In this paper we consider an alternative strategy.

As a cost-effective alternative, learning-free PTQ schemes have been proposed. However, the performance is somewhat limited because they cannot consider the inter-layer dependency within the attention module, which is a significant feature of Transformers. In this paper, we thus propose a novel PTQ algorithm that balances accuracy and efficiency. The key idea of the proposed algorithm called aespa is to perform quantization layer-wise for efficiency while targeting attention-wise reconstruction to consider the cross-layer dependency.

Current instance segmentation models achieve high performance on average predictions, but lack principled uncertainty quantification: their outputs are not calibrated, and there is no guarantee that a predicted mask is close to the ground truth. To address this limitation, we introduce a conformal prediction algorithm to generate adaptive confidence sets for instance segmentation. Given an image and a pixel coordinate query, our algorithm generates a confidence set of instance predictions for that pixel, with a provable guarantee for the probability that at least one of the predictions has high Intersection-Over-Union (IoU) with the true object instance mask. We apply our algorithm to instance segmentation examples in agricultural field delineation, cell segmentation, and vehicle detection. Empirically, we find that our prediction sets vary in size based on query difficulty and attain the target coverage, outperforming existing baselines such as Learn Then Test, Conformal Risk Control, and morphological dilation-based methods. We provide versions of the algorithm with asymptotic and finite sample guarantees.

Conformal prediction (CP) offers a principled framework for uncertainty quantification, but it fails to guarantee coverage when faced with missing covariates. In addressing the heterogeneity induced by various missing patterns, Mask-Conditional Valid (MCV) Coverage has emerged as a more desirable property than Marginal Coverage. In this work, we adapt split CP to handle missing values by proposing a preimpute-mask-then-correct framework that can offer valid coverage. We show that our method provides guaranteed Marginal Coverage and Mask-Conditional Validity for general missing data mechanisms. A key component of our approach is a reweighted conformal prediction procedure that corrects the prediction sets after distributional imputation (multiple imputation) of the calibration dataset, making our method compatible with standard imputation pipelines. We derive two algorithms, and we show that they are approximately marginally valid and MCV. We evaluate them on synthetic and real-world datasets. It reduces significantly the width of prediction intervals w.r.t standard MCV methods, while maintaining the target guarantees.

Large language models (LLMs) are increasingly used as evaluators in lieu of humans. While scalable, their judgments are noisy due to imperfect specificity and sensitivity of LLMs, leading to biased accuracy estimates. Although bias-correction methods exist, they are underutilized in LLM research and typically assume exact knowledge of the model's specificity and sensitivity. Furthermore, in general we only have estimates of these values and it is not well known how to properly construct confidence intervals using only estimates. This work presents a simple plug-in framework that corrects such bias and constructs confidence intervals reflecting uncertainty from both test and calibration dataset, enabling practical and statistically sound LLM-based evaluation. Additionally, to reduce uncertainty in the accuracy estimate, we introduce an adaptive algorithm that efficiently allocates calibration sample sizes.

Sparse Mixture-of-Experts (SMoE) architectures have enabled a new frontier in scaling Large Language Models (LLMs), offering superior performance by activating only a fraction of their total parameters during inference. However, their practical deployment is severely hampered by substantial static memory overhead, as all experts must be loaded into memory. Existing post-training pruning methods, while reducing model size, often derive their pruning criteria from a single, general-purpose corpus. This leads to a critical limitation: a catastrophic performance degradation when the pruned model is applied to other domains, necessitating a costly re-pruning for each new domain. To address this generalization gap, we introduce Mosaic Pruning (MoP). The core idea of MoP is to construct a functionally comprehensive set of experts through a structured ``cluster-then-select" process. This process leverages a similarity metric that captures expert performance across different task domains to functionally cluster the experts, and subsequently selects the most representative expert from each cluster based on our proposed Activation Variability Score. Unlike methods that optimize for a single corpus, our proposed Mosaic Pruning ensures that the pruned model retains a functionally complementary set of experts, much like the tiles of a mosaic that together form a complete picture of the original model's capabilities, enabling it to handle diverse downstream tasks.Extensive experiments on various MoE models demonstrate the superiority of our approach. MoP significantly outperforms prior work, achieving a 7.24\% gain on general tasks and 8.92\% on specialized tasks like math reasoning and code generation.

Mixture-of-Experts (MoE) models have shown strong potential in scaling language models efficiently by activating only a small subset of experts per input. However, their widespread deployment remains limited due to the high memory overhead associated with storing all expert parameters, particularly as the number of experts increases. To address this challenge, prior works have explored expert dropping and merging strategies, yet they often suffer from performance drop at high compression ratios. In this paper, we introduce PuzzleMoE, a training-free MoE compression method that achieves both high accuracy and efficient inference through two key innovations: First, PuzzleMoE performs sparse expert merging by identifying element-wise weight redundancy and specialization. It uses a dual-mask to capture both shared and expert-specific parameters. Second, to avoid the overhead of storing binary masks and signs, PuzzleMoE introduces a bit-packed encoding scheme that reuses underutilized exponent bits, enabling efficient MoE inference on GPUs. Extensive experiments demonstrate that PuzzleMoE can compress MoE models by up to 50% while maintaining accuracy across various tasks. Specifically, it outperforms prior MoE compression methods by up to 16.7% on MMLU at 50% compression ratio, and achieves up to 1.28\times inference speedup.

Pruning provides a practical solution to reduce the resources required to run large language models (LLMs) to benefit from their effective capabilities as well as control their cost for training and inference. Research on LLM pruning often ranks the importance of LLM parameters using their magnitudes and calibration-data activations and removes (or masks) the less important ones, accordingly reducing LLMs' size. However, these approaches primarily focus on preserving the LLM's ability to generate fluent sentences, while neglecting performance on specific domains and tasks. In this paper, we propose a simple yet effective pruning approach for LLMs that preserves task-specific capabilities while shrinking their parameter space. We first analyze how conventional pruning minimizes loss perturbation under general-domain calibration and extend this formulation by incorporating task-specific feature distributions into the importance computation of existing pruning algorithms. Thus, our framework computes separate importance scores using both general and task-specific calibration data, partitions parameters into shared and exclusive groups based on activation-norm differences, and then fuses their scores to guide the pruning process. This design enables our method to integrate seamlessly with various foundation pruning techniques and preserve the LLM's specialized abilities under compression. Experiments on widely used benchmarks demonstrate that our approach is effective and consistently outperforms the baselines with identical pruning ratios and different settings.

Structured pruning is a practical approach to deploying large language models (LLMs) efficiently, as it yields compact, hardware-friendly architectures. However, the dominant local paradigm is task-agnostic: by optimizing layer-wise reconstruction rather than task objectives, it tends to preserve perplexity or generic zero-shot behavior but fails to capitalize on modest task-specific calibration signals, often yielding limited downstream gains. We revisit global structured pruning and present GISP-Global Iterative Structured Pruning-a post-training method that removes attention heads and MLP channels using first-order, loss-based important weights aggregated at the structure level with block-wise normalization. An iterative schedule, rather than one-shot pruning, stabilizes accuracy at higher sparsity and mitigates perplexity collapse without requiring intermediate fine-tuning; the pruning trajectory also forms nested subnetworks that support a "prune-once, deploy-many" workflow. Furthermore, because importance is defined by a model-level loss, GISP naturally supports task-specific objectives; we instantiate perplexity for language modeling and a margin-based objective for decision-style tasks. Extensive experiments show that across Llama2-7B/13B, Llama3-8B, and Mistral-0.3-7B, GISP consistently lowers WikiText-2 perplexity and improves downstream accuracy, with especially strong gains at 40-50% sparsity; on DeepSeek-R1-Distill-Llama-3-8B with GSM8K, task-aligned calibration substantially boosts exact-match accuracy.