Pith. sign in

REVIEW 3 major objections 1 cited by

If a hypergraph has vanishing 2-degree Turán density, its edges must admit a global 2-vanishing order.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-15 14:05 UTC pith:SE64HJOD

load-bearing objection We only have the abstract of the vanishing-order paper; the supplied body is an unrelated code-completion paper, so the proofs cannot be checked. the 3 major comments →

arxiv 2603.05973 v2 pith:SE64HJOD submitted 2026-03-06 math.CO

Vanishing orders, suspensions and zero degree Tur\'an densities

Jiangdong Ai , Laihao Ding , Hong Liu , Haotian Yang This is my paper
classification math.CO MSC 05C6505C3505D05
keywords degree Turán density2-vanishing orderhypergraph Turán problemsuspension principlecodegree densityextremal hypergraph theory
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Classical Turán theory says a k-uniform hypergraph has zero edge-density threshold for forcing a copy of F precisely when F is k-partite. This paper studies the analogous question for higher-degree thresholds: when the minimum number of edges through every pair (or every ℓ-set) can be forced arbitrarily small yet still avoid F. The central claim for pairs is structural: whenever that 2-degree threshold vanishes, the vertices of F can be ordered so that every edge sits in a canonical way relative to its pairs. The authors also prove a suspension bridge between consecutive degree parameters—adding an apex vertex preserves vanishing of the density—and deduce that every non-classical degree Turán density accumulates at zero. The result supplies a concrete obstruction (lack of a 2-vanishing order) that prevents the 2-degree threshold from disappearing, extending a known 3-uniform phenomenon to all uniformities.

Core claim

For every uniformity k ≥ 3, a k-graph F satisfies π₂(F) = 0 if and only if it admits a 2-vanishing order: a linear order on its vertices under which every edge is canonically aligned with respect to its pairs. Separately, suspending a (k−1)-graph F by an apex yields a k-graph whose ℓ-degree Turán density vanishes exactly when the (ℓ−1)-degree density of F vanishes; consequently all degree Turán densities other than the classical edge density accumulate at zero.

What carries the argument

The 2-vanishing order (a global vertex ordering forcing every edge into a canonical pair-alignment) together with the suspension operator S_F that lifts vanishing statements between consecutive degree parameters; the constructive half is realized by random geometric blocks, design-theoretic gluing, and random sparsification that keep positive minimum 2-degree while locally avoiding non-ordered F.

Load-bearing premise

The construction that glues random geometric pieces via designs and then sparsifies them really does produce host hypergraphs with positive minimum 2-degree that still contain no copy of any F lacking a 2-vanishing order.

What would settle it

Exhibit a single k-uniform F that has no 2-vanishing order yet still satisfies π₂(F) = 0, or show that every host built by the geometric-design-sparsification scheme inevitably forces such an F once the minimum 2-degree stays bounded away from zero.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Absence of a 2-vanishing order is a concrete structural certificate that π₂(F) > 0.
  • Vanishing results for any degree parameter can be lifted or lowered across uniformities by repeated suspension.
  • All ℓ-degree Turán densities with ℓ ≥ 2 accumulate at zero, in contrast to the classical edge-density case.
  • The same obstruction-and-construction template may extend to characterize vanishing of π_ℓ for ℓ > 2.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The geometric-design construction may be adaptable to give quantitative lower bounds on π₂(F) whenever a 2-vanishing order is absent, not merely positivity.
  • If the suspension principle iterates cleanly, the only degree parameter whose zero-set is “large” (the multipartite hypergraphs) is the classical edge density; every higher-degree zero-set is sparse in a strong sense.
  • A natural next test is whether an analogous “ℓ-vanishing order” fully characterizes π_ℓ(F) = 0 for each fixed ℓ between 3 and k−2.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 0 minor

Summary. From the abstract, the paper studies structural consequences of vanishing ℓ-degree Turán densities π_ℓ(F) for k-uniform hypergraphs. The main claim is that for every k≥3, π_2(F)=0 forces F to admit a 2-vanishing order (a global vertex ordering under which every edge is canonically aligned with respect to its pairs). A suspension principle is also claimed: for a (k-1)-graph F and its k-uniform suspension S_F, π_ℓ(S_F)=0 if and only if π_{ℓ-1}(F)=0 (2≤ℓ<k). As an application, all non-classical degree Turán densities are said to accumulate at zero. The abstract sketches a proof strategy for the main theorem that combines random geometric building blocks, design-theoretic gluing, and random sparsification.

Significance. If the stated theorems hold, the work would be a substantial contribution to extremal hypergraph theory. It would extend Erdős’s classical characterization of zero Turán density (k-partiteness) to the 2-degree setting for all uniformities, generalize a known 3-graph phenomenon, and supply a clean bridge (via suspensions) that lifts vanishing results across degree parameters and uniformities. The accumulation statement would also clarify the landscape of intermediate degree Turán densities. These are natural, high-value structural questions; machine-checkable or fully constructive proofs of the geometric/design/sparsification engine would further strengthen the contribution.

major comments (3)
  1. The supplied full manuscript body is not the paper described by the title, abstract, and arXiv identifier 2603.05973. The body is instead the complete text of a software-engineering paper (MCCom / local-cloud model cascading for code completion, arXiv:2603.05974). Consequently the central combinatorial constructions (random geometric building blocks, design-theoretic gluing, random sparsification), all lemmas, and the proofs of the 2-vanishing-order theorem and the suspension principle are absent and cannot be checked.
  2. Because the constructive half of the main obstruction (producing host k-graphs with positive minimum 2-degree that still avoid every F lacking a 2-vanishing order) is only named in the abstract and never appears in the supplied text, the load-bearing technical engine of the principal theorem is unevaluable. No verification of soundness, circularity, or completeness of the argument is possible from the material provided.
  3. The suspension equivalence and the accumulation-at-zero corollary likewise rest on arguments that are not present in the supplied manuscript. Until the correct body is furnished, these claims remain unrefereed assertions rather than established theorems.

Circularity Check

0 steps flagged

No circularity detectable: abstract states structural implications only; supplied body is an unrelated SE paper, so no derivation chain exists to reduce.

full rationale

The supplied CACHEABLE full manuscript is arXiv:2603.05974 (MCCom code-completion cascading), not the claimed math.CO paper 2603.05973. Only the abstract of the Turán-density paper is present. That abstract asserts pure structural theorems (π₂(F)=0 implies existence of a 2-vanishing order; the suspension equivalence π_ℓ(S_F)=0 ⇔ π_{ℓ-1}(F)=0) with no free parameters, no fitted quantities re-labeled as predictions, and no self-referential definitions. No equations, lemmas, or constructions appear that could be checked for self-definitional reduction, fitted-input-as-prediction, or load-bearing self-citation. Consequently the derivation chain cannot be walked, no circular step can be exhibited by quotation, and the honest score is 0. Ordinary dependence on the classical definitions of π_ℓ is not circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 2 invented entities

Abstract-only review of a pure extremal-combinatorics paper. No numerical free parameters appear. Background axioms are the standard definitions of ℓ-degree Turán density and the classical Erdős multipartite characterization for π₁=0. The 2-vanishing order and the suspension S_F are the paper’s own structural notions; they are definitional rather than physical entities, so they are recorded as invented entities with independent_evidence false pending external use.

axioms (3)
  • domain assumption Standard definition of the ℓ-degree Turán density π_ℓ(F) as the liminf of the minimum ℓ-degree threshold forcing a copy of F.
    Invoked throughout the abstract as the object of study; taken from the existing higher-degree Turán literature.
  • standard math Erdős’ characterization: π₁(F)=0 if and only if F is k-partite.
    Cited as the classical fact of which the 2-vanishing-order theorem is the higher-degree analogue.
  • ad hoc to paper Existence of random geometric constructions and design-theoretic gluings that can enforce positive minimum 2-degree while preserving local vanishing structure.
    The abstract states that the main proof combines these ingredients; their success is an unproved (from the abstract) technical premise of the obstruction theorem.
invented entities (2)
  • 2-vanishing order no independent evidence
    purpose: Global vertex ordering that forces every edge of F to align canonically with respect to its pairs; used as the structural certificate for π₂(F)=0.
    Defined in the paper as the higher-uniformity analogue of multipartite orderings; no independent external verification supplied in the abstract.
  • Suspension S_F no independent evidence
    purpose: k-graph obtained by adding an apex and replacing each edge e of a (k-1)-graph F by v∪e; used to transfer vanishing between consecutive degree parameters.
    Construction introduced to prove the equivalence π_ℓ(S_F)=0 ⇔ π_{ℓ-1}(F)=0; definitional, not an external object.

pith-pipeline@v1.1.0-grok45 · 27622 in / 2672 out tokens · 26962 ms · 2026-07-15T14:05:08.688655+00:00 · methodology

0 comments
read the original abstract

For integers $1\le \ell<k$, the $\ell$-degree Tur\'an density $\pi_\ell(F)$ measures the minimum $\ell$-degree threshold that forces a copy of a fixed $k$-uniform hypergraph $F$, generalizing both the classical Tur\'an density $\pi_1$ and the codegree Tur\'an density $\pi_{k-1}$. Motivated by Erd\H{o}s' characterization of $k$-graphs with zero Tur\'an density, we study the structural implications of vanishing $\ell$-degree Tur\'an density. Our main result concerns the case $\ell=2$. We prove that, for every $k\ge3$, if a $k$-graph $F$ satisfies $\pi_2(F)=0$, then $F$ admits a $2$-vanishing order, that is, a global vertex ordering under which all edges align canonically with respect to their pairs. This extends to all uniformities a structural phenomenon previously known for $3$-graphs, and gives a higher-degree analogue of the classical fact that $\pi_1(F)=0$ forces $F$ to be $k$-partite. In particular, the absence of a $2$-vanishing order is a structural obstruction to vanishing $2$-degree Tur\'an density. We also establish a suspension principle connecting consecutive degree parameters. Given a $(k-1)$-graph $F$, let $\mathcal{S}_F$ be the $k$-graph obtained by adding an apex vertex $v$ and replacing each edge $e\in E(F)$ with $v\cup e$. We show that, for $2\le \ell<k$, $\pi_{\ell}(\mathcal{S}_F)=0$ if and only if $\pi_{\ell-1}(F)=0$. This provides a bridge between different degree Tur\'an densities and allows vanishing results to be lifted across uniformities and degree parameters. As an application, we prove that except the classical Tur\'an density, all other degree Tur\'an densities accumulate at zero. The proof of our main result combines random geometric building blocks, a design-theoretic gluing scheme, and random sparsification to reconcile positive $2$-degree with local vanishing structure.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Tree suspensions and transfer functions for single degree Tur\'an spectra

    math.CO 2026-07 accept novelty 7.5

    Tree suspensions realize three transfer functions on single degree Turán spectra, yielding infinitely many accumulation points for all parameters and arbitrarily high algebraic degree for ordinary and half-or-higher d...

Reference graph

Works this paper leans on

49 extracted references · 13 linked inside Pith · cited by 1 Pith paper

  1. [1]

    d.].Amazon CodeWhisperer

    [n. d.].Amazon CodeWhisperer. https://aws.amazon.com/codewhisperer/

  2. [2]

    GitHub 2025.GitHub Copilot�Your AI Pair Programmer. GitHub. https://github.com/features/copilot

    2025

  3. [3]

    d.].Tabnine AI Code Assistant | Private, Personalized, Protected

    Tabnine [n. d.].Tabnine AI Code Assistant | Private, Personalized, Protected. Tabnine. https://www.tabnine.com/ Proc. ACM Softw. Eng., Vol. 3, No. FSE, Article FSE060. Publication date: July 2026. Balancing Latency and Accuracy of Code Completion via Local-Cloud Model Cascading FSE060:21

    2026

  4. [4]

    d.].TongYiLingMa

    [n. d.].TongYiLingMa. https://lingma.aliyun.com/

  5. [5]

    d.].Windsurf (Formerly Codeium) - The Most Powerful AI Code Editor

    [n. d.].Windsurf (Formerly Codeium) - The Most Powerful AI Code Editor. https://windsurf.com/

  6. [6]

    Sven Amann, Sebastian Proksch, Sarah Nadi, and Mira Mezini. 2016. A study of visual studio usage in practice. In2016 IEEE 23rd International Conference on Software Analysis, Evolution, and Reengineering, Vol. 1. IEEE, 124–134

    2016

  7. [7]

    Muhammad Asaduzzaman, Chanchal K Roy, Kevin A Schneider, and Daqing Hou. 2014. Cscc: Simple, efficient, context sensitive code completion. In2014 IEEE International Conference on Software Maintenance and Evolution. IEEE, 71–80

    2014

  8. [8]

    Mohammad Bavarian, Heewoo Jun, Nikolas Tezak, John Schulman, Christine McLeavey, Jerry Tworek, and Mark Chen

  9. [9]

    Efficient training of language models to fill in the middle.arXiv preprint arXiv:2207.14255(2022)

    Pith/arXiv arXiv 2022

  10. [10]

    Marcel Bruch, Martin Monperrus, and Mira Mezini. 2009. Learning from examples to improve code completion systems. InProceedings of the 7th joint meeting of the European software engineering conference and the ACM SIGSOFT symposium on the foundations of software engineering. 213–222

    2009

  11. [11]

    Boyuan Chen, Mingzhi Zhu, Brendan Dolan-Gavitt, Muhammad Shafique, and Siddharth Garg. 2025. Model cascading for code: A cascaded black-box multi-model framework for cost-efficient code completion with self-testing. In2025 International Joint Conference on Neural Networks. IEEE, 1–9

    2025

  12. [12]

    Lingjiao Chen, Matei Zaharia, and James Zou. 2024. FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance.Transactions on Machine Learning Research(2024)

    2024

  13. [13]

    Wei Chen and Zhiyuan Li. 2024. Octopus v4: Graph of language models.arXiv preprint arXiv:2404.19296(2024)

    Pith/arXiv arXiv 2024

  14. [14]

    Ziyi Chen, Xiaocong Yang, Jiacheng Lin, Chenkai Sun, Kevin Chang, and Jie Huang. 2024. Cascade speculative drafting for even faster llm inference.Advances in Neural Information Processing Systems37 (2024), 86226–86242

    2024

  15. [15]

    Le Deng, Xiaoxue Ren, Chao Ni, Ming Liang, David Lo, and Zhongxin Liu. 2025. Enhancing project-specific code completion by inferring internal api information.IEEE Transactions on Software Engineering(2025)

    2025

  16. [16]

    Yangruibo Ding, Zijian Wang, Wasi Ahmad, Murali Krishna Ramanathan, Ramesh Nallapati, Parminder Bhatia, Dan Roth, and Bing Xiang. 2024. Cocomic: Code completion by jointly modeling in-file and cross-file context. InProceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation. 3433–3445

    2024

  17. [17]

    Aryaz Eghbali and Michael Pradel. 2024. De-hallucinator: Iterative grounding for llm-based code completion.arXiv preprint arXiv:2401.01701(2024)

    Pith/arXiv arXiv 2024

  18. [18]

    Tao Feng, Yanzhen Shen, and Jiaxuan You. 2025. GraphRouter: A Graph-based Router for LLM Selections. InThe Thirteenth International Conference on Learning Representations

    2025

  19. [19]

    2002.Facts and fallacies of software engineering

    Robert L Glass. 2002.Facts and fallacies of software engineering. Addison-Wesley Professional

    2002

  20. [20]

    Daya Guo, Qihao Zhu, Dejian Yang, Zhenda Xie, Kai Dong, Wentao Zhang, Guanting Chen, Xiao Bi, Yu Wu, YK Li, et al. 2024. DeepSeek-Coder: When the Large Language Model Meets Programming–The Rise of Code Intelligence. arXiv preprint arXiv:2401.14196(2024)

    Pith/arXiv arXiv 2024

  21. [21]

    Daqing Hou and David M Pletcher. 2010. Towards a better code completion system by API grouping, filtering, and popularity-based ranking. InProceedings of the 2nd International Workshop on Recommendation Systems for Software Engineering. 26–30

    2010

  22. [22]

    Shengding Hu, Yuge Tu, Xu Han, Chaoqun He, Ganqu Cui, Xiang Long, Zhi Zheng, Yewei Fang, Yuxiang Huang, Weilin Zhao, et al. 2024. Minicpm: Unveiling the potential of small language models with scalable training strategies. arXiv preprint arXiv:2404.06395(2024)

    Pith/arXiv arXiv 2024

  23. [23]

    Siming Huang, Tianhao Cheng, Jason Klein Liu, Weidi Xu, Jiaran Hao, Liuyihan Song, Yang Xu, Jian Yang, Jiaheng Liu, Chenchen Zhang, et al. 2025. Opencoder: The open cookbook for top-tier code large language models. InProceedings of the 63rd Annual Meeting of the Association for Computational Linguistics. 33167–33193

    2025

  24. [24]

    Binyuan Hui, Jian Yang, Zeyu Cui, Jiaxi Yang, Dayiheng Liu, Lei Zhang, Tianyu Liu, Jiajun Zhang, Bowen Yu, Keming Lu, et al. 2024. Qwen2. 5-coder technical report.arXiv preprint arXiv:2409.12186(2024)

    Pith/arXiv arXiv 2024

  25. [25]

    Siyuan Jiang, Jia Li, He Zong, Huanyu Liu, Hao Zhu, Shukai Hu, Erlu Li, Jiazheng Ding, Yu Han, Wei Ning, et al

  26. [26]

    In2025 IEEE/ACM 47th International Conference on Software Engineering: Software Engineering in Practice

    aiXcoder-7B: A Lightweight and Effective Large Language Model for Code Processing. In2025 IEEE/ACM 47th International Conference on Software Engineering: Software Engineering in Practice. IEEE, 215–226

  27. [27]

    Vladimir I Levenshtein et al. 1966. Binary codes capable of correcting deletions, insertions, and reversals. InSoviet physics doklady, Vol. 10. Soviet Union, 707–710

    1966

  28. [28]

    Yaniv Leviathan, Matan Kalman, and Yossi Matias. 2023. Fast inference from transformers via speculative decoding. In International Conference on Machine Learning. PMLR, 19274–19286

    2023

  29. [29]

    Wei Liu, Ailun Yu, Daoguang Zan, Bo Shen, Wei Zhang, Haiyan Zhao, Zhi Jin, and Qianxiang Wang. 2024. Graphcoder: Enhancing repository-level code completion via code context graph-based retrieval and language model.arXiv preprint arXiv:2406.07003(2024)

    Pith/arXiv arXiv 2024

  30. [30]

    Anton Lozhkov, Raymond Li, Loubna Ben Allal, Federico Cassano, Joel Lamy-Poirier, Nouamane Tazi, Ao Tang, Dmytro Pykhtar, Jiawei Liu, Yuxiang Wei, et al. 2024. Starcoder 2 and the stack v2: The next generation.arXiv preprint arXiv:2402.19173(2024). Proc. ACM Softw. Eng., Vol. 3, No. FSE, Article FSE060. Publication date: July 2026. FSE060:22 Hanzhen Lu, L...

    Pith/arXiv arXiv 2024

  31. [31]

    Gail C Murphy, Mik Kersten, and Leah Findlater. 2006. How are Java software developers using the Eclipse IDE?IEEE software23, 4 (2006), 76–83

    2006

  32. [32]

    Huy Nhat Phan, Hoang Nhat Phan, Tien N Nguyen, and Nghi DQ Bui. 2024. Repohyper: Better context retrieval is all you need for repository-level code completion.CoRR(2024)

    2024

  33. [33]

    Mohammad Reza Rezaei and Adji Bousso Dieng. 2025. Vendi-rag: Adaptively trading-off diversity and quality significantly improves retrieval augmented generation with llms.arXiv preprint arXiv:2502.11228(2025)

    Pith/arXiv arXiv 2025

  34. [34]

    Baptiste Roziere, Jonas Gehring, Fabian Gloeckle, Sten Sootla, Itai Gat, Xiaoqing Ellen Tan, Yossi Adi, Jingyu Liu, Romain Sauvestre, Tal Remez, et al. 2023. Code llama: Open foundation models for code.arXiv preprint arXiv:2308.12950 (2023)

    Pith/arXiv arXiv 2023

  35. [35]

    Hitesh Sagtani, Rishabh Mehrotra, and Beyang Liu. 2025. Improving fim code completions via context & curriculum based learning. InProceedings of the Eighteenth ACM International Conference on Web Search and Data Mining. 801–810

    2025

  36. [36]

    Anton Semenkin, Vitaliy Bibaev, Yaroslav Sokolov, Kirill Krylov, Alexey Kalina, Anna Khannanova, Danila Savenkov, Darya Rovdo, Igor Davidenko, Kirill Karnaukhov, et al. 2025. Full line code completion: Bringing ai to desktop. In2025 IEEE/ACM 47th International Conference on Software Engineering: Software Engineering in Practice. IEEE, 563–574

    2025

  37. [37]

    Liangying Shao, Yanfu Yan, Denys Poshyvanyk, and Jinsong Su. 2025. UniGenCoder: Merging SEQ2SEQ and SEQ2TREE Paradigms for Unified Code Generation. In2025 IEEE/ACM 47th International Conference on Software Engineering: New Ideas and Emerging Results (ICSE-NIER). 71–75. doi:10.1109/ICSE-NIER66352.2025.00020

    doi:10.1109/icse-nier66352.2025.00020 2025

  38. [38]

    Tal Shnitzer, Anthony Ou, Mírian Silva, Kate Soule, Yuekai Sun, Justin Solomon, Neil Thompson, and Mikhail Yurochkin

  39. [39]

    Large language model routing with benchmark datasets.arXiv preprint arXiv:2309.15789(2023)

    Pith/arXiv arXiv 2023

  40. [40]

    2025.GitHub Copilot Chat Explained: The Life of a Prompt

    sombanerjee. 2025.GitHub Copilot Chat Explained: The Life of a Prompt. All things Azure. https://devblogs.microsoft. com/all-things-azure/github-copilot-chat-explained-the-life-of-a-prompt/

    2025

  41. [41]

    TechEpiphany. 2025. GPU Retail Sales June ’25 Amazon US. https://x.com/TechEpiphanyYT/status/ 1941064015016341736?s=20

    2025

  42. [42]

    Chaozheng Wang, Junhao Hu, Cuiyun Gao, Yu Jin, Tao Xie, Hailiang Huang, Zhenyu Lei, and Yuetang Deng. 2023. Practitioners’ expectations on code completion.arXiv preprint arXiv:2301.03846(2023)

    Pith/arXiv arXiv 2023

  43. [43]

    Yanlin Wang, Yanli Wang, Daya Guo, Jiachi Chen, Ruikai Zhang, Yuchi Ma, and Zibin Zheng. 2024. RLCoder: Reinforcement Learning for Repository-Level Code Completion. In2025 IEEE/ACM 47th International Conference on Software Engineering. IEEE Computer Society, 165–177

    2024

  44. [44]

    Di Wu, Wasi Uddin Ahmad, Dejiao Zhang, Murali Krishna Ramanathan, and Xiaofei Ma. 2024. REPOFORMER: selective retrieval for repository-level code completion. InProceedings of the 41st International Conference on Machine Learning. 53270–53290

    2024

  45. [45]

    Jingrong Yu, Zhipeng Gao, Lingfeng Bao, and Zhongxin Liu. 2025. Enhancing Domain-Specific Code Completion via Collaborative Inference with Large and Small Language Models.ACM Transactions on Software Engineering and Methodology(2025)

    2025

  46. [46]

    Fengji Zhang, Bei Chen, Yue Zhang, Jacky Keung, Jin Liu, Daoguang Zan, Yi Mao, Jian-Guang Lou, and Weizhu Chen

  47. [47]

    InProceedings of the 2023 Conference on Empirical Methods in Natural Language Processing

    RepoCoder: Repository-Level Code Completion Through Iterative Retrieval and Generation. InProceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 2471–2484

    2023

  48. [48]

    Sheng Zhang, Yifan Ding, Shuquan Lian, Shun Song, and Hui Li. 2025. CodeRAG: Finding Relevant and Necessary Knowledge for Retrieval-Augmented Repository-Level Code Completion. InProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 23289–23299

    2025

  49. [49]

    Weilin Zhao, Yuxiang Huang, Xu Han, Chaojun Xiao, Zhiyuan Liu, and Maosong Sun. 2024. Ouroboros: Speculative decoding with large model enhanced drafting.arXiv preprint arXiv:2402.13720(2024). Received 2025-09-11; accepted 2025-12-22 Proc. ACM Softw. Eng., Vol. 3, No. FSE, Article FSE060. Publication date: July 2026

    Pith/arXiv arXiv 2024