), a paradigm that relies on expensive, non-scalable data annotated by humans or proprietary models. This approach often leads to models that memorize specific ground-truth answers, limiting their generality and ability to generate diverse, creative descriptions. To overcome the limitation of SFT, we propose applying the Reinforcement Learning with Verifiable Rewards (RLVR) paradigm to the open-ended task of image captioning. A primary challenge, however, is designing an objective reward function for the inherently subjective nature of what constitutes a "good" caption. We introduce Captioning Reinforcement Learning (CapRL), a novel training framework that redefines caption quality through its utility: a high-quality caption should enable a non-visual language model to accurately answer questions about the corresponding image. CapRL employs a decoupled two-stage pipeline where an LVLM generates a caption, and the objective reward is derived from the accuracy of a separate, vision-free LLM answering Multiple-Choice Questions based solely on that caption. As the first study to apply RLVR to the subjective image captioning task, we demonstrate that CapRL significantly enhances multiple settings. Pretraining on the CapRL-5M caption dataset annotated by CapRL-3B results in substantial gains across 12 benchmarks. Moreover, within the Prism Framework for caption quality evaluation, CapRL achieves performance comparable to Qwen2.5-VL-72B, while exceeding the baseline by an average margin of 8.4%. Code is available here: https://github.com/InternLM/CapRL.

In-Loop Filtering Using Learned Look-Up Tables for Video Coding

In-loop filtering (ILF) is a key technology in video coding standards to reduce artifacts and enhance visual quality. Recently, neural network-based ILF schemes have achieved remarkable coding gains, emerging as a powerful candidate for next-generation video coding standards. However, the use of deep neural networks (DNN) brings significant computational and time complexity or high demands for dedicated hardware, making it challenging for general use. To address this limitation, we study a practical ILF solution by adopting look-up tables (LUTs). After training a DNN with a restricted reference range for ILF, all possible inputs are traversed, and the output values of the DNN are cached into LUTs. During the coding process, the filtering process is performed by simply retrieving the filtered pixel through locating the input pixels and interpolating between the cached values, instead of relying on heavy inference computations. In this paper, we propose a universal LUT-based ILF framework, termed LUT-ILF++. First, we introduce the cooperation of multiple kinds of filtering LUTs and propose a series of customized indexing mechanisms to enable better filtering reference perception with limited storage consumption. Second, we propose the cross-component indexing mechanism to enable the filtering of different color components jointly. Third, in order to make our solution practical for coding uses, we propose the LUT compaction scheme to enable the LUT pruning, achieving a lower storage cost of the entire solution. The proposed framework is implemented in the VVC reference software. Experimental results show that the proposed framework achieves on average 0.82%/2.97%/1.63% and 0.85%/4.11%/2.06% bitrate reduction for common test sequences, under the AI and RA configurations, respectively. Compared to DNN-based solutions, our proposed solution has much lower time complexity and storage cost.

Scaling Learned Image Compression Models up to 1 Billion

Recent advances in large language models (LLMs) highlight a strong connection between intelligence and compression. Learned image compression, a fundamental task in modern data compression, has made significant progress in recent years. However, current models remain limited in scale, restricting their representation capacity, and how scaling model size influences compression performance remains unexplored. In this work, we present a pioneering study on scaling up learned image compression models and revealing the performance trends through scaling laws. Using the recent state-of-the-art HPCM model as baseline, we scale model parameters from 68.5 millions to 1 billion and fit power-law relations between test loss and key scaling variables, including model size and optimal training compute. The results reveal a scaling trend, enabling extrapolation to larger scale models. Experimental results demonstrate that the scaled-up HPCM-1B model achieves state-of-the-art rate-distortion performance. We hope this work inspires future exploration of large-scale compression models and deeper investigations into the connection between compression and intelligence.

ScaleCap: Inference-Time Scalable Image Captioning via Dual-Modality Debiasing

This paper presents ScaleCap, an inference-time scalable image captioning strategy that generates comprehensive and detailed image captions. The key challenges of high-quality image captioning lie in the inherent biases of LVLMs: multimodal bias resulting in imbalanced descriptive granularity, offering detailed accounts of some elements while merely skimming over others; linguistic bias leading to hallucinated descriptions of non-existent objects. To address these issues, we propose a scalable debiased captioning strategy, which continuously enriches and calibrates the caption with increased inference budget. Specifically, we propose two novel components: heuristic question answering and contrastive sentence rating. The former generates content-specific questions based on the image and answers them to progressively inject relevant information into the caption. The latter employs sentence-level offline contrastive decoding to effectively identify and eliminate hallucinations caused by linguistic biases. With increased inference cost, more heuristic questions are raised by ScaleCap to progressively capture additional visual details, generating captions that are more accurate, balanced, and informative. Extensive modality alignment experiments demonstrate the effectiveness of ScaleCap. Annotating 450K images with ScaleCap and using them for LVLM pretraining leads to consistent performance gains across 11 widely used benchmarks. Furthermore, ScaleCap showcases superb richness and fidelity of generated captions with two additional tasks: replacing images with captions in VQA task, and reconstructing images from captions to assess semantic coverage. Code is available at https://github.com/Cooperx521/ScaleCap.

Alleviating User-Sensitive bias with Fair Generative Sequential Recommendation Model

Recommendation fairness has recently attracted much attention. In the real world, recommendation systems are driven by user behavior, and since users with the same sensitive feature (e.g., gender and age) tend to have the same patterns, recommendation models can easily capture the strong correlation preference of sensitive features and thus cause recommendation unfairness. Diffusion model (DM) as a new generative model paradigm has achieved great success in recommendation systems. DM's ability to model uncertainty and represent diversity, and its modeling mechanism has a high degree of adaptability with the real-world recommendation process with bias. Therefore, we use DM to effectively model the fairness of recommendation and enhance the diversity. This paper proposes a FairGENerative sequential Recommendation model based on DM, FairGENRec. In the training phase, we inject random noise into the original distribution under the guidance of the sensitive feature recognition model, and a sequential denoise model is designed for the reverse reconstruction of items. Simultaneously, recommendation fairness modeling is completed by injecting multi-interests representational information that eliminates the bias of sensitive user features into the generated results. In the inference phase, the model obtains the noise in the form of noise addition by using the history interactions which is followed by reverse iteration to reconstruct the target item representation. Finally, our extensive experiments on three datasets demonstrate the dual enhancement effect of FairGENRec on accuracy and fairness, while the statistical analysis of the cases visualizes the degree of improvement on the fairness of the recommendation.