搜索结果：Rebuild

Containerization has revolutionized software deployment, with Docker leading the way due to its ease of use and consistent runtime environment. As Docker usage grows, optimizing Dockerfile performance, particularly by reducing rebuild time, has become essential for maintaining efficient CI/CD pipelines. However, existing optimization approaches primarily address single builds without considering the recurring rebuild costs associated with modifications and evolution, limiting long-term efficiency gains. To bridge this gap, we present Doctor, a method for improving Dockerfile build efficiency through instruction re-ordering that addresses key challenges: identifying instruction dependencies, predicting future modifications, ensuring behavioral equivalence, and managing the optimization computational complexity. We developed a comprehensive dependency taxonomy based on Dockerfile syntax and a historical modification analysis to prioritize frequently modified instructions. Using a weighted topological sorting algorithm, Doctor optimizes instruction order to minimize future rebuild time while maintaining functionality. Experiments on 2,000 GitHub repositories show that Doctor improves

This paper presents MiniGPT, a compact from-scratch implementation of GPT-style autoregressive language modeling in PyTorch. The aim is to rebuild the core GPT pipeline from first principles after studying the design of nanoGPT by Andrej Karpathy, while keeping the model and training code independently written in a single notebook. MiniGPT implements token and positional embeddings, causal multi-head self-attention, pre-LayerNorm Transformer blocks, residual connections, feed-forward MLP layers, next-token cross-entropy training (teacher forcing), validation tracking, checkpoint selection, and autoregressive text generation. This paper evaluates the implementation on Tiny Shakespeare dataset using character-level tokenization. A baseline 0.83M-parameter model reaches a validation loss of 1.7236 after 3000 training iterations. A stronger 10.77M-parameter configuration, using a larger context length and improved training settings, reaches a best validation loss of 1.4780 and generates text with recognizable Shakespeare-style dialogue structure. MiniGPT does not introduce a new language-model architecture. Instead, it documents a clear and reproducible implementation path from raw tex

We revisit zigzag array codes, a family of MDS codes known for achieving optimal access and optimal rebuilding ratio in single-node repair. In this work, we endow zigzag codes with two new properties: small field size and low skip cost. First, we prove that when the row-indexing group is $\mathcal{G} = \mathbb{Z}_2^m$ and the field has characteristic two, explicit coefficients over any field with $|\mathcal{F}|\ge N$ guarantee the MDS property, thereby decoupling the dependence among $p$, $k$, and $M$. Second, we introduce an ordering-and-subgroup framework that yields repair-by-transfer schemes with bounded skip cost and low repair-fragmentation ratio (RFR), while preserving optimal access and optimal rebuilding ratio. Our explicit constructions include families with zero skip cost whose rates approach $2/3$, and families with bounded skip cost whose rates approach $3/4$ and $4/5$. These rates are comparable to those of MDS array codes widely deployed in practice. Together, these results demonstrate that zigzag codes can be made both more flexible in theory and more practical for modern distributed storage systems.

Turning ideas into full software projects from scratch has become a popular use case for language models. Agents are being deployed to seed, maintain, and grow codebases over extended periods with minimal human oversight. Such settings require models to make high-level software architecture decisions. However, existing benchmarks measure focused, limited tasks such as fixing a single bug or developing a single, specified feature. We therefore introduce ProgramBench to measure the ability of software engineering agents to develop software holisitically. In ProgramBench, given only a program and its documentation, agents must architect and implement a codebase that matches the reference executable's behavior. End-to-end behavioral tests are generated via agent-driven fuzzing, enabling evaluation without prescribing implementation structure. Our 200 tasks range from compact CLI tools to widely used software such as FFmpeg, SQLite, and the PHP interpreter. We evaluate 9 LMs and find that none fully resolve any task, with the best model passing 95\% of tests on only 3\% of tasks. Models favor monolithic, single-file implementations that diverge sharply from human-written code.

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.

Pre-trained code models rely heavily on high-quality pre-training data, particularly human-written reference comments that bridge code and natural language. However, these comments often become outdated as software evolves, degrading model performance. Large language models (LLMs) excel at generating high-quality code comments. We investigate whether replacing human-written comments with LLM-generated ones improves pre-training datasets. Since standard metrics cannot assess reference comment quality, we propose two novel reference-free evaluation tasks: code-comment inconsistency detection and semantic code search. Results show that LLM-generated comments are more semantically consistent with code than human-written ones, as confirmed by manual evaluation. Leveraging this finding, we rebuild the CodeSearchNet dataset with LLM-generated comments and re-pre-train CodeT5. Evaluations demonstrate that models trained on LLM-enhanced data outperform those using original human comments in code summarization, generation, and translation tasks. This work validates rebuilding pre-training datasets with LLMs to advance code intelligence, challenging the traditional reliance on human reference

Jailbreak attacks pose persistent threats to large language models (LLMs). Current safety alignment methods have attempted to address these issues, but they experience two significant limitations: insufficient safety alignment depth and unrobust internal defense mechanisms. These limitations make them vulnerable to adversarial attacks such as prefilling and refusal direction manipulation. We introduce DeepRefusal, a robust safety alignment framework that overcomes these issues. DeepRefusal forces the model to dynamically rebuild its refusal mechanisms from jailbreak states. This is achieved by probabilistically ablating the refusal direction across layers and token depths during fine-tuning. Our method not only defends against prefilling and refusal direction attacks but also demonstrates strong resilience against other unseen jailbreak strategies. Extensive evaluations on four open-source LLM families and six representative attacks show that DeepRefusal reduces attack success rates by approximately 95%, while maintaining model capabilities with minimal performance degradation.

This paper proposes smaRTLy: a new optimization technique for multiplexers in Register-Transfer Level (RTL) logic synthesis. Multiplexer trees are very common in RTL designs, and traditional tools like Yosys optimize them by traversing the tree and monitoring control port values. However, this method does not fully exploit the intrinsic logical relationships among signals or the potential for structural optimization. To address these limitations, we develop innovative strategies to remove redundant multiplexer trees and restructure the remaining ones, significantly reducing the overall gate count. We evaluate smaRTLy on the IWLS-2005 and RISC-V benchmarks, achieving an additional 8.95% reduction in AIG area compared to Yosys. We also evaluate smaRTLy on an industrial benchmark in the scale of millions of gates, results show that smaRTLy can remove 47.2% more AIG area than Yosys. These results demonstrate the effectiveness of our logic inferencing and structural rebuilding techniques in enhancing the RTL optimization process, leading to more efficient hardware designs.

A full power flow (PF) model is a complete representation of the physical power network. Traditional model-based methods rely on the full PF model to implement power flow analysis. In practice, however, some PF model parameters can be inaccurate or even unavailable due to the uncertainties or dynamics in the power systems. Moreover, because the power network keeps evolving with possibly changing topology, the generalizability of a PF model to different network sizes and typologies should be considered. In this paper, we propose a PF rebuild model based on graph attention networks (GAT) by constructing a new graph based on the real and imaginary parts of voltage at each bus. By comparing with two state-of-the-art PF rebuild models for different standard IEEE power system cases and their modified topology variants, we demonstrate the feasibility of our method. Experimental results show that our proposed model achieves better accuracy for a changing network and can generalize to different networks with less accuracy discount.

The Santa cruz Extreme Adaptive optics Lab (SEAL) is a visible/near-infrared wavelength testbed designed to support technology development for high contrast imaging on large, segmented, ground-based telescopes. SEAL saw first light in 2021 as a transmissive, visible-wavelength AO testbed. In this paper, we present four major upgrades to SEAL: (1) the testbed has been rebuilt with custom off-axis parabolic mirrors, enabling operation in both near-infrared and visible wavelengths; (2) the suite of wavefront sensors now includes a Shack-Hartmann, transmissive four-sided pyramid, vector-Zernike, and, in the muirSEAL testbed, a photonic lantern; (3) the testbed includes a vector-vortex coronagraph and will soon include a hybrid astrophotonic coronagraph; (4) in addition to its original Keck-heritage RTC, SEAL now includes two additional control software packages: Catkit, originally developed for the HiCAT testbed at the Space Telescope Science Institute, and the RTC Compute And Control for Adaptive Optics (CACAO), originally designed for Subaru/SCExAO. We discuss the performance of the testbed after the reflective rebuild and on-going technology development work at SEAL.

Modern software projects depend on many third-party libraries, complicating reproducible and secure builds. Several package managers address this with the generation of a lockfile that freezes dependency versions and can be used to verify the integrity of dependencies. Yet, Maven, one of the most important package managers in the Java ecosystem, lacks native support for a lockfile. We present Maven-Lockfile to generate and update lockfiles, with support for rebuilding projects from past versions. Our lockfiles capture all direct and transitive dependencies with their checksums, enabling high integrity builds. Our evaluation shows that Maven-Lockfile can reproduce builds from historical commits and is able to detect tampered artifacts. With minimal configuration, Maven-Lockfile equips Java projects with modern build integrity and build reproducibility, and fosters future research on software supply chain security in Java.

In post-disaster contexts, design is not only about rebuilding structures but also about reimagining how architecture can become a communicative medium that supports recovery, resilience, and collective memory. While recent studies have expanded the understanding of media architecture from aesthetic urban screens to participatory civic infrastructures, there remains limited empirical research on its potential role in post-disaster contexts. In particular, opportunities exist to explore how architecture and interaction design might speculate on media architecture's role in rebuilding and recovery efforts for post-disaster permanent housing, especially when conceptualizing disasters as active agents that reshape design processes. Following to Kahramanmaras earthquake on February 6, 2023, we conducted two focus groups with architects and interaction designers in the case of Antakya, Turkey, building on affected residents' expectations for post-earthquake permanent housing. Our analysis revealed three critical dimensions of how future media architecture may support post-disaster housing: (1) as a facilitator of individuals' social connections to their community, (2) as an enabler of mu

Large language models (LLMs) have seen substantial growth, necessitating efficient model pruning techniques. Existing post-training pruning methods primarily measure weight importance in converged dense models, often overlooking changes in weight significance during the pruning process, leading to performance degradation. To address this issue, we present LLM-Barber (Block-Aware Rebuilder for Sparsity Mask in One-Shot), a novel one-shot pruning framework that rebuilds the sparsity mask of pruned models without any retraining or weight reconstruction. LLM-Barber incorporates block-aware error optimization across Self-Attention and MLP blocks, facilitating global performance optimization. We are the first to employ the product of weights and gradients as a pruning metric in the context of LLM post-training pruning. This enables accurate identification of weight importance in massive models and significantly reduces computational complexity compared to methods using secondorder information. Our experiments show that LLM-Barber efficiently prunes models from LLaMA and OPT families (7B to 13B) on a single A100 GPU in just 30 minutes, achieving state-of-the-art results in both perplexity

MDS (maximum distance separable) array codes are widely used in storage systems due to their computationally efficient encoding and decoding procedures. An MDS code with r redundancy nodes can correct any r erasures by accessing (reading) all the remaining information in both the systematic nodes and the parity (redundancy) nodes. However, in practice, a single erasure is the most likely failure event; hence, a natural question is how much information do we need to access in order to rebuild a single storage node? We define the rebuilding ratio as the fraction of remaining information accessed during the rebuilding of a single erasure. In our previous work we showed that the optimal rebuilding ratio of 1/r is achievable (using our newly constructed array codes) for the rebuilding of any systematic node, however, all the information needs to be accessed for the rebuilding of the parity nodes. Namely, constructing array codes with a rebuilding ratio of 1/r was left as an open problem. In this paper, we solve this open problem and present array codes that achieve the lower bound of 1/r for rebuilding any single systematic or parity node.

MDS array codes are widely used in storage systems due to their computationally efficient encoding and decoding procedures. An MDS code with $r$ redundancy nodes can correct any $r$ node erasures by accessing all the remaining information in the surviving nodes. However, in practice, $e$ erasures is a more likely failure event, for $1\le e<r$. Hence, a natural question is how much information do we need to access in order to rebuild $e$ storage nodes? We define the rebuilding ratio as the fraction of remaining information accessed during the rebuilding of $e$ erasures. In our previous work we constructed MDS codes, called zigzag codes, that achieve the optimal rebuilding ratio of $1/r$ for the rebuilding of any systematic node when $e=1$, however, all the information needs to be accessed for the rebuilding of the parity node erasure. The (normalized) repair bandwidth is defined as the fraction of information transmitted from the remaining nodes during the rebuilding process. For codes that are not necessarily MDS, Dimakis et al. proposed the regenerating codes framework where any $r$ erasures can be corrected by accessing some of the remaining information, and any $e=1$ erasure

We present an axiomatic approach to combination theorems for various homological properties of groups and, more generally, of chain complexes. Examples of such properties include algebraic finiteness properties, $\ell^2$-invisibility, $\ell^2$-acyclicity, lower bounds for Novikov--Shubin invariants, and vanishing of homology growth. As a key example, we introduce an algebraic version of Abért--Bergeron--Frączyk--Gaboriau's cheap rebuilding property that implies vanishing of torsion homology growth and fits into our axiomatic framework for combination theorems. In particular, we obtain that certain graphs of groups with amenable vertex groups and elementary amenable edge groups have vanishing torsion homology growth.