跳到正文
研踪PaperTrail

论文库

检索计算机领域论文,按研究方向、作者或关键词筛选。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+1680新增论文
+1333其中 AI 相关
+187已有论文修订
+59新增 AI 导读

arXiv 最近完成扫描:2026-11-18 04:28 · 全站最近入库:2026-11-17 23:35各信源进度 →

统计口径与数据新鲜度

统计 2026-11-17 11:27 — 2026-11-18 11:27(北京时间) · 点击数字查看对应简报;AI 相关包含在新增论文内。

最近完成检查:2026-11-18 04:28最近发现新增版本:2026-11-17 23:35库内论文最新更新时间:2026-11-17 02:59

每分钟刷新已保存的数据。空检查不代表采集异常;新收录可能包含历史论文。AI 导读按新增版本计数。

25,786 已收录论文21,491 AI 相关7,725 有修订112 HF 社区发现

浏览论文库

时间范围跟随排序:版本更新、本站收录或首次发表。最新收录可能包含历史论文。部分信源异常,查看状态 →

找到 3 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-18 04:27(北京时间)
新论文 · v1arXiv:2611.17650

Classifier-Guided Latent Diffusion for Small-Molecule Generation

Ali Forooghi, Luis Rueda, Alioune Ngom

原文摘要

Diffusion models have emerged as a promising class of generative models for molecular design because of their ability to learn complex chemical distributions and support conditional or guided generation. In this work, we investigate a latent diffusion pipeline for property-guided small-molecule generation. The proposed framework represents canonicalized SMILES strings using a frozen ChemBERTa encoder, trains a Transformer-based decoder to reconstruct SELFIES strings from molecular latent embeddings, and learns an unconditional denoising diffusion model in the ChemBERTa latent space. To enable property-directed generation, a differentiable latent property predictor is used to guide the diffusion sampling process. Generated SELFIES strings are converted to SMILES, validated using RDKit, and evaluated using molecular validity, uniqueness, novelty, and predicted property scores. We study guidance toward two drug-relevant binary endpoints: blood--brain barrier permeability and inhibition of human $β$-secretase 1. The results suggest that latent diffusion over pretrained molecular embeddings can generate chemically valid and novel molecules while enabling controllable steering toward desired molecular properties.

cs.LG

新论文 · v1arXiv:2611.15001

Understanding Suggested Edits in Stack Overflow: A Large-Scale Empirical Study and Review Prioritization Framework

Mehedi Hasan Shanto, Muhammad Asaduzzaman, Md. Ahasanuzzaman, Alioune Ngom

原文摘要

Suggested edits allow low-reputation users to improve Stack Overflow posts and contribute to the platform's knowledge base. However, we still know little about why these edits are accepted or rejected and how reviewers can efficiently identify problematic edits. In this study, we analyze 3,888,805 suggested edits submitted between 2011 and May 2023. First, we compare accepted and rejected edits using 35 features. We find significant differences between them, with the largest differences related to users' editing history. Second, we manually analyze rejected edits and identify 17 fine-grained rejection reasons. These reasons provide a more detailed understanding of why edits are rejected than Stack Overflow's predefined rejection reasons. Third, we evaluate three models for prioritizing edits based on their likelihood of rejection. The Feature Model, built using the 35 studied features, provides the best trade-off between review rate and rejected-edit recall. At a decision threshold of $τ= 0.40$, it prioritizes 52.3% of edits for human review while capturing 80.8% of rejected edits. In comparison, the Embedding Model captures 78.8% of rejected edits while prioritizing 60.9% of edits, and the EditEx Model captures 72.5% while prioritizing 59.0%. We also conduct a complementary zero-shot evaluation on a separate set of 500 edits. The Feature Model achieves 78.0% recall at $τ= 0.40$, compared with 23.6%-31.2% for GPT-4o, Qwen2.5-32B, and Llama-3.3-70B. Overall, our findings provide a detailed understanding of suggested-edit rejection and show that rejection-risk prediction can help prioritize potentially problematic edits for human review.

cs.SE

版本更新 · v2arXiv:2611.02383

Toward an Evidence-Based Template for Agent Instruction Files: An Empirical Study of Adoption, Consistency, and Maintenance

Mehedi Hasan Shanto, Md. Anaytul Islam, Muhammad Asaduzzaman, Alioune Ngom

原文摘要

AI coding agents rely on repository-level instruction files to understand project conventions, development workflows, and constraints. However, different agents support different instruction formats, and limited empirical evidence exists about how these files are adopted, what information they contain, whether coexisting formats provide consistent guidance, and how they are maintained. We conduct an empirical study of 3,387 instruction files from 1,000 GitHub repositories across four formats: AGENTS.md, CLAUDE.md, GitHub Copilot instructions, and Cursor rules. We find that 42.2% of instruction-bearing repositories in the higher-popularity cohort maintain multiple formats. Among 386 manually examined pairs of corresponding sections, 16.1% provide divergent guidance. Our maintenance analysis further shows that 40.4% of the analyzed files have only one recorded modification event and that maintenance patterns differ across formats. Based on category prevalence and category addition ratios (CAR), we derive an evidence-based template containing ten Required, four Gap, and two Recommended categories. In an exploratory generation experiment, explicit template guidance increases structural category coverage from 18.7% to 82.8%. We further introduce AGIT-Instruct, a benchmark for evaluating instruction-file generation through structural coverage and content alignment. Across four generation configurations, retrieval-augmented generation achieves the highest full-file ROUGE-1 F1 of 0.459. A manual assessment of 278 generated category instances finds that 19.4% fully preserve and 61.2% partially preserve developer-written intent. Our findings provide empirical guidance for organizing and maintaining agent instruction

cs.SE