Searcharxiv⌕ Search

arXiv subjects

Zhikai Li

Publications and source records attributed to Zhikai Li.

At least 19 recordsLinked to original sources
arXiv2026

TailSieve: Partial-Rollout-Guided Tail Routing for LLM Rollouts

Large-scale rollouts have become a core component of modern LLM systems, spanning reinforcement learning (RL) post-training, on-policy distillation (OPD), and sampling-heavy evaluation pipelines. Unlike online serving, which is typically optimized for request-level latency and throughput, a small number of long-tail generations can dominate the end-to-end makespan of an entire rollout step. In practice, rollout requests are often routed uniformly across replicas, which can place extremely long generations inside high-concurrency decoding batches. To address this, we present TailSieve, a partial-rollout-guided framework that jointly controls tail routing and replica allocation for LLM rollouts. In an idealized setting with known completion lengths, we show that makespan-optimal routing in the long-tail regime combines tail isolation with load balancing, and that a simple top-k policy closely approximates this offline optimum. Leveraging the observation that long-tail prompts tend to remain long-tailed across policy updates, TailSieve uses partial rollouts as a training-free signal for identifying candidate tail groups. A hierarchical controller then jointly adapts the number of isolated groups and the replica split between the tail and bulk pools using collected response-work history and a measured concurrency-throughput model. TailSieve achieves up to 1.67x routing-only speedup over uniform group routing. The resulting low-concurrency tail pool further enables route-specialized speculative decoding with MTP or DFlash, achieving up to 2.59x speedup over uniform routing. Selected prompts are regenerated under the current policy, preserving on-policy generation and avoiding additional routing-induced length bias in steady state.

cs.AI↗
arXiv2026

Post-Training Pruning for Diffusion Transformers

Diffusion Transformers (DiTs) have demonstrated impressive performance in image generation but suffer from substantial computational overhead and resource consumption. Post-training pruning offers a promising solution; however, due to DiTs' unique architectural design and parameter distribution, traditional pruning methods are inapplicable, leading to significant performance degradation. Specifically, prior methods developed for LLMs, which derive metrics through a series of approximations, amplify the relative contribution of weights in the saliency metric. In addition, weights in DiTs exhibit significantly larger magnitudes than those in LLMs. Moreover, existing pruning granularity overlooks variations in model structures. In this paper, we propose DiT-Pruning, which improves pruning performance by introducing customized saliency criteria and pruning granularity. We design a novel metric that balances the contributions of weights and activations from an energy-based perspective, enabling more effective identification of important elements. Furthermore, we observe distinct clustering patterns in the two-dimensional weight space. Accordingly, we adopt a clustering-aware pruning granularity, enabling effective sparse allocation. Extensive evaluations on various DiTs show that our method consistently preserves image quality, especially under high sparsity. For FLUX.1-dev at 512x512 resolution on MJHQ, DiT-Pruning achieves only a 0.001 loss in CLIP score at 50% sparsity, dramatically outperforming recent pruning methods.

cs.CV↗
arXiv2026

LEAD: Length-Efficient Adaptive and Dynamic Reasoning for Large Language Models

Large reasoning models, such as OpenAI o1 and DeepSeek-R1, tend to become increasingly verbose as their reasoning capabilities improve. These inflated Chain-of-Thought (CoT) trajectories often exceed what the underlying problems require, wasting compute, latency, and context budgets. While introducing length-based efficiency rewards during reinforcement learning offers a natural remedy, existing methods struggle with two fundamental challenges: the optimal balance between correctness and efficiency is non-stationary throughout training, and intrinsic reasoning budgets vary drastically across problems. Relying on static reward weights and global length constraints inevitably forces a compromise between degraded accuracy and unrealized compression. To overcome these limitations, we propose LEAD (Length-Efficient Adaptive and Dynamic reasoning), a method that replaces static heuristics with online, self-adaptive mechanisms. LEAD dynamically calibrates the correctness-efficiency trade-off at each step using a Potential-Scaled Instability, directing optimization capacity to the most informative learning signal. Furthermore, it estimates an adaptive per-problem target length online based on the model's own correct rollouts, applying a symmetric efficiency reward that penalizes both overthinking and over-compression. Evaluated on five mathematical reasoning benchmarks, LEAD achieves the highest accuracy and Accuracy-Efficiency Score among RL-trained efficient-reasoning methods while producing substantially shorter outputs than the base model.

cs.LG↗
arXiv2026

OSAQ: Outlier Self-Absorption for Accurate Low-bit LLM Quantization

Large Language Models (LLMs) have demonstrated remarkable capabilities. However, their massive parameter scale leads to significant resource consumption and latency during inference. Post-training weight-only quantization offers a promising solution by reducing model size and accelerating token generation through alleviating the memory-bound issue. Nevertheless, the presence of inherent systematic outliers in weights continues to be a major obstacle. While existing methods, such as scaling and rotation, attempt to address this issue, the performance remains unsatisfactory. In this paper, we propose Outlier Self-Absorption Quantization (OSAQ), which performs additive weight suppression guided by the second-order low-rank property for low-bit weight-only quantization of LLMs. Specifically, we observe that the Hessian exhibits low-rank consistency across different inputs, with certain directions consistently showing vanishing curvature. Leveraging this property, we identify a stable null space of the Hessian and then construct an additive weight transformation by linearly combining the vectors within this null space, thereby suppressing weight outliers without affecting the task loss. This additive transformation can be absorbed into the weights offline, requiring no inter-layer transformations and introducing no inference overhead. Moreover, the construction is efficiently achieved by a closed-form solution, without resource-intensive training or iterative procedures. Extensive experiments demonstrate that OSAQ effectively suppresses outliers and enhances low-bit quantization performance. For instance, in 2-bit quantization, OSAQ, when integrated with GPTQ, achieves over 40% lower perplexity compared to vanilla GPTQ.

cs.LG↗
arXiv2026

TS-Verkle: A TypeScript Native Verkle Library With On-chain Verifier

Blockchain systems face significant scalability challenges due to growing data volumes and increasing transaction demands, necessitating more efficient data structures and verification mechanisms. Verkle trees, a novel data structure combining the efficiency of Merkle trees with the compactness of vector commitments, have gained attention for their potential to optimize blockchain storage and improve scalability. However, their practical implementation, especially at the smart contract level, has remained unexplored. To address these challenges, we present TS-verkle, the first known TypeScript-native implementation of Verkle trees designed for web3 backend compatibility, coupled with a corresponding on-chain verifier written in Solidity. Our work bridges this gap by providing a concrete implementation of Verkle trees and demonstrating their feasibility for on-chain verification. While previous literature suggests Verkle trees should outperform Merkle trees due to their succinct proof size, our empirical evaluation reveals that basic implementations of Verkle trees actually incur higher costs than Merkle trees without advanced optimization techniques. This finding represents a crucial insight for blockchain developers and researchers considering Verkle tree adoption. The paper discusses implementation strategies and performance characteristics while exploring implications for scaling and data availability in decentralized blockchain systems.

cs.DC↗
arXiv2026

Arena as Offline Reward: Efficient Fine-Grained Preference Optimization for Diffusion Models

Reinforcement learning from human feedback (RLHF) effectively promotes preference alignment of text-to-image (T2I) diffusion models. To improve computational efficiency, direct preference optimization (DPO), which avoids explicit reward modeling, has been widely studied. However, its reliance on binary feedback limits it to coarse-grained modeling on chosen-rejected pairs, resulting in suboptimal optimization. In this paper, we propose ArenaPO, which leverages Arena scores as offline rewards to provide refined feedback, thus achieving efficient and fine-grained optimization without a reward model. This enables ArenaPO to benefit from both the rich rewards of traditional RLHF and the efficiency of DPO. Specifically, we first construct a model Arena in which each model's capability is represented as a Gaussian distribution, and infer these capabilities by traversing the annotated pairwise preferences. Each output image is treated as a sample from the corresponding capability distribution. Then, for a image pair, conditioned on the two capability distributions and the observed pairwise preference, the absolute quality gap is estimated using latent-variable inference based on truncated normal distribution, which serves as fine-grained feedback during training. It does not require a reward model and can be computed offline, thus introducing no additional training overhead. We conduct ArenaPO training on Pick-a-Pic v2 and HPD v3 datasets, showing that ArenaPO consistently outperforms existing baselines.

cs.CV↗
arXiv2026

QFT: Quantized Full-parameter Tuning of LLMs with Affordable Resources

Large Language Models (LLMs) have showcased remarkable impacts across a wide spectrum of natural language processing tasks. Fine-tuning these pretrained models on downstream datasets provides further significant performance gains; however, this process typically requires a large number of expensive, high-end GPUs. Although there have been efforts focused on parameter-efficient fine-tuning, they cannot fully unlock the powerful potential of full-parameter fine-tuning. In this paper, we propose QFT, a Quantized Full-parameter Tuning framework for LLMs that quantizes and stores all training states, including weights, gradients, and optimizer states, in INT8 format to reduce training memory, thereby enabling full-parameter fine-tuning on existing GPUs at an affordable cost. To ensure training performance, we make two key efforts: i) for quantized gradients and optimizer states, we theoretically prove that the Lion optimizer, with its property of consistent update magnitudes, is highly robust to quantization; ii) and for quantized weights, we employ the hybrid feature quantizer, which identifies and protects a small subset of sparse critical features while quantizing the remaining dense features, thus ensuring accurate weight updates without FP32 backups. Moreover, to support backpropagation in the integer context, we develop a stack-based gradient flow scheme with O(1) complexity, forming a unified integer training pipeline. As a result, QFT reduces the model state memory to 21% of the standard solution while achieving comparable performance, e.g., tuning a LLaMA-7B model requires only <30GB of memory, making it feasible on a single A6000 GPU.

cs.CL↗
arXiv2026

Sparsity Induction for Accurate Post-Training Pruning of Large Language Models

Large language models have demonstrated capabilities in text generation, while their increasing parameter scales present challenges in computational and memory efficiency. Post-training sparsity (PTS), which reduces model cost by removing weights from dense networks, is an effective approach. However, native dense matrices lack high sparsity, making existing approaches that directly remove weights disrupt model states, resulting in unsatisfactory performance recovery even with post-tuning. We propose Sparsity Induction, which promotes models toward higher sparsity at both distribution and feature levels before pruning, to push the limits of PTS. At the distribution level, we enhance distributional sparsity through mathematically equivalent scaling transformations, which are fully absorbable and incur no extra parameters or inference-time overhead. At the feature level, we introduce Spectral Norm Loss to promote feature sparsity from a low-rank perspective. Experiments across diverse model architectures and tasks demonstrate that our method further enhances sparsity-friendliness, achieving superior pruning performance over existing approaches.

cs.CL↗
arXiv2026

Efficient-SAM2: Accelerating SAM2 with Object-Aware Visual Encoding and Memory Retrieval

Segment Anything Model 2 (SAM2) shows excellent performance in video object segmentation tasks; however, the heavy computational burden hinders its application in real-time video processing. Although there have been efforts to improve the efficiency of SAM2, most of them focus on retraining a lightweight backbone, with little exploration into post-training acceleration. In this paper, we observe that SAM2 exhibits sparse perception pattern as biological vision, which provides opportunities for eliminating redundant computation and acceleration: i) In mask decoder, the attention primarily focuses on the foreground objects, whereas the image encoder in the earlier stage exhibits a broad attention span, which results in unnecessary computation to background regions. ii) In memory bank, only a small subset of tokens in each frame contribute significantly to memory attention, and the salient regions exhibit temporal consistency, making full-token computation redundant. With these insights, we propose Efficient-SAM2, which promotes SAM2 to adaptively focus on object regions while eliminating task-irrelevant computations, thereby significantly improving inference efficiency. Specifically, for image encoder, we propose object-aware Sparse Window Routing (SWR), a window-level computation allocation mechanism that leverages the consistency and saliency cues from the previous-frame decoder to route background regions into a lightweight shortcut branch. Moreover, for memory attention, we propose object-aware Sparse Memory Retrieval (SMR), which allows only the salient memory tokens in each frame to participate in computation, with the saliency pattern reused from their first recollection. With negligible additional parameters and minimal training overhead, Efficient-SAM2 delivers 1.68x speedup on SAM2.1-L model with only 1.0% accuracy drop on SA-V test set.

cs.CV↗
arXiv2026

K-Sort Eval: Efficient Preference Evaluation for Visual Generation via Corrected VLM-as-a-Judge

The rapid development of visual generative models raises the need for more scalable and human-aligned evaluation methods. While the crowdsourced Arena platforms offer human preference assessments by collecting human votes, they are costly and time-consuming, inherently limiting their scalability. Leveraging vision-language model (VLMs) as substitutes for manual judgments presents a promising solution. However, the inherent hallucinations and biases of VLMs hinder alignment with human preferences, thus compromising evaluation reliability. Additionally, the static evaluation approach lead to low efficiency. In this paper, we propose K-Sort Eval, a reliable and efficient VLM-based evaluation framework that integrates posterior correction and dynamic matching. Specifically, we curate a high-quality dataset from thousands of human votes in K-Sort Arena, with each instance containing the outputs and rankings of K models. When evaluating a new model, it undergoes (K+1)-wise free-for-all comparisons with existing models, and the VLM provide the rankings. To enhance alignment and reliability, we propose a posterior correction method, which adaptively corrects the posterior probability in Bayesian updating based on the consistency between the VLM prediction and human supervision. Moreover, we propose a dynamic matching strategy, which balances uncertainty and diversity to maximize the expected benefit of each comparison, thus ensuring more efficient evaluation. Extensive experiments show that K-Sort Eval delivers evaluation results consistent with K-Sort Arena, typically requiring fewer than 90 model runs, demonstrating both its efficiency and reliability.

cs.CV↗
arXiv2026

PTQ4ARVG: Post-Training Quantization for AutoRegressive Visual Generation Models

AutoRegressive Visual Generation (ARVG) models retain an architecture compatible with language models, while achieving performance comparable to diffusion-based models. Quantization is commonly employed in neural networks to reduce model size and computational latency. However, applying quantization to ARVG remains largely underexplored, and existing quantization methods fail to generalize effectively to ARVG models. In this paper, we explore this issue and identify three key challenges: (1) severe outliers at channel-wise level, (2) highly dynamic activations at token-wise level, and (3) mismatched distribution information at sample-wise level. To these ends, we propose PTQ4ARVG, a training-free post-training quantization (PTQ) framework consisting of: (1) Gain-Projected Scaling (GPS) mitigates the channel-wise outliers, which expands the quantization loss via a Taylor series to quantify the gain of scaling for activation-weight quantization, and derives the optimal scaling factor through differentiation.(2) Static Token-Wise Quantization (STWQ) leverages the inherent properties of ARVG, fixed token length and position-invariant distribution across samples, to address token-wise variance without incurring dynamic calibration overhead.(3) Distribution-Guided Calibration (DGC) selects samples that contribute most to distributional entropy, eliminating the sample-wise distribution mismatch. Extensive experiments show that PTQ4ARVG can effectively quantize the ARVG family models to 8-bit and 6-bit while maintaining competitive performance. Code is available at http://github.com/BienLuky/PTQ4ARVG .

cs.CV↗
arXiv2025

Validating the 1D-3D coupling stellar models via Asteroseismology of 18 Kepler main-sequence stars

Standard 1D stellar evolution model has poor descriptions of the near-surface layers of stars, and this can be improved by using the atmosphere model computed from 3D hydrodynamical simulations. In this work, we validated the model inferences of the 1D-3D coupling models using 18 well-studied stars from the Kepler LEGACY Sample. We compared our estimates of the fundamental parameters determined with other six pipelines and obtained good consistency. The results indicate that the 1D-3D coupling models can be applied to characterizing solar-like stars with confidence. Our analysis showed similar pattern for the surface term in stars with effective temperature range from ~5000 K to ~6400 K, suggesting that the surface term of the 1D-3D coupling models is simpler and easier to deal with than that of models using classical atmosphere.

astro-ph.SR↗
arXiv2025

Rectified SpaAttn: Revisiting Attention Sparsity for Efficient Video Generation

Diffusion Transformers dominate video generation, but the quadratic complexity of attention computation introduces substantial latency. Attention sparsity reduces computational costs by focusing on critical tokens while ignoring non-critical tokens. However, existing methods suffer from severe performance degradation. In this paper, we revisit attention sparsity and reveal that existing methods induce systematic biases in attention allocation: (1) excessive focus on critical tokens amplifies their attention weights; (2) complete neglect of non-critical tokens causes the loss of relevant attention weights. To address these issues, we propose Rectified SpaAttn, which rectifies attention allocation with implicit full attention reference, thereby enhancing the alignment between sparse and full attention maps. Specifically: (1) for critical tokens, we show that their bias is proportional to the sparse attention weights, with the ratio governed by the amplified weights. Accordingly, we propose Isolated-Pooling Attention Reallocation, which calculates accurate rectification factors by reallocating multimodal pooled weights. (2) for non-critical tokens, recovering attention weights from the pooled query-key yields attention gains but also introduces pooling errors. Therefore, we propose Gain-Aware Pooling Rectification, which ensures that the rectified gain consistently surpasses the induced error. Moreover, we customize and integrate the Rectified SpaAttn kernel using Triton, achieving up to 3.33 and 2.08 times speedups on HunyuanVideo and Wan 2.1, respectively, while maintaining high generation quality. We release Rectified SpaAttn as open-source at https://github.com/BienLuky/Rectified-SpaAttn .

cs.CV↗
arXiv2025

SAQ-SAM: Semantically-Aligned Quantization for Segment Anything Model

Segment Anything Model (SAM) exhibits remarkable zero-shot segmentation capability; however, its prohibitive computational costs make edge deployment challenging. Although post-training quantization (PTQ) offers a promising compression solution, existing methods yield unsatisfactory results when applied to SAM, owing to its specialized model components and promptable workflow: (i) The mask decoder's attention exhibits extreme activation outliers, and we find that aggressive clipping (even 100x), without smoothing or isolation, is effective in suppressing outliers while maintaining performance. Unfortunately, traditional distribution-based metrics (e.g., MSE) fail to provide such large-scale clipping. (ii) Existing quantization reconstruction methods neglect semantic interactivity of SAM, leading to misalignment between image feature and prompt intention. To address the above issues, we propose SAQ-SAM in this paper, which boosts PTQ for SAM from the perspective of semantic alignment. Specifically, we propose Perceptual-Consistency Clipping, which exploits attention focus overlap to promote aggressive clipping while preserving semantic capabilities. Furthermore, we propose Prompt-Aware Reconstruction, which incorporates image-prompt interactions by leveraging cross-attention in mask decoder, thus facilitating alignment in both distribution and semantic. Moreover, to ensure the interaction efficiency, we design a layer-skipping strategy for image tokens in encoder. Extensive experiments are conducted on various SAM sizes and tasks, including instance segmentation, oriented object detection, and semantic segmentation, and the results show that our method consistently exhibits advantages. For example, when quantizing SAM-B to 4-bit, SAQ-SAM achieves 11.7% higher mAP than the baseline in instance segmentation task.

cs.CV↗
arXiv2025

DilateQuant: Accurate and Efficient Diffusion Quantization via Weight Dilation

Model quantization is a promising method for accelerating and compressing diffusion models. Nevertheless, since post-training quantization (PTQ) fails catastrophically at low-bit cases, quantization-aware training (QAT) is essential. Unfortunately, the wide range and time-varying activations in diffusion models sharply increase the complexity of quantization, making existing QAT methods inefficient. Equivalent scaling can effectively reduce activation range, but previous methods remain the overall quantization error unchanged. More critically, these methods significantly disrupt the original weight distribution, resulting in poor weight initialization and challenging convergence during QAT training. In this paper, we propose a novel QAT framework for diffusion models, called DilateQuant. Specifically, we propose Weight Dilation (WD) that maximally dilates the unsaturated in-channel weights to a constrained range through equivalent scaling. WD decreases the activation range while preserving the original weight range, which steadily reduces the quantization error and ensures model convergence. To further enhance accuracy and efficiency, we design a Temporal Parallel Quantizer (TPQ) to address the time-varying activations and introduce a Block-wise Knowledge Distillation (BKD) to reduce resource consumption in training. Extensive experiments demonstrate that DilateQuant significantly outperforms existing methods in terms of accuracy and efficiency. Code is available at http://github.com/BienLuky/DilateQuant .

cs.CV↗
arXiv2025

EDA-DM: Enhanced Distribution Alignment for Post-Training Quantization of Diffusion Models

Diffusion models have achieved great success in image generation tasks. However, the lengthy denoising process and complex neural networks hinder their low-latency applications in real-world scenarios. Quantization can effectively reduce model complexity, and post-training quantization (PTQ), which does not require fine-tuning, is highly promising for compressing and accelerating diffusion models. Unfortunately, we find that due to the highly dynamic activations, existing PTQ methods suffer from distribution mismatch issues at both calibration sample level and reconstruction output level, which makes the performance far from satisfactory. In this paper, we propose EDA-DM, a standardized PTQ method that efficiently addresses the above issues. Specifically, at the calibration sample level, we extract information from the density and diversity of latent space feature maps, which guides the selection of calibration samples to align with the overall sample distribution; and at the reconstruction output level, we theoretically analyze the reasons for previous reconstruction failures and, based on this insight, optimize block reconstruction using the Hessian loss of layers, aligning the outputs of quantized model and full-precision model at different network granularity. Extensive experiments demonstrate that EDA-DM significantly outperforms the existing PTQ methods across various models and datasets. Our method achieves a 1.83 times speedup and 4 times compression for the popular Stable-Diffusion on MS-COCO, with only a 0.05 loss in CLIP score. Code is available at http://github.com/BienLuky/EDA-DM .

cs.CV↗
arXiv2025

Unsupervised Learning for Class Distribution Mismatch

Class distribution mismatch (CDM) refers to the discrepancy between class distributions in training data and target tasks. Previous methods address this by designing classifiers to categorize classes known during training, while grouping unknown or new classes into an "other" category. However, they focus on semi-supervised scenarios and heavily rely on labeled data, limiting their applicability and performance. To address this, we propose Unsupervised Learning for Class Distribution Mismatch (UCDM), which constructs positive-negative pairs from unlabeled data for classifier training. Our approach randomly samples images and uses a diffusion model to add or erase semantic classes, synthesizing diverse training pairs. Additionally, we introduce a confidence-based labeling mechanism that iteratively assigns pseudo-labels to valuable real-world data and incorporates them into the training process. Extensive experiments on three datasets demonstrate UCDM's superiority over previous semi-supervised methods. Specifically, with a 60% mismatch proportion on Tiny-ImageNet dataset, our approach, without relying on labeled data, surpasses OpenMatch (with 40 labels per class) by 35.1%, 63.7%, and 72.5% in classifying known, unknown, and new classes.

cs.CV↗
arXiv2025

K-Sort Arena: Efficient and Reliable Benchmarking for Generative Models via K-wise Human Preferences

The rapid advancement of visual generative models necessitates efficient and reliable evaluation methods. Arena platform, which gathers user votes on model comparisons, can rank models with human preferences. However, traditional Arena methods, while established, require an excessive number of comparisons for ranking to converge and are vulnerable to preference noise in voting, suggesting the need for better approaches tailored to contemporary evaluation challenges. In this paper, we introduce K-Sort Arena, an efficient and reliable platform based on a key insight: images and videos possess higher perceptual intuitiveness than texts, enabling rapid evaluation of multiple samples simultaneously. Consequently, K-Sort Arena employs K-wise comparisons, allowing K models to engage in free-for-all competitions, which yield much richer information than pairwise comparisons. To enhance the robustness of the system, we leverage probabilistic modeling and Bayesian updating techniques. We propose an exploration-exploitation-based matchmaking strategy to facilitate more informative comparisons. In our experiments, K-Sort Arena exhibits 16.3x faster convergence compared to the widely used ELO algorithm. To further validate the superiority and obtain a comprehensive leaderboard, we collect human feedback via crowdsourced evaluations of numerous cutting-edge text-to-image and text-to-video models. Thanks to its high efficiency, K-Sort Arena can continuously incorporate emerging models and update the leaderboard with minimal votes. Our project has undergone several months of internal testing and is now available at https://huggingface.co/spaces/ksort/K-Sort-Arena

cs.AI↗