代码大模型 / AI 编程

A Frozen 12B Beats Frontier Models on Verified Work: 100% Accuracy, 0 Tokens, Bit-Exact, Forever

一个冻结的12B模型在经过验证的任务上超越前沿模型：100%准确率、零token、位精确、永远如此

Sietse Schelpe

专题命中 仓库级理解 ：分类 cs.CL、cs.AI、cs.LG；repository(comments)

AI总结 研究提出一种让模型冻结，通过增长持久内存来解决问题的方法。在多个问题族实例上，四种架构得分优异，执行与参数扩展解耦。该方法在开放式推理中也有效，内存选择快，存储规模大，在已验证任务上超越前沿模型。

Comments Industry experience report. 14 pages, 8 figures. Public testbench: https://corbenic-galahad-bench.hf.space; companion repository with SHA-256 provenance manifest: https://github.com/corbenicai/galahad-bench

Model-Informed Joint Material-Structural Optimization of Hard-Magnetic Soft Materials

硬磁软材料的模型驱动联合材料-结构优化

Ian Galloway, Prashant K. Jha

专题命中 仓库级理解 ：repository(abstract,comments)

AI总结 针对硬磁软材料，开发模型驱动框架，将多种关系纳入统一剪切模量框架获多个本构模型，经实验选代表性关系，再用所选模型开发联合材料-结构优化框架，可处理不同设计情况，产生规定变形响应设计并开源实现。

Comments 53 pages, 20 figures, 8 tables; Repository: https://github.com/CEADpx/top_optim

NRR-Core: Non-Resolution Reasoning as a Computational Framework for Contextual Identity and Ambiguity Preservation

NRR-Core：非解析推理作为情境身份和歧义保留的计算框架

Kei Saito

机构 * Independent Researcher, Japan（日本独立研究员） 纠错

专题命中 仓库级理解 ：分类 cs.CL、cs.AI、cs.LG；repository(comments)

AI总结 本文提出NRR框架，通过多向量嵌入、非坍缩注意力和情境身份追踪，解决人工智能系统过早解析歧义的问题，保持歧义保留以提升推理灵活性。

Comments 14 pages, 2 figures, 2 tables. Replacement synced to the current GitHub repository snapshot. Series hub: https://github.com/kei-saito-research/nrr-series-hub

Fine-Grained Uncertainty Quantification for Long-Form Language Model Outputs: A Comparative Study

长文本语言模型输出的细粒度不确定性量化：一项比较研究

Dylan Bouchard, Mohit Singh Chauhan, Viren Bajaj, David Skarbrevik

机构 * CVS Health（CVS健康公司） ； Wellesley, MA（马萨诸塞州韦尔士利） 纠错

专题命中 仓库级理解 ：分类 cs.CL、cs.AI、cs.LG；repository(comments)

AI总结 针对长文本生成，提出细粒度不确定性量化分类法，通过响应分解、单元级评分和响应级聚合三阶段区分方法，引入FactScore-STEM-Geo数据集，实验发现声明-响应蕴含评分优于复杂方法，声明级评分优于句子级，不确定性感知解码有效提升事实性。

Comments Accepted by TMLR; UQLM repository: https://github.com/cvs-health/uqlm

Journal ref Transactions on Machine Learning Research, 2026

Quantifying Evidential Rigor in Meta-Analytic Corpora: A Simulation-Characterized, Bias-Robust Bayesian Workflow with a Nutrition Case Study

量化元分析语料库中的证据严谨性：一个模拟表征、偏差稳健的贝叶斯工作流及营养学案例研究

Matt Hester

专题命中 仓库级理解 ：repository(abstract,comments)

AI总结 提出一个语料库规模的贝叶斯证据审计工作流，通过联合贝叶斯因子综合效应/无效应证据和偏差成分缺失来量化证据严谨性，并使用ADEMP框架的模拟/重采样设计进行表征，以营养干预语料库作为案例研究。

Comments 59 pages, 13 figures; supplementary material included as ancillary file; companion repository archived at Zenodo, DOI: 10.5281/zenodo.20467258

The QUEST Database of Highly-Accurate Excitation Energies

高精度激发能数据库QUEST

Pierre-François Loos, Martial Boggio-Pasqua, Aymeric Blondel, Filippo Lipparini, Denis Jacquemin

专题命中 仓库级理解 ：repository(abstract,comments)

AI总结 本文介绍了QUEST数据库，该数据库提供了大量激发态和分子的垂直跃迁能的理论最佳估计，包含多种参考基组的计算结果，并评估了各种单参考和多参考波函数方法的性能。

Comments 28 pages, 7 figures (Supporting Information available). For associated git repository, see https://github.com/pfloos/QUESTDB

Journal ref J. Chem. Theory Comput. 21, 8010 (2025)

citecheck: An MCP Server for Automated Bibliographic Verification and Repair in Scholarly Manuscripts

citecheck: 一种MCP服务器，用于学术手稿中的自动文献验证与修复

Junhyeok Lee

专题命中 仓库级理解 ：repository(abstract,comments)

AI总结 citecheck通过自动化验证和修复学术手稿中的文献错误，包括识别错误、元数据不完整、作者归属错误及预印本与发表版本不一致等问题，提供结构化修复建议和安全诊断。

Comments 6 pages, 1 figure. Software paper on bibliography verification and repair for scholarly manuscripts; includes MCP server implementation and evaluation on repository-backed tests

UQLM: A Python Package for Uncertainty Quantification in Large Language Models

UQLM：用于大型语言模型不确定性量化的Python包

Dylan Bouchard, Mohit Singh Chauhan, David Skarbrevik, Ho-Kyeong Ra, Viren Bajaj, Zeya Ahmad

机构 * CVS Health（CVS健康） 纠错

专题命中 仓库级理解 ：分类 cs.CL、cs.AI、cs.LG；repository(comments)

AI总结 UQLM是一个用于检测大型语言模型幻觉的Python工具包，通过先进的不确定性量化技术提升模型输出的可靠性。

Comments Accepted by JMLR; UQLM Repository: https://github.com/cvs-health/uqlm

Journal ref Journal of Machine Learning Research, 2026

Dyad: a binary-star dynamics and statistics library for Python

Dyad：一个用于Python的双星动力学与统计库

Amery Gration

专题命中 仓库级理解 ：repository(abstract,comments)

AI总结 Dyad 是一个用于双星动力学和统计分析的 Python 库，提供概率分布模型和二体系统模拟工具。

Comments Published in The Journal of Open Source Software. Software repository at https://github.com/AmeryGration/dyad

Journal ref Journal of Open Source Software, 11(118), 9011 (2026)

Uncertainty Quantification for Language Models: A Suite of Black-Box, White-Box, LLM Judge, and Ensemble Scorers

语言模型的不确定性量化：一套黑盒、白盒、LLM评判者和集成评分器

Dylan Bouchard, Mohit Singh Chauhan

机构 * CVS Health（CVS健康公司） 纠错

专题命中 仓库级理解 ：分类 cs.CL、cs.AI、cs.LG；repository(comments)

AI总结 本文提出了一种灵活的框架，利用黑盒、白盒、LLM评判者和集成评分器来检测语言模型的幻觉问题，通过可调节的集成方法提升检测性能。

Comments Accepted by TMLR; UQLM repository: https://github.com/cvs-health/uqlm

Journal ref Transactions on Machine Learning Research, 2025

Formalizing Computational Paths and Fundamental Groups in Lean

在Lean中形式化计算路径与基本群

Arthur F. Ramos, Anjolina G. de Oliveira, Ruy J. G. B. de Queiroz, Tiago M. L. de Veras

专题命中 仓库级理解 ：repository(abstract,comments)

AI总结 本文在Lean 4中形式化了计算路径理论，并通过六个代数拓扑例子展示了其在同伦计算中的应用。

Comments 27 pages, 2 figures. All definitions and proofs are available in the ComputationalPathsLean GitHub repository

crate2bib: Citing Rust crates made easy

Jonas Pleyer

专题命中 仓库级理解 ：repository(abstract,comments)

Comments See the Github repository: https://github.com/jonaspleyer/crate2bib/

Improving the solver for the Balitsky-Kovchegov evolution equation with Automatic Differentiation

Florian Cougoulic, Piotr Korcyl, Tomasz Stebel

专题命中 仓库级理解 ：repository(abstract,comments)

Comments 17 pages, 4 figures, source code is published in the repository

Journal ref Comput.Phys.Commun. 313 (2025) 109616

The Local Volume Database: a library of the observed properties of nearby dwarf galaxies and star clusters

Andrew B. Pace

专题命中 仓库级理解 ：repository(abstract,comments)

Comments 22 pages, 13 figures, Published in the Open Journal of Astrophysics, GitHub repository and catalogs are available at https://github.com/apace7/local_volume_database

A neural network model for quickly solving multiple-band light curves of contact binaries

Kai Li, Li-Heng Wang, Xiang Gao

专题命中 仓库级理解 ：repository(abstract,comments)

Comments 10pages, 4 figures, 3 tables, accepted by ApJS, Data available via China-VO PaperData repository

A Survey: Learning Embodied Intelligence from Physical Simulators and World Models

Xiaoxiao Long, Qingrui Zhao, Kaiwen Zhang, Zihao Zhang, Dingrui Wang, Yumeng Liu, Zhengjie Shu, Yi Lu, Shouzheng Wang, Xinzhe Wei, Wei Li, Wei Yin, Yao Yao, Jia Pan, Qiu Shen, Ruigang Yang, Xun Cao, Qionghai Dai

机构 * Nanjing University（南京大学） ； University of Hong Kong（香港大学） ； Central South University（中南大学） ； Horizon Robotics（Horizon机器人） ； Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所） ； University of Chinese Academy of Sciences（中国科学院大学） ； Shanghai Jiao Tong University（上海交通大学） ； Technical University of Munich（慕尼黑技术大学） ； Tsinghua University（清华大学） 纠错

专题命中 仓库级理解 ：repository(abstract,comments)

Comments Update with recent progresses. 49pages, 25figures, 6tables, github repository avalible in https://github.com/NJU3DV-LoongGroup/Embodied-World-Models-Survey

QUEENS: An Open-Source Python Framework for Solver-Independent Analyses of Large-Scale Computational Models

Jonas Biehler, Jonas Nitzler, Sebastian Brandstaeter, Maximilian Dinkel, Volker Gravemeier, Lea J. Haeusel, Gil Robalo Rei, Harald Willmann, Barbara Wirthl, Wolfgang A. Wall

专题命中 仓库级理解 ：repository(abstract,comments)

Comments 17 pages, 6 figures, one Github repository https://github.com/queens-py/queens

mcdust: A 2D Monte Carlo code for dust coagulation in protoplanetary disks

Vignesh Vaikundaraman, Nerea Gurrutxaga, Joanna Drążkowska

专题命中 仓库级理解 ：code model(abstract)；repository(comments)

Comments Submitted to the Journal of Open Source Software (JOSS). Documentation: https://mcdust.readthedocs.io/en/latest/. Repository: https://github.com/vicky1997/mcdust

Graph-of-Causal Evolution: Challenging Chain-of-Model for Reasoning

Libo Wang

专题命中 仓库级理解 ：分类 cs.CL、cs.AI、cs.LG；repository(comments)

Comments The relevant code has been uploaded to the publicly available GitHub repository. The link is: https://github.com/brucewang123456789/GeniusTrail/tree/main/GoCE

On the Ubiquity Of Electromagnetic-Duality Rotations in 4D, N = 1 Holoraumy Tensors for On-Shell 4D Supermultiplets

S. James Gates,, Daniel Lay, S. -N. Hazel Mak, Brock Peters, Aravind Ramakrishnan, Kory Stiffler, Zachary Wimpee, Xiao Xiao, Yifan Yuan, Jinjie Zhang, Peter V. Zhou

专题命中 仓库级理解 ：repository(abstract,comments)

Comments Latex twice, 21 pages, final version includes one additional reference, teo citation correction, and 2-form clarifying language, link to the HEPTHools Data Repository included

Journal ref Int.J.Mod.Phys.A 35 (2020) 01, 2050008

Exploring the Abelian 4D, $\mathcal{N}$ = 4 Vector-Tensor Supermultiplet and Its Off-Shell Central Charge Structure

S. James Gates,, Kory Stiffler

专题命中 仓库级理解 ：repository(abstract,comments)

Comments HEPTHools Data Repository available open-source at https://hepthools.github.io/Data/, added references and related content, corrected group from SU(2) to GL(2,R)

Journal ref JHEP 03 (2019) 129

pyBoLaNO: A Python symbolic package for normal ordering involving bosonic ladder operators

Hendry M. Lim, Donny Dwiputra, M. Shoufie Ukhtary, Ahmad R. T. Nugraha

专题命中 仓库级理解 ：repository(abstract,comments)

Comments 14 pages, 2 figures; GitHub repository at https://github.com/hendry24/pyBoLaNO; More detailed analysis in the Performance section; Addressed the ambiguity in the terminology; Added LaTeX renders of the demonstration outputs; Typesetting fixes and other small tweaks

Journal ref Computer Physics Communications 313 (2025); 109622

Wormhole Memory: A Rubik's Cube for Cross-Dialogue Retrieval

Libo Wang

专题命中 仓库级理解 ：分类 cs.CL、cs.AI、cs.LG；repository(comments)

Comments The experimental process and code have been uploaded to the Github repository, the link is: https://github.com/brucewang123456789/GeniusTrail/tree/main/Wormhole%20Memory%20Module

PyLongslit: a simple manual Python pipeline for processing of astronomical long-slit spectra recorded with CCD detectors

Kostas Valeckas, Johan Peter Uldall Fynbo, Jens-Kristian Krogager, Kasper Elm Heintz

专题命中 仓库级理解 ：repository(abstract,comments)

Comments Submitted to JOSS. Software repository: https://github.com/KostasValeckas/PyLongslit

Image Pre-Processing Framework for Time-Domain Astronomy in the Artificial Intelligence Era

Liang Cao, Peng Jia, Jiaxin Li, Yu Song, Chengkun Hou, Yushan Li

专题命中 仓库级理解 ：repository(abstract,comments)

Comments Accepted by the AJ. The Docker Version could be found in the PaperData Repository powered by China-VO

Large Language Models and Synthetic Data for Monitoring Dataset Mentions in Research Papers

Aivin V. Solatorio, Rafael Macalaba, James Liounis

专题命中 仓库级理解 ：分类 cs.CL、cs.AI、cs.LG；repository(comments)

Comments Project GitHub repository at https://github.com/worldbank/ai4data-use

Brain in the Dark: Design Principles for Neuromimetic Inference under the Free Energy Principle

Mehran H. Bazargani, Szymon Urbas, Karl Friston

专题命中 仓库级理解 ：repository(abstract,comments)

Comments 14 pages, 4 figures, 1 GitHub repository

A ripple in time: a discontinuity in American history

Alexander Kolpakov, Igor Rivin

专题命中 仓库级理解 ：分类 cs.CL、cs.AI、cs.LG；repository(comments)

Comments 6 pages, 8 figures ; GitHub repository https://github.com/sashakolpakov/ripple_in_time ; to appear in 8th NLPIR Okayama, Japan | December 13-15, 2024 as "Discovering temporal and personality aspects in meager and highly variable text samples"

From Generalist to Specialist: A Survey of Large Language Models for Chemistry

Yang Han, Ziping Wan, Lu Chen, Kai Yu, Xin Chen

专题命中 仓库级理解 ：分类 cs.CL、cs.AI、cs.LG；repository(comments)

Comments COLING2025,We maintain an up-to-date Github repository at: https://github.com/OpenDFM/LLM4Chemistry

PaCE: Parsimonious Concept Engineering for Large Language Models

Jinqi Luo, Tianjiao Ding, Kwan Ho Ryan Chan, Darshan Thaker, Aditya Chattopadhyay, Chris Callison-Burch, René Vidal

专题命中 仓库级理解 ：分类 cs.CL、cs.AI、cs.LG；repository(comments)

Comments Accepted in NeurIPS 2024. GitHub repository at https://github.com/peterljq/Parsimonious-Concept-Engineering