pith. sign in

arxiv: 2607.01929 · v1 · pith:A4STEOHJnew · submitted 2026-07-02 · 💻 cs.SE

Beyond Textual Repository Exploration: Dual-Modal Structural Reasoning for Agentic Issue Resolution

Jiayi Zhang , Kai Huang , Yang Liu , Chunyang Chen This is my paper

Pith reviewed 2026-07-03 09:02 UTC · model grok-4.3

classification 💻 cs.SE
keywords DUALVIEWdual-modal reasoningrepository explorationissue resolutiongraph viewsSWE-benchvisual scaffoldingstructural dependencies
0
0 comments X

The pith

DUALVIEW improves agent issue resolution by giving persistent visual and textual graph views of code repositories.

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

The paper establishes that text-only navigation forces agents to rebuild repository structure from fragmented observations at every step, which produces drift in large long-horizon codebases. DUALVIEW supplies four persistent graph views—Module Coupling Graph, Function Call Graph, Class Hierarchy Graph, and Program Dependence Graph—through a queryable interface that returns both visual and textual responses. A sympathetic reader would care because the framework raises issue-resolution rates on SWE-bench Pro and Verified for multiple agent architectures and model families. Ablation results indicate that the visual externalization itself, not only the structural data, drives part of the gain.

Core claim

DUALVIEW represents repository structure through four complementary graph views: Module Coupling Graph (MCG), Function Call Graph (FCG), Class Hierarchy Graph (CHG), and Program Dependence Graph (PDG), and exposes them through a queryable interface with visual and textual responses. Rather than reconstructing repository structure from a sequence of textual observations, agents can directly reason over persistent visual representations of code dependencies, enabling more effective exploration and understanding of long-horizon codebases. Evaluation on SWE-bench Pro and Verified shows consistent improvements across agent architectures and model families, with ablation studies attributing gains

What carries the argument

DUALVIEW dual-modal scaffolding framework that maintains four repository graphs and returns both visual and textual answers to agent queries.

If this is right

  • Higher issue-resolution rates on SWE-bench Pro and Verified.
  • Consistent gains across different agent architectures and model families.
  • Performance lift arises from both textual structural data and visual externalization.
  • Visual graph views reduce exploration drift in long-horizon tasks.

Where Pith is reading between the lines

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

  • The same persistent visual graphs could be tested in other agent tasks such as test generation or multi-file refactoring.
  • One could measure whether particular graphs (for example PDG versus MCG) matter more for certain classes of bugs.
  • Persistent visuals might allow agents to operate effectively with shorter context windows by storing structure outside the prompt.
  • Training pipelines for code agents could incorporate the four graph views as additional observation channels.

Load-bearing premise

Visual externalization of repository dependencies better supports long-horizon exploration than textual observations alone.

What would settle it

An experiment in which agents receive only the textual forms of the four graphs and match the full DUALVIEW success rates on SWE-bench Pro and Verified.

Figures

Figures reproduced from arXiv: 2607.01929 by Chunyang Chen, Jiayi Zhang, Kai Huang, Yang Liu.

Figure 1
Figure 1. Figure 1: Text-centric vs. Graph-guided repository exploration. exploration dominates the agent’s workload. In practice, some refine and extend richer file-system and navigation primitives (e.g., purpose-built file viewers, search, and edit commands) to make code legible to the agent [1], [2]. Recent work such as mini-SWE-agent lets the model drive repository exploration through raw shell commands [15]. Another line… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of DUALVIEW. slice. The agent combines these dual-modal structural observa￾tions with native agent tools to navigate the codebase, inspect implementation details, and progressively localize and repair the bug. The remainder of this section first introduces the graph-view construction, then presents the dual-modal struc￾tural representations, and finally describes how DUALVIEW is integrated into ex… view at source ↗
Figure 3
Figure 3. Figure 3: Dual-modal observation of DUALVIEW. underlying repository. The textual representation records four categories of source-grounded information: • Entity Identity. Each graph node is mapped to its exact repository entity, allowing symbols with identical display names to be distinguished unambiguously. This information is particularly important for the FCG and CHG, where overloaded methods, inherited implement… view at source ↗
Figure 4
Figure 4. Figure 4: Adaptive reasoning procedure of DUALVIEW. that the behavior around GetSites may need to be understood both from its upstream use and from its downstream helper calls. The textual structural representation grounds the same structure in concrete source locations. It separates callers from callees, records each relation as a calls edge, and provides full paths and line numbers. Specifically for the three GetN… view at source ↗
Figure 5
Figure 5. Figure 5: Comparison with graph repository exploration methods. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: The case study of vuls-edb324. representations allow the model to identify dependency paths, branching structures, and highly connected regions without reconstructing them from lengthy textual descriptions. Conse￾quently, the agent can navigate the repository more efficiently and reason about long-range dependencies with fewer explo￾ration steps, mitigating Limitation ❷. However, visual representations alo… view at source ↗
Figure 8
Figure 8. Figure 8: The CHG case study of openlibrary-7bf323. and 75 instances respectively. These results show that each graph view provides useful structural evidence compared to DUALVIEWbase. While [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Heatmap of different variants of DUALVIEW invoking corresponding graph representation(s) frequency. work could further reduce potential variance by averaging results over multiple independent runs. Visual rendering. The effectiveness of visual reasoning may depend on how repository structures are rendered. To reduce this threat, all graph views are generated using deterministic Graphviz layouts with fixed … view at source ↗
read the original abstract

Recent advances in agentic program repair have significantly improved issue resolution by enabling iterative repository exploration. However, existing approaches predominantly rely on sequential, text-based code navigation, which fundamentally limits their ability to reason over large-scale long-horizon repositories with complex and long-range dependencies. As issue-resolution agents traverse repositories through fragmented textual observations, structural information such as module organization, call relationships, and dependency chains must be repeatedly reconstructed across interaction steps, often leading to exploration drift and incomplete localization. We present DUALVIEW, a dual-modal structural scaffolding framework that brings visual reasoning into repository exploration for issue-resolution agents. DUALVIEW represents repository structure through four complementary graph views: Module Coupling Graph (MCG), Function Call Graph (FCG), Class Hierarchy Graph (CHG), and Program Dependence Graph (PDG), and exposes them through a queryable interface with visual and textual responses. Rather than reconstructing repository structure from a sequence of textual observations, agents can directly reason over persistent visual representations of code dependencies, enabling more effective exploration and understanding of long-horizon codebases. We evaluate DUALVIEW on SWE-bench Pro and Verified. Results show that DUALVIEW consistently improves issue-resolution performance across different agent architectures and model families. Further ablation studies demonstrate that the gains arise not only from textual structural information but also from visual externalization of repository dependencies, which better supports long-horizon repository exploration.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The manuscript introduces DUALVIEW, a dual-modal structural scaffolding framework for agentic issue resolution. It represents repositories via four graph views (Module Coupling Graph (MCG), Function Call Graph (FCG), Class Hierarchy Graph (CHG), and Program Dependence Graph (PDG)) exposed through a queryable interface providing both visual and textual responses. The central claim is that this enables more effective long-horizon exploration than text-only approaches, yielding consistent improvements on SWE-bench Pro and Verified across agent architectures and model families, with ablation studies attributing the gains specifically to visual externalization of dependencies.

Significance. If the performance gains and their attribution to visual modality hold under controlled conditions, the work could advance agentic program repair by addressing exploration drift in large repositories through persistent visual representations. The evaluation on established SWE-bench benchmarks provides a clear point of comparison, which is a positive aspect of the experimental design.

major comments (1)
  1. [Abstract] Abstract (and experimental evaluation section): The claim that 'gains arise not only from textual structural information but also from visual externalization of repository dependencies' rests on ablation studies, but no details are supplied on whether these studies hold the underlying graph content (MCG/FCG/CHG/PDG) and query interface fixed when comparing text-only structural interfaces against the dual-modal version. Without such controls, measured improvements could arise from richer textual serialization, multi-turn context, or redundant formatting rather than visual reasoning per se. This directly undermines the attribution of benefits to the visual modality, which is load-bearing for the central claim.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on the ablation studies and their role in supporting the central claim. We address the single major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract (and experimental evaluation section): The claim that 'gains arise not only from textual structural information but also from visual externalization of repository dependencies' rests on ablation studies, but no details are supplied on whether these studies hold the underlying graph content (MCG/FCG/CHG/PDG) and query interface fixed when comparing text-only structural interfaces against the dual-modal version. Without such controls, measured improvements could arise from richer textual serialization, multi-turn context, or redundant formatting rather than visual reasoning per se. This directly undermines the attribution of benefits to the visual modality, which is load-bearing for the central claim.

    Authors: We agree that the current manuscript does not provide sufficient detail on the ablation controls, which weakens the attribution to the visual modality. In the revised manuscript we will expand the experimental evaluation section (and update the abstract accordingly) to explicitly state that the text-only structural baseline uses identical graph content (MCG, FCG, CHG, PDG) and the same queryable interface, differing solely in the absence of visual responses. We will also report additional controls (e.g., token-matched textual serializations and fixed multi-turn context lengths) to rule out alternative explanations. These clarifications will be added as a dedicated subsection with pseudocode and example outputs. revision: yes

Circularity Check

0 steps flagged

No circularity; empirical claims rest on external benchmarks

full rationale

The paper introduces DUALVIEW as a new dual-modal framework with four graph views and evaluates it empirically on SWE-bench Pro and Verified. Performance gains and ablation results are measured against these established external benchmarks rather than derived from internal definitions or self-citations. No equations, fitted parameters renamed as predictions, or load-bearing self-citations appear in the abstract or described chain; the central claim about visual externalization is supported by comparative experiments whose inputs (agent architectures, model families) are independent of the reported outcomes.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides no details on free parameters, axioms, or invented entities; the framework introduces new named graph views but their construction, query mechanisms, and any fitting choices are absent.

pith-pipeline@v0.9.1-grok · 5783 in / 1254 out tokens · 51447 ms · 2026-07-03T09:02:51.735201+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

49 extracted references · 14 canonical work pages · 7 internal anchors

  1. [1]

    Swe-agent: Agent-computer interfaces enable automated soft- ware engineering,

    J. Yang, C. E. Jimenez, A. Wettig, K. Lieret, S. Yao, K. Narasimhan, and O. Press, “Swe-agent: Agent-computer interfaces enable automated soft- ware engineering,”Advances in Neural Information Processing Systems, vol. 37, pp. 50 528–50 652, 2024

    2024

  2. [2]

    OpenHands: An Open Platform for AI Software Developers as Generalist Agents

    X. Wang, B. Li, Y . Song, F. F. Xu, X. Tang, M. Zhuge, J. Pan, Y . Song, B. Li, J. Singhet al., “Openhands: An open platform for ai software developers as generalist agents,”arXiv preprint arXiv:2407.16741, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  3. [3]

    ReAct: Synergizing Reasoning and Acting in Language Models

    S. Yao, J. Zhao, D. Yu, N. Du, I. Shafran, K. Narasimhan, and Y . Cao, “React: Synergizing reasoning and acting in language models,” 2023. [Online]. Available: https://arxiv.org/abs/2210.03629

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  4. [4]

    Autocoderover: Autonomous program improvement,

    Y . Zhang, H. Ruan, Z. Fan, and A. Roychoudhury, “Autocoderover: Autonomous program improvement,” inProceedings of the 33rd ACM SIGSOFT International Symposium on Software Testing and Analysis, 2024, pp. 1592–1604

    2024

  5. [5]

    Repairagent: An autonomous, llm-based agent for program repair,

    I. Bouzenia, P. Devanbu, and M. Pradel, “Repairagent: An autonomous, llm-based agent for program repair,” inIEEE/ACM 47th International Conference on Software Engineering (ICSE), 2025, pp. 2188–2200

    2025

  6. [6]

    Repair ingredients are all you need: Improving large language model-based program repair via repair ingredients search,

    J. Zhang, K. Huang, J. Zhang, Y . Liu, and C. Chen, “Repair ingredients are all you need: Improving large language model-based program repair via repair ingredients search,”arXiv preprint arXiv:2506.23100, 2025

    work page arXiv 2025

  7. [7]

    Live-SWE-agent: Can software engineering agents self-evolve on the fly?

    C. S. Xia, Z. Wang, Y . Yang, Y . Wei, and L. Zhang, “Live-swe-agent: Can software engineering agents self-evolve on the fly?”arXiv preprint arXiv:2511.13646, 2025

    work page arXiv 2025

  8. [8]

    Trae agent: An llm-based agent for software engineering with test-time scaling,

    T. R. Team, P. Gao, Z. Tian, X. Meng, X. Wang, R. Hu, Y . Xiao, Y . Liu, Z. Zhang, J. Chen, C. Gao, Y . Lin, Y . Xiong, C. Peng, and X. Liu, “Trae agent: An llm-based agent for software engineering with test-time scaling,” 2025. [Online]. Available: https://arxiv.org/abs/2507.23370

    work page arXiv 2025

  9. [9]

    Demystifying llm-based software engineering agents,

    C. S. Xia, Y . Deng, S. Dunn, and L. Zhang, “Demystifying llm-based software engineering agents,”Proceedings of the ACM on Software Engineering, vol. 2, no. FSE, pp. 801–824, 2025

    2025

  10. [10]

    Swe-bench: Can language models resolve real-world github issues?

    C. E. Jimenez, J. Yang, A. Wettig, S. Yao, K. Pei, O. Press, and K. Narasimhan, “Swe-bench: Can language models resolve real-world github issues?” in12th International Conference on Learning Represen- tations, ICLR 2024, 2024

    2024

  11. [11]

    SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?

    X. Deng, J. Da, E. Pan, Y . Y . He, C. Ide, K. Garg, N. Lauffer, A. Park, N. Pasari, C. Raneet al., “Swe-bench pro: Can ai agents solve long- horizon software engineering tasks?”arXiv preprint arXiv:2509.16941, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  12. [12]

    Swe-bench multimodal: Do ai systems generalize to visual software domains?

    J. Yang, C. E. Jimenez, A. L. Zhang, K. Lieret, J. Yang, X. Wu, O. Press, N. Muennighoff, G. Synnaeve, K. R. Narasimhanet al., “Swe-bench multimodal: Do ai systems generalize to visual software domains?” in The Thirteenth International Conference on Learning Representations

  13. [13]

    Swe-bench verified,

    C. E. Jimenez, J. Yang, A. Wettig, S. Yao, K. Pei, O. Press, and K. R. Narasimhan, “Swe-bench verified,” https://www.swebench.com/verified. html, 2025

    2025

  14. [14]

    SWE-Pruner: Self-Adaptive Context Pruning for Coding Agents

    Y . Wang, Y . Shi, M. Yang, R. Zhang, S. He, H. Lian, Y . Chen, S. Ye, K. Cai, and X. Gu, “Swe-pruner: Self-adaptive context pruning for coding agents,”arXiv preprint arXiv:2601.16746, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  15. [15]

    mini-swe-agent: The minimal ai software engineering agent,

    J. Yang, C. E. Jimenez, A. Wettig, K. Lieret, S. Yao, K. R. Narasimhan, and O. Press, “mini-swe-agent: The minimal ai software engineering agent,” https://github.com/SWE-agent/mini-swe-agent, 2026

    2026

  16. [16]

    Repograph: Enhancing ai software engineer- ing with repository-level code graph,

    S. Ouyang, W. Yu, K. Ma, Z. Xiao, Z. Zhang, M. Jia, J. Han, H. Zhang, and D. Yu, “Repograph: Enhancing ai software engineer- ing with repository-level code graph,” inInternational Conference on Learning Representations, vol. 2025, 2025, pp. 30 098–30 121

    2025

  17. [17]

    codegraph: Pre-indexed code knowledge graph for claude code, codex, cursor, opencode, and hermes agent,

    C. McHenryet al., “codegraph: Pre-indexed code knowledge graph for claude code, codex, cursor, opencode, and hermes agent,” 2026. [Online]. Available: https://github.com/colbymchenry/codegraph

    2026

  18. [18]

    Prometheus: Towards long-horizon codebase navigation for repository-level problem solving,

    Y . Pan, Z. Chen, S. Lu, Z. Chu, X. Li, H. Li, Y . Feng, C. Le Goues, F. Sarro, M. Monperruset al., “Prometheus: Towards long-horizon codebase navigation for repository-level problem solving,”arXiv e- prints, pp. arXiv–2507, 2025

    2025

  19. [19]

    ARISE: A Repository-level Graph Representation and Toolset for Agentic Fault Localization and Program Repair

    S. Seddik and F. Fard, “Arise: A repository-level graph representation and toolset for agentic fault localization and program repair,”arXiv preprint arXiv:2605.03117, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  20. [20]

    DeepSeek-OCR: Contexts Optical Compression

    H. Wei, Y . Sun, and Y . Li, “Deepseek-ocr: Contexts optical compres- sion,”arXiv preprint arXiv:2510.18234, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  21. [21]

    Kimi k2.5: Visual agentic intelligence,

    MoonshotAI, “Kimi k2.5: Visual agentic intelligence,” Tech. Rep.,

  22. [22]

    Available: https://www.kimi.com/blog/kimi-k2-5

    [Online]. Available: https://www.kimi.com/blog/kimi-k2-5

  23. [23]

    Introducing agentic vision in gemini 3 flash,

    DeepMind, “Introducing agentic vision in gemini 3 flash,” Tech. Rep., 2026. [Online]. Available: https://blog.google/innovation-and-ai/ technology/developers-tools/agentic-vision-gemini-3-flash/

    2026

  24. [24]

    Dualview website,

    DualView, “Dualview website,” 2026. [Online]. Available: https: //sites.google.com/view/dual-view

    2026

  25. [25]

    Dualview mcp service,

    ——, “Dualview mcp service,” 2026. [Online]. Available: https: //github.com/AutoVisualCoder/dualview

    2026

  26. [26]

    A systematic literature review on large language models for auto- mated program repair,

    Q. Zhang, C. Fang, Y . Xie, Y . Ma, W. Sun, Y . Yang, and Z. Chen, “A systematic literature review on large language models for auto- mated program repair,”ACM Transactions on Software Engineering and Methodology, 2024

    2024

  27. [27]

    Agentic software issue resolution with large language models: A survey,

    Z. Jiang, D. Lo, and Z. Liu, “Agentic software issue resolution with large language models: A survey,”arXiv preprint arXiv:2512.22256, 2025

    work page arXiv 2025

  28. [28]

    Using automatic clustering to produce high-level system organizations of source code,

    S. Mancoridis, B. S. Mitchell, C. Rorres, Y . Chen, and E. R. Gansner, “Using automatic clustering to produce high-level system organizations of source code,” inProceedings. 6th International Workshop on Program Comprehension, 1998, pp. 45–52

    1998

  29. [29]

    On the criteria to be used in decomposing systems into modules,

    D. L. Parnas, “On the criteria to be used in decomposing systems into modules,”Communications of the ACM, vol. 15, no. 12, pp. 1053–1058, 1972

    1972

  30. [30]

    The structure and value of modularity in software design,

    K. J. Sullivan, W. G. Griswold, Y . Cai, and B. Hallen, “The structure and value of modularity in software design,”ACM SIGSOFT Software Engineering Notes, vol. 26, no. 5, pp. 99–108, 2001

    2001

  31. [31]

    How effective developers investigate source code: an exploratory study,

    M. Robillard, W. Coelho, and G. Murphy, “How effective developers investigate source code: an exploratory study,”IEEE Transactions on Software Engineering, vol. 30, no. 12, pp. 889–903, 2004

    2004

  32. [32]

    Call graph con- struction in object-oriented languages,

    D. Grove, G. DeFouw, J. Dean, and C. Chambers, “Call graph con- struction in object-oriented languages,” inProceedings of the 12th ACM SIGPLAN conference on Object-oriented programming, systems, languages, and applications, 1997, pp. 108–124

    1997

  33. [33]

    Optimization of object-oriented programs using static class hierarchy analysis,

    J. Dean, D. Grove, and C. Chambers, “Optimization of object-oriented programs using static class hierarchy analysis,” inEuropean conference on object-oriented programming, 1995, pp. 77–101

    1995

  34. [34]

    The program dependence graph and its use in optimization,

    J. Ferrante, K. J. Ottenstein, and J. D. Warren, “The program dependence graph and its use in optimization,”ACM Transactions on Programming Languages and Systems (TOPLAS), vol. 9, no. 3, pp. 319–349, 1987

    1987

  35. [35]

    Is a picture worth a thousand words? delving into spatial reasoning for vision language models,

    J. Wang, Y . Ming, Z. Shi, V . Vineet, X. Wang, Y . Li, and N. Joshi, “Is a picture worth a thousand words? delving into spatial reasoning for vision language models,”Advances in Neural Information Processing Systems, vol. 37, pp. 75 392–75 421, 2024

    2024

  36. [36]

    Visual sketchpad: Sketching as a visual chain of thought for multimodal language models,

    Y . Hu, W. Shi, X. Fu, D. Roth, M. Ostendorf, L. Zettlemoyer, N. A. Smith, and R. Krishna, “Visual sketchpad: Sketching as a visual chain of thought for multimodal language models,”Advances in Neural Information Processing Systems, vol. 37, pp. 139 348–139 379, 2024

    2024

  37. [37]

    OpenCode: The open source AI coding agent,

    Anomaly, “OpenCode: The open source AI coding agent,” 2026. [Online]. Available: https://opencode.ai/

    2026

  38. [38]

    Claude sonnet 4.5,

    Anthropic, “Claude sonnet 4.5,” https://www.anthropic.com/news/ claude-sonnet-4-5, 2026

    2026

  39. [39]

    Kimi k2.5,

    Moonshot, “Kimi k2.5,” https://www.kimi.com/ai-models/kimi-k2-5, 2026

    2026

  40. [40]

    Gemini 3 flash,

    Google, “Gemini 3 flash,” https://docs.cloud.google.com/ gemini-enterprise-agent-platform/models/gemini/3-flash, 2026

    2026

  41. [41]

    Swe-bench pro leaderboard ai coding benchmark (public dataset),

    S. AI, “Swe-bench pro leaderboard ai coding benchmark (public dataset),” Tech. Rep., 2026. [Online]. Available: https://labs.scale.com/ leaderboard/swe bench pro public

    2026

  42. [42]

    Issue localization via llm-driven iterative code graph searching,

    Z. Jiang, X. Ren, M. Yan, W. Jiang, Y . Li, and Z. Liu, “Issue localization via llm-driven iterative code graph searching,” in2025 40th IEEE/ACM International Conference on Automated Software Engineering (ASE), 2025, pp. 3034–3045

    2025

  43. [43]

    Enhancing repository-level software repair via repository-aware knowledge graphs,

    B. Yang, J. Ren, S. Jin, Y . Liu, F. Liu, B. Le, and H. Tian, “Enhancing repository-level software repair via repository-aware knowledge graphs,” arXiv preprint arXiv:2503.21710, 2025

    work page arXiv 2025

  44. [44]

    Sgagent: Suggestion-guided llm-based multi-agent framework for repository-level software repair,

    Q. Zhang, C. Gao, Y . Han, Y . Shang, C. Fang, Z. Chen, and L. Xiao, “Sgagent: Suggestion-guided llm-based multi-agent framework for repository-level software repair,”ACM Transactions on Software Engineering and Methodology, 2026

    2026

  45. [45]

    Seeing is fixing: Cross- modal reasoning with multimodal llms for visual software issue repair,

    K. Huang, J. Zhang, X. Xie, and C. Chen, “Seeing is fixing: Cross- modal reasoning with multimodal llms for visual software issue repair,” in40th IEEE/ACM International Conference on Automated Software Engineering (ASE), 2025, pp. 1156–1168

    2025

  46. [46]

    Svrepair: Structured visual reasoning for automated program repair,

    X. Tang, J. Wang, L. Luo, J. Xu, S. Zhou, D. Chen, W. Jiang, and Y . Li, “Svrepair: Structured visual reasoning for automated program repair,” arXiv preprint arXiv:2602.06090, 2026

    work page arXiv 2026

  47. [47]

    Longcodezip: Compress long context for code language models,

    Y . Shi, Y . Qian, H. Zhang, B. Shen, and X. Gu, “Longcodezip: Compress long context for code language models,” in40th IEEE/ACM International Conference on Automated Software Engineering (ASE), 2025, pp. 141–153

    2025

  48. [48]

    CodeOCR: On the Effectiveness of Vision Language Models in Code Understanding

    Y . Shi, C. Xie, Z. Sun, Y . Chen, C. Zhang, L. Yun, C. Wan, H. Zhang, D. Lo, and X. Gu, “Codeocr: On the effectiveness of vision language models in code understanding,”arXiv preprint arXiv:2602.01785, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  49. [49]

    VEA and Baselines Implementation Details

    J. Zhong, G. Li, C. Zhi, J. Han, Z. Qin, X. Zhao, N. Wang, S. Deng, and J. Yin, “Can vision-language models handle long-context code? an em- pirical study on visual compression,”arXiv preprint arXiv:2602.00746, 2026

    work page arXiv 2026