Does Compression Preserve Uncertainty? A Unified Benchmark for Quantized and Sparse LLMs via Conformal Prediction

Model compression techniques such as quantization and pruning are widely used to reduce the deployment cost of large language models (LLMs), with existing evaluations focusing almost exclusively on accuracy preservation. However, in safety-critical applications, a model's ability to reliably quantify its own uncertainty is equally important. We ask: does compression preserve this ability? To answer this question, we benchmark 12 LLMs under various compression configurations across five NLP tasks, using conformal prediction to provide a rigorous, distribution-free measure of uncertainty. Our experiments reveal that: (I) compression frequently decouples accuracy from uncertainty; (II) larger models absorb compression-induced uncertainty far more effectively than smaller ones; and (III) uncertainty inflation is often threshold-like rather than gradual. These results suggest that accuracy-only evaluation is insufficient for assessing the deployment readiness of compressed LLMs, and that uncertainty-aware benchmarking should be a standard component of model compression pipelines.

Causal Disentanglement for Full-Reference Image Quality Assessment

Existing deep network-based full-reference image quality assessment (FR-IQA) models typically work by performing pairwise comparisons of deep features from the reference and distorted images. In this paper, we approach this problem from a different perspective and propose a novel FR-IQA paradigm based on causal inference and decoupled representation learning. Unlike typical feature comparison-based FR-IQA models, our approach formulates degradation estimation as a causal disentanglement process guided by intervention on latent representations. We first decouple degradation and content representations by exploiting the content invariance between the reference and distorted images. Second, inspired by the human visual masking effect, we design a masking module to model the causal relationship between image content and degradation features, thereby extracting content-influenced degradation features from distorted images. Finally, quality scores are predicted from these degradation features using either supervised regression or label-free dimensionality reduction. Extensive experiments demonstrate that our method achieves highly competitive performance on standard IQA benchmarks across fully supervised, few-label, and label-free settings. Furthermore, we evaluate the approach on diverse non-standard natural image domains with scarce data, including underwater, radiographic, medical, neutron, and screen-content images. Benefiting from its ability to perform scenario-specific training and prediction without labeled IQA data, our method exhibits superior cross-domain generalization compared to existing training-free FR-IQA models.

AgentSentry: Mitigating Indirect Prompt Injection in LLM Agents via Temporal Causal Diagnostics and Context Purification

Large language model (LLM) agents increasingly rely on external tools and retrieval systems to autonomously complete complex tasks. However, this design exposes agents to indirect prompt injection (IPI), where attacker-controlled context embedded in tool outputs or retrieved content silently steers agent actions away from user intent. Unlike prompt-based attacks, IPI unfolds over multi-turn trajectories, making malicious control difficult to disentangle from legitimate task execution. Existing inference-time defenses primarily rely on heuristic detection and conservative blocking of high-risk actions, which can prematurely terminate workflows or broadly suppress tool usage under ambiguous multi-turn scenarios. We propose AgentSentry, a novel inference-time detection and mitigation framework for tool-augmented LLM agents. To the best of our knowledge, AgentSentry is the first inference-time defense to model multi-turn IPI as a temporal causal takeover. It localizes takeover points via controlled counterfactual re-executions at tool-return boundaries and enables safe continuation through causally guided context purification that removes attack-induced deviations while preserving task-relevant evidence. We evaluate AgentSentry on the \textsc{AgentDojo} benchmark across four task suites, three IPI attack families, and multiple black-box LLMs. AgentSentry eliminates successful attacks and maintains strong utility under attack, achieving an average Utility Under Attack (UA) of 74.55 %, improving UA by 20.8 to 33.6 percentage points over the strongest baselines without degrading benign performance.

PuriLight: A Lightweight Shuffle and Purification Framework for Monocular Depth Estimation

We propose PuriLight, a lightweight and efficient framework for self-supervised monocular depth estimation, to address the dual challenges of computational efficiency and detail preservation. While recent advances in self-supervised depth estimation have reduced reliance on ground truth supervision, existing approaches remain constrained by either bulky architectures compromising practicality or lightweight models sacrificing structural precision. These dual limitations underscore the critical need to develop lightweight yet structurally precise architectures. Our framework addresses these limitations through a three-stage architecture incorporating three novel modules: the Shuffle-Dilation Convolution (SDC) module for local feature extraction, the Rotation-Adaptive Kernel Attention (RAKA) module for hierarchical feature enhancement, and the Deep Frequency Signal Purification (DFSP) module for global feature purification. Through effective collaboration, these modules enable PuriLight to achieve both lightweight and accurate feature extraction and processing. Extensive experiments demonstrate that PuriLight achieves state-of-the-art performance with minimal training parameters while maintaining exceptional computational efficiency. Codes will be available at https://github.com/ishrouder/PuriLight.

Prune, Don't Rebuild: Efficiently Tuning $α$-Reachable Graphs for Nearest Neighbor Search

Vector similarity search is an essential primitive in modern AI and ML applications. Most vector databases adopt graph-based approximate nearest neighbor (ANN) search algorithms, such as DiskANN (Subramanya et al., 2019), which have demonstrated state-of-the-art empirical performance. DiskANN's graph construction is governed by a reachability parameter $α$, which gives a trade-off between construction time, query time, and accuracy. However, adaptively tuning this trade-off typically requires rebuilding the index for different $α$ values, which is prohibitive at scale. In this work, we propose RP-Tuning, an efficient post-hoc routine, based on DiskANN's pruning step, to adjust the $α$ parameter without reconstructing the full index. Within the $α$-reachability framework of prior theoretical works (Indyk and Xu, 2023; Gollapudi et al., 2025), we prove that pruning an initially $α$-reachable graph with RP-Tuning preserves worst-case reachability guarantees in general metrics and improved guarantees in Euclidean metrics. Empirically, we show that RP-Tuning accelerates DiskANN tuning on four public datasets by up to $43\times$ with negligible overhead.

Still Manual? Automated Linter Configuration via DSL-Based LLM Compilation of Coding Standards

Coding standards are essential for maintaining consistent and high-quality code across teams and projects. Linters help developers enforce these standards by detecting code violations. However, manual linter configuration is complex and expertise-intensive, and the diversity and evolution of programming languages, coding standards, and linters lead to repetitive and maintenance-intensive configuration work. To reduce manual effort, we propose LintCFG, a domain-specific language (DSL)-driven, LLM-based compilation approach to automate linter configuration generation for coding standards, independent of programming languages, coding standards, and linters. Inspired by compiler design, we first design a DSL to express coding rules in a tool-agnostic, structured, readable, and precise manner. Then, we build linter configurations into DSL configuration instructions. For a given natural language coding standard, the compilation process parses it into DSL coding standards, matches them with the DSL configuration instructions to set configuration names, option names and values, verifies consistency between the standards and configurations, and finally generates linter-specific configurations. Experiments with Checkstyle for Java coding standard show that our approach achieves over 90% precision and recall in DSL representation, with accuracy, precision, recall, and F1-scores close to 70% (with some exceeding 70%) in fine-grained linter configuration generation. Notably, our approach outperforms baselines by over 100% in precision. A user study further shows that our approach improves developers' efficiency in configuring linters for coding standards. Finally, we demonstrate the generality of the approach by generating ESLint configurations for JavaScript coding standards, showcasing its broad applicability across other programming languages, coding standards, and linters.

Forget by Uncertainty: Orthogonal Entropy Unlearning for Quantized Neural Networks

The deployment of quantized neural networks on edge devices, combined with privacy regulations like GDPR, creates an urgent need for machine unlearning in quantized models. However, existing methods face critical challenges: they induce forgetting by training models to memorize incorrect labels, conflating forgetting with misremembering, and employ scalar gradient reweighting that cannot resolve directional conflicts between gradients. We propose OEU, a novel Orthogonal Entropy Unlearning framework with two key innovations: 1) Entropy-guided unlearning maximizes prediction uncertainty on forgotten data, achieving genuine forgetting rather than confident misprediction, and 2) Gradient orthogonal projection eliminates interference by projecting forgetting gradients onto the orthogonal complement of retain gradients, providing theoretical guarantees for utility preservation under first-order approximation. Extensive experiments demonstrate that OEU outperforms existing methods in both forgetting effectiveness and retain accuracy.

SAGE: Accelerating Vision-Language Models via Entropy-Guided Adaptive Speculative Decoding

Speculative decoding has emerged as a promising approach to accelerate inference in vision-language models (VLMs) by enabling parallel verification of multiple draft tokens. However, existing methods rely on static tree structures that remain fixed throughout the decoding process, failing to adapt to the varying prediction difficulty across generation steps. This leads to suboptimal acceptance lengths and limited speedup. In this paper, we propose SAGE, a novel framework that dynamically adjusts the speculation tree structure based on real-time prediction uncertainty. Our key insight is that output entropy serves as a natural confidence indicator with strong temporal correlation across decoding steps. SAGE constructs deeper-narrower trees for high-confidence predictions to maximize speculation depth, and shallower-wider trees for uncertain predictions to diversify exploration. SAGE improves acceptance lengths and achieves faster acceleration compared to static tree baselines. Experiments on multiple benchmarks demonstrate the effectiveness of SAGE: without any loss in output quality, it delivers up to $3.36\times$ decoding speedup for LLaVA-OneVision-72B and $3.18\times$ for Qwen2.5-VL-72B.

AscendCraft: Automatic Ascend NPU Kernel Generation via DSL-Guided Transcompilation

The performance of deep learning models critically depends on efficient kernel implementations, yet developing high-performance kernels for specialized accelerators remains time-consuming and expertise-intensive. While recent work demonstrates that large language models (LLMs) can generate correct and performant GPU kernels, kernel generation for neural processing units (NPUs) remains largely underexplored due to domain-specific programming models, limited public examples, and sparse documentation. Consequently, directly generating AscendC kernels with LLMs yields extremely low correctness, highlighting a substantial gap between GPU and NPU kernel generation. We present AscendCraft, a DSL-guided approach for automatic AscendC kernel generation. AscendCraft introduces a lightweight DSL that abstracts non-essential complexity while explicitly modeling Ascend-specific execution semantics. Kernels are first generated in the DSL using category-specific expert examples and then transcompiled into AscendC through structured, constraint-driven LLM lowering passes. Evaluated on MultiKernelBench across seven operator categories, AscendCraft achieves 98.1% compilation success and 90.4% functional correctness. Moreover, 46.2% of generated kernels match or exceed PyTorch eager execution performance, demonstrating that DSL-guided transcompilation can enable LLMs to generate both correct and competitive NPU kernels. Beyond benchmarks, AscendCraft further demonstrates its generality by successfully generating two correct kernels for newly proposed mHC architecture, achieving performance that substantially surpasses PyTorch eager execution.

Bits for Privacy: Evaluating Post-Training Quantization via Membership Inference

Deep neural networks are widely deployed with quantization techniques to reduce memory and computational costs by lowering the numerical precision of their parameters. While quantization alters model parameters and their outputs, existing privacy analyses primarily focus on full-precision models, leaving a gap in understanding how bit-width reduction can affect privacy leakage. We present the first systematic study of the privacy-utility relationship in post-training quantization (PTQ), a versatile family of methods that can be applied to pretrained models without further training. Using membership inference attacks as our evaluation framework, we analyze three popular PTQ algorithms-AdaRound, BRECQ, and OBC-across multiple precision levels (4-bit, 2-bit, and 1.58-bit) on CIFAR-10, CIFAR-100, and TinyImageNet datasets. Our findings consistently show that low-precision PTQs can reduce privacy leakage. In particular, lower-precision models demonstrate up to an order of magnitude reduction in membership inference vulnerability compared to their full-precision counterparts, albeit at the cost of decreased utility. Additional ablation studies on the 1.58-bit quantization level show that quantizing only the last layer at higher precision enables fine-grained control over the privacy-utility trade-off. These results offer actionable insights for practitioners to balance efficiency, utility, and privacy protection in real-world deployments.