Tables & Resources: Knowledge Explorer for Automated Issue Resolution¶

This page is the public query hub for the reviewed corpus. Instead of mirroring the survey chapter-by-chapter, it is organized around four reader tasks: understanding the field quickly, finding papers efficiently, comparing methods meaningfully, and locating reusable resources.

Quick Insights¶

These summary rows are intended to answer the most common first-pass questions a reader has when arriving at the survey site.

Question	Answer	Evidence
Most dominant knowledge type	Existing Code Knowledge remains the universal core of current issue-resolution systems.	128 / 133 papers
Fastest-rising procedural factor	Development Tool Knowledge and execution-grounded interaction increasingly dominate post-2025 systems.	49 papers in latest-half slice
Main shift in recent work	The field is moving from static code retrieval toward feedback-driven, process-aware, and history-aware agent workflows.	Development Tool Knowledge + Existing Code Knowledge is the top co-occurrence
Most underexplored knowledge	Domain, language, and architecture knowledge remain sparse despite their importance in real-world repair settings.	Each remains below 10% of the included corpus
Workflow hotspots	Localization, patch generation, and patch validation absorb the largest share of knowledge-intensive systems.	These stages dominate the corpus-wide usage-stage distribution

Explore the Field¶

Knowledge Coverage Summary¶

This table compresses the survey's core coding dimensions into a single field map, showing how each knowledge type is typically sourced, represented, used, and applied.

Knowledge Type	Share	Main Source	Main Stage	Common Representation	Common Usage	Trend	Key Papers
Existing Code Knowledge	96.2%	Codebase	Localization	Unstructured Text	Tool Invocation	Stable core	GraphLocator: Graph-guided Causal Reasoning for Issue Localization Agentic Rubrics as Contextual Verifiers for SWE Agents Dars: Dynamic action re-sampling to enhance coding agent performance by adaptive tree traversal
Development Tool Knowledge	77.4%	Codebase	Localization	Unstructured Text	Tool Invocation	Rising	Otter: Generating tests from issues to validate swe patches SWE-Replay: Efficient Test-Time Scaling for Software Engineering Agents
Experiential Knowledge	51.1%	Dynamic Execution Results	Localization	Unstructured Text	Tool Invocation	Rising fast	Swe-exp: Experience-driven software issue resolution SWE-Master: Unleashing the Potential of Software Engineering Agents via Post-Training
Repository Evolution Knowledge	32.3%	Codebase	Localization	Unstructured Text	Direct Prompt	Rising	BLAZE: Cross-language and cross-project bug localization via dynamic chunking and hard example learning Closing the Loop: Universal Repository Representation with RPG-Encoder SIADAFIX: issue description response for adaptive program repair
Design Architecture Knowledge	9.8%	Codebase	Patch Generation	Unstructured Text	Direct Prompt	Emerging	BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? Architecture-Aware Multi-Design Generation for Repository-Level Feature Addition
Technology Stack Knowledge	3.0%	Codebase	Localization	Unstructured Text	Tool Invocation	Sparse but important	BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? ArkEval: Benchmarking and Evaluating Automated CodeRepair for ArkTS
Programming Language Knowledge	2.3%	Codebase	Localization	Unstructured Text	Direct Prompt	Sparse but important	Evaluating and Improving Automated Repository-Level Rust Issue Resolution with LLM-based Agents ArkEval: Benchmarking and Evaluating Automated CodeRepair for ArkTS
Domain-specific Knowledge	0.8%	Codebase	Environment Building	Unstructured Text	Direct Prompt	Highly underexplored	BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing?

Underexplored Opportunities¶

These rows translate low-coverage knowledge dimensions into concrete research opportunities rather than leaving them as isolated counts.

Gap	Current Coverage	Why It Matters	Candidate Papers	Opportunity
Domain-specific knowledge	0.8% of papers	Real-world issue resolution often depends on domain semantics that generic repository reasoning cannot recover from code alone.	BeyondSWE	High
Programming-language knowledge	2.3% of papers	Multilingual and low-resource repair settings need stronger language-semantic grounding than current Python-heavy pipelines provide.	ArkEval Rust issue resolution SWE-Rank+	High
Design / architecture knowledge	9.8% of papers	Passing tests does not guarantee alignment with repository-wide constraints, module responsibilities, or intended architecture.	RPG-Encoder Agentic Rubrics ToM-SWE	Very high
Document-grounded reasoning	11 document-linked papers	Specifications, migration notes, and framework guidance are increasingly necessary in non-trivial issue settings.	BeyondSWE ArkEval Agentic Rubrics	Medium-high

Reading Entry Table¶

If a reader enters the field from a specific angle, this table gives a better starting path than a raw tag directory.

If You Care About...	Start With	Why
Execution feedback	Swe-agent: Agent-computer interfaces enable automated software engineering R2e-gym: Procedural environments and hybrid verifiers for scaling open-weights swe agents Otter: Generating tests from issues to validate swe patches	These systems make tool interaction, runtime evidence, and executable validation central to the repair loop.
Memory / experience	EXPEREPAIR: Dual-Memory Enhanced LLM-based Repository-Level Program Repair Swe-exp: Experience-driven software issue resolution Reasoningbank: Scaling agent self-evolving with reasoning memory	They show how prior trajectories, reflections, and reusable strategy memory can guide future issue resolution.
Repository history	KGCompass: Knowledge Graph Enhanced Repository-Level Software Repair SweRank: Software Issue Localization with Code Ranking Improving Code Localization with Repository Memory	These works foreground commits, issues, historical fixes, and repository memory as explicit retrieval signals.
Architecture-aware repair	Closing the Loop: Universal Repository Representation with RPG-Encoder Agentic Rubrics as Contextual Verifiers for SWE Agents Tom-swe: User mental modeling for software engineering agents	They expose how design intent and global repository constraints enter planning, patching, and validation.
Benchmark / training resources	SWE-Factory: Your Automated Factory for Issue Resolution Training Data and Evaluation Benchmarks Training software engineering agents and verifiers with swe-gym R2e-gym: Procedural environments and hybrid verifiers for scaling open-weights swe agents	These papers are the best entry points if the goal is to build training environments, evaluation infrastructure, or reusable baselines.

Find Papers¶

Knowledge Taxonomy Explorer¶

The taxonomy explorer remains as a fast navigation entry point, but its role here is to drive filtering rather than duplicate the whole survey narrative.

L1 Background Knowledge¶

Programming Language Knowledge¶

3 papers / 2.3%

This category captures language-level constraints that a repository-grounded agent cannot ignore, including syntax, type systems, runtime semantics, and idiomatic repair patterns. In the survey corpus, it appears most clearly when issue resolution must generalize across languages or obey language-specific rules such as Rust ownership or ArkTS compilation constraints.

Localization (3)

Patch Generation (2)

Patch Validation (2)

Reproduction Test Generation (2)

View papers

Domain-specific Knowledge¶

1 papers / 0.8%

This category covers application semantics that live outside generic software engineering knowledge, such as scientific concepts, platform-specific business rules, or field-specific constraints. Although rare in the current corpus, it becomes essential once issue resolution moves beyond standard single-repository bug fixing into tasks where correct patches depend on understanding the target domain itself.

Environment Building (1)

Technology Stack Knowledge¶

4 papers / 3.0%

This category refers to libraries, frameworks, APIs, platform conventions, and dependency ecosystems that shape what counts as a compatible repair. It is especially important in migration, compatibility, and platform-constrained tasks, where an agent must reason about external components rather than only the repository's local implementation.

Localization (4)

Patch Generation (4)

Patch Validation (4)

Environment Building (3)

View papers

L2 Repository Knowledge¶

Existing Code Knowledge¶

128 papers / 96.2%

This is the dominant knowledge type in the survey and forms the core of most issue-resolution systems. It includes concrete repository-native artifacts such as files, classes, functions, call sites, dependency relations, and local implementation context that agents rely on for localization, editing, and validation.

Localization (116)

Patch Generation (100)

Patch Validation (88)

Model Training (50)

View papers

Design Architecture Knowledge¶

13 papers / 9.8%

This category captures higher-level repository intent: architectural decomposition, component responsibilities, design constraints, and cross-module coordination logic. It becomes important when systems must produce patches that preserve global structure, satisfy implicit project conventions, or align edits with intended repository organization rather than only local code correctness.

Patch Generation (11)

Localization (10)

Patch Validation (10)

Task Planning (9)

View papers

Repository Evolution Knowledge¶

43 papers / 32.3%

This category covers the repository's historical memory, including issues, pull requests, commits, prior patches, and longer development-process traces. It allows systems to retrieve precedents, infer repair patterns, and ground current decisions in how the project has changed over time.

Localization (38)

Patch Generation (31)

Patch Validation (30)

Model Training (29)

View papers

L3 Procedural Knowledge¶

Development Tool Knowledge¶

103 papers / 77.4%

This category describes operational knowledge about tools, interfaces, environments, and execution feedback channels used during issue resolution. It includes how agents search, run tests, interact with shells and sandboxes, interpret validation results, and orchestrate tool-assisted workflows across the repair loop.

Localization (92)

Patch Validation (86)

Patch Generation (85)

Reproduction Test Generation (42)

View papers

Experiential Knowledge¶

68 papers / 51.1%

This category focuses on reusable experience distilled from prior trajectories, failures, reflections, and successful repair behaviors. In the survey, it appears both as explicit memory structures and as training-time data that teaches agents how to plan, retrieve, edit, and validate more effectively over time.

Localization (61)

Patch Generation (56)

Patch Validation (52)

Model Training (36)

View papers

Preset Filters¶

These presets expose higher-level reading intentions directly, so readers do not need to build every query manually from raw dropdowns.

Multilingual / low-resource

Agent training / RL

Master Paper Table¶

The controls below support shareable prefilters through the page URL, so links from Home or preset cards can open this table already narrowed to a knowledge type, source, stage, or method family.

Code = Existing Code Knowledge

Tool = Development Tool Knowledge

History = Repository Evolution Knowledge

Memory = Experiential Knowledge

Arch = Design Architecture Knowledge

Lang = Programming Language Knowledge

Stack = Technology Stack Knowledge

Domain = Domain-specific Knowledge

Showing 133 of 133 papers.

Paper	Year	Venue	Focus	Knowledge Profile	Method Style	Benchmark / Setting	Link
BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? `chen2026beyondswe` · Chen, Guoxin; Meng, Fanzhe; Zhao, Jiale; Li, Minghao; Cheng, Daixuan; Song, Huatong; Chen, Jie; Lin, Yuzhi; Chen, Hui; Zhao, Xin; others	2026	arXiv preprint arXiv:2603.03194	Validation / testing	Arch Tool Domain Code	Tool-centric	BeyondSWE	arXiv:2603.03194
SWE-CI: Evaluating Agent Capabilities in Maintaining Codebases via Continuous Integration `chen2026swe` · Chen, Jialong; Xu, Xander; Wei, Hu; Chen, Chuan; Zhao, Bing	2026	arXiv preprint arXiv:2603.03823	Benchmark / infrastructure		Agentic	—	arXiv:2603.03823
SWE-Replay: Efficient Test-Time Scaling for Software Engineering Agents `ding2026swe` · Ding, Yifeng; Zhang, Lingming	2026	arXiv preprint arXiv:2601.22129	Validation / testing	Tool Code Memory	Tool-centric	—	arXiv:2601.22129
SWE-Prot$\backslash$'eg$\backslash$'e: Learning to Selectively Collaborate With an Expert Unlocks Small Language Models as Software Engineering Agents `kon2026swe` · Kon, Patrick Tser Jern; Pradeep, Archana; Chen, Ang; Ellis, Alexander P; Hunt, Warren; Wang, Zijian; Yang, John; Thompson, Samuel	2026	arXiv preprint arXiv:2602.22124	Training / post-training	Tool Code Memory	Training-based / RL	—	arXiv:2602.22124
Architecture-Aware Multi-Design Generation for Repository-Level Feature Addition `liu2026architecture` · Liu, Mingwei; Chen, Zhenxi; Pei, Zheng; Wang, Zihao; Wang, Yanlin; Zheng, Zibin	2026	arXiv preprint arXiv:2603.01814	Architecture / constraint-aware	Arch Tool Code	Graph-based	—	arXiv:2603.01814
Dynamic analysis enhances issue resolution `liu2026dynamic` · Liu, Mingwei; Wang, Zihao; Chen, Zhenxi; Pei, Zheng; Wang, Yanlin; Zheng, Zibin	2026	arXiv preprint arXiv:2603.22048	Validation / testing	Tool Code	Tool-centric	—	arXiv:2603.22048
Closing the Loop: Universal Repository Representation with RPG-Encoder `luo2026closing` · Luo, Jane; Yin, Chengyu; Zhang, Xin; Li, Qingtao; Liu, Steven; Huang, Yiming; Wu, Jie; Liu, Hao; Huang, Yangyu; Kang, Yu; others	2026	arXiv preprint arXiv:2602.02084	History-aware reasoning	Arch Tool Code History	Graph-based	—	arXiv:2602.02084
Agentic Rubrics as Contextual Verifiers for SWE Agents `raghavendra2026agentic` · Raghavendra, Mohit; Gunjal, Anisha; Liu, Bing; He, Yunzhong	2026	arXiv preprint arXiv:2601.04171	Training / post-training	Arch Tool Code	Training-based / RL	—	arXiv:2601.04171
SWE-Master: Unleashing the Potential of Software Engineering Agents via Post-Training `song2026swe` · Song, Huatong; Huang, Lisheng; Sun, Shuang; Jiang, Jinhao; Le, Ran; Cheng, Daixuan; Chen, Guoxin; Hu, Yiwen; Chen, Zongchao; Jia, Yiming; others	2026	arXiv preprint arXiv:2602.03411	Training / post-training	Tool Code Memory	Training-based / RL	—	arXiv:2602.03411
SWE-Tester: Training Open-Source LLMs for Issue Reproduction in Real-World Repositories `soni2026swe` · Soni, Aditya Bharat; Ghosh, Rajat; Bhargava, Vaishnavi; Chen, Valerie; Dutta, Debojyoti	2026	arXiv preprint arXiv:2601.13713	History-aware reasoning	Tool Code History	Training-based / RL	—	arXiv:2601.13713
Swe-lego: Pushing the limits of supervised fine-tuning for software issue resolving `tao2026swe` · Tao, Chaofan; Chen, Jierun; Jiang, Yuxin; Kou, Kaiqi; Wang, Shaowei; Wang, Ruoyu; Li, Xiaohui; Yang, Sidi; Du, Yiming; Dai, Jianbo; others	2026	arXiv preprint arXiv:2601.01426	History-aware reasoning	Tool Code Memory History	Training-based / RL	—	arXiv:2601.01426
Empowering Autonomous Debugging Agents with Efficient Dynamic Analysis `xiang2026empowering` · Xiang, Jiahong; Xu, Xiaoyang; Chu, Xiaopan; Tian, Hongliang; Zhang, Yuqun	2026	arXiv preprint arXiv:2604.24212	Localization	Tool Code	Tool-centric	—	arXiv:2604.24212
Evaluating and Improving Automated Repository-Level Rust Issue Resolution with LLM-based Agents `xiang2026evaluating` · Xiang, Jiahong; He, Wenxiao; Wang, Xihua; Tian, Hongliang; Zhang, Yuqun	2026	arXiv preprint arXiv:2602.22764	Benchmark / infrastructure	Tool Code Lang	Tool-centric	Language-specific	arXiv:2602.22764
ArkEval: Benchmarking and Evaluating Automated CodeRepair for ArkTS `xie2026arkeval` · Xie, Bang; Zhang, Senjian; Peng, Zhiyuan; Chen, Wei; Ying, Chenhao; Luo, Yuan	2026	arXiv preprint arXiv:2602.08866	Benchmark / infrastructure	Tool Code Lang History	Tool-centric	ArkTS / HarmonyOS	arXiv:2602.08866
Does Pass Rate Tell the Whole Story? Evaluating Design Constraint Compliance in LLM-based Issue Resolution `yu2026does` · Yu, Kai; Zhou, Zhenhao; Zeng, Junhao; Wang, Ying; Du, Xueying; Yuan, Zhiqiang; Liu, Junwei; Zhou, Ziyu; Wang, Yujia; Wang, Chong; others	2026	arXiv preprint arXiv:2604.05955	Benchmark / infrastructure	Arch Code History	Pipeline	—	arXiv:2604.05955
SWE-MiniSandbox: Container-Free Reinforcement Learning for Building Software Engineering Agents `yuan2026swe` · Yuan, Danlong; Wu, Wei; Wang, Zhengren; Zhao, Xueliang; Zhang, Huishuai; Zhao, Dongyan	2026	arXiv preprint arXiv:2602.11210	Training / post-training	Tool Code Memory	Training-based / RL	Evaluation / infra	arXiv:2602.11210
Dars: Dynamic action re-sampling to enhance coding agent performance by adaptive tree traversal `aggarwal2025dars` · Aggarwal, Vaibhav; Kamal, Ojasv; Japesh, Abhinav; Jin, Zhijing; Sch{\"o}lkopf, Bernhard	2025	arXiv preprint arXiv:2503.14269	Training / post-training	Tool Code Memory	Graph-based	—	arXiv:2503.14269
Otter: Generating tests from issues to validate swe patches `ahmed2025otter` · Ahmed, Toufique; Ganhotra, Jatin; Pan, Rangeet; Shinnar, Avraham; Sinha, Saurabh; Hirzel, Martin	2025	arXiv preprint arXiv:2502.05368	History-aware reasoning	Tool Code History	Tool-centric	—	arXiv:2502.05368
SIADAFIX: issue description response for adaptive program repair `cao2025siadafix` · Cao, Xin; Yu, Nan	2025	arXiv preprint arXiv:2510.16059	History-aware reasoning	Arch Tool Code Memory	Training-based / RL	—	arXiv:2510.16059
BLAZE: Cross-language and cross-project bug localization via dynamic chunking and hard example learning `chakraborty2025blaze` · Chakraborty, Partha; Alfadel, Mahmoud; Nagappan, Meiyappan	2025	IEEE Transactions on Software Engineering	History-aware reasoning	Code History	Graph-based	—	Scholar
Bridging Bug Localization and Issue Fixing: A Hierarchical Localization Framework Leveraging Large Language Models `chang2025bridging` · Chang, Jianming; Zhou, Xin; Wang, Lulu; Lo, David; Li, Bixin	2025	arXiv preprint arXiv:2502.15292	History-aware reasoning	Code History	Graph-based	—	arXiv:2502.15292
Locagent: Graph-guided llm agents for code localization `chen2025locagent` · Chen, Zhaoling; Tang, Xiangru; Deng, Gangda; Wu, Fang; Wu, Jialong; Jiang, Zhiwei; Prasanna, Viktor; Cohan, Arman; Wang, Xingyao	2025	arXiv preprint arXiv:2503.09089	Training / post-training	Code Memory	Graph-based	—	arXiv:2503.09089
When Old Meets New: Evaluating the Impact of Regression Tests on SWE Issue Resolution `chen2025old` · Chen, Yang; Ahmed, Toufique; Jabbarvand, Reyhaneh; Hirzel, Martin	2025	arXiv preprint arXiv:2510.18270	Benchmark / infrastructure	Tool Code	Tool-centric	—	arXiv:2510.18270
Prometheus: Unified knowledge graphs for issue resolution in multilingual codebases `chen2025prometheus` · Chen, Zimin; Pan, Yue; Lu, Siyu; Xu, Jiayi; Goues, Claire Le; Monperrus, Martin; Ye, He	2025	arXiv preprint arXiv:2507.19942	Validation / testing	Tool Code	Graph-based	Language-specific	arXiv:2507.19942
Swe-exp: Experience-driven software issue resolution `chen2025swe` · Chen, Silin; Lin, Shaoxin; Gu, Xiaodong; Shi, Yuling; Lian, Heng; Yun, Longfei; Chen, Dong; Sun, Weiguo; Cao, Lin; Wang, Qianxiang	2025	arXiv preprint arXiv:2507.23361	Memory / experience	Tool Code Memory	Tool-centric	—	arXiv:2507.23361
Cwm: An open-weights llm for research on code generation with world models `copet2025cwm` · Copet, Jade; Carbonneaux, Quentin; Cohen, Gal; Gehring, Jonas; Kahn, Jacob; Kossen, Jannik; Kreuk, Felix; McMilin, Emily; Meyer, Michel; Wei, Yuxiang; others	2025	arXiv preprint arXiv:2510.02387	History-aware reasoning	Tool Code Memory History	Training-based / RL	—	arXiv:2510.02387
Agent-RLVR: Training Software Engineering Agents via Guidance and Environment Rewards `da2025agent` · Da, Jeff; Wang, Clinton; Deng, Xiang; Ma, Yuntao; Barhate, Nikhil; Hendryx, Sean	2025	arXiv preprint arXiv:2506.11425	Training / post-training	Tool Code Memory	Training-based / RL	—	arXiv:2506.11425
Lita: Light Agent Uncovers the Agentic Coding Capabilities of LLMs `dai2025lita` · Dai, Hankun; Wang, Maoquan; Qi, Mengnan; Zhang, Yikai; Jin, Zijian; Yao, Yongqiang; Huang, Yufan; Fu, Shengyu; Nallipogu, Elsie	2025	arXiv preprint arXiv:2509.25873	Validation / testing	Tool Code	Tool-centric	—	arXiv:2509.25873
Codemonkeys: Scaling test-time compute for software engineering `ehrlich2025codemonkeys` · Ehrlich, Ryan; Brown, Bradley; Juravsky, Jordan; Clark, Ronald; R{\'e}, Christopher; Mirhoseini, Azalia	2025	arXiv preprint arXiv:2501.14723	Validation / testing	Tool Code	Tool-centric	—	arXiv:2501.14723
When agents go astray: Course-correcting swe agents with prms `gandhi2025agents` · Gandhi, Shubham; Tsay, Jason; Ganhotra, Jatin; Kate, Kiran; Rizk, Yara	2025	arXiv preprint arXiv:2509.02360	Validation / testing	Tool Code Memory	Tool-centric	—	arXiv:2509.02360
Trae agent: An llm-based agent for software engineering with test-time scaling `gao2025trae` · Gao, Pengfei; Tian, Zhao; Meng, Xiangxin; Wang, Xinchen; Hu, Ruida; Xiao, Yuanan; Liu, Yizhou; Zhang, Zhao; Chen, Junjie; Gao, Cuiyun; others	2025	arXiv preprint arXiv:2507.23370	Validation / testing	Tool Code Memory	Tool-centric	—	arXiv:2507.23370
Training long-context, multi-turn software engineering agents with reinforcement learning `golubev2025training` · Golubev, Alexander; Trofimova, Maria; Polezhaev, Sergei; Badertdinov, Ibragim; Nekrashevich, Maksim; Shevtsov, Anton; Karasik, Simon; Abramov, Sergey; Andriushchenko, Andrei; Fisin, Filipp; others	2025	arXiv preprint arXiv:2508.03501	Training / post-training	Tool Code Memory	Training-based / RL	—	arXiv:2508.03501
SWE-Factory: Your Automated Factory for Issue Resolution Training Data and Evaluation Benchmarks `guo2025swe` · Guo, Lianghong; Wang, Yanlin; Li, Caihua; Yang, Pengyu; Chen, Jiachi; Tao, Wei; Zou, Yingtian; Tang, Duyu; Zheng, Zibin	2025	arXiv preprint arXiv:2506.10954	Benchmark / infrastructure	Tool Code History	Training-based / RL	Evaluation / infra	arXiv:2506.10954
SACL: Understanding and Combating Textual Bias in Code Retrieval with Semantic-Augmented Reranking and Localization `gupta2025sacl` · Gupta, Dhruv; Lakshmy, Gayathri Ganesh; Xie, Yiqing	2025	arXiv preprint arXiv:2506.20081	Benchmark / infrastructure	Code	RAG-based	—	arXiv:2506.20081
Seeing is Fixing: Cross-Modal Reasoning with Multimodal LLMs for Visual Software Issue Fixing `huang2025seeing` · Huang, Kai; Zhang, Jian; Xie, Xiaofei; Chen, Chunyang	2025	arXiv preprint arXiv:2506.16136	Validation / testing	Arch Code	RAG-based	—	arXiv:2506.16136
R2e-gym: Procedural environments and hybrid verifiers for scaling open-weights swe agents `jain2025r2e` · Jain, Naman; Singh, Jaskirat; Shetty, Manish; Zheng, Liang; Sen, Koushik; Stoica, Ion	2025	arXiv preprint arXiv:2504.07164	Benchmark / infrastructure	Tool Code Memory History	Training-based / RL	Training environment	arXiv:2504.07164
CoSIL: Software Issue Localization via LLM-Driven Code Repository Graph Searching `jiang2025cosil` · Jiang, Zhonghao; Ren, Xiaoxue; Yan, Meng; Jiang, Wei; Li, Yong; Liu, Zhongxin	2025	arXiv preprint arXiv:2503.22424	Patching	Code	Graph-based	—	arXiv:2503.22424
Putting It All into Context: Simplifying Agents with LCLMs `jiang2025putting` · Jiang, Mingjian; Ruan, Yangjun; Lastras, Luis; Kapanipathi, Pavan; Hashimoto, Tatsunori	2025	arXiv preprint arXiv:2505.08120	Validation / testing	Tool Code	Tool-centric	—	arXiv:2505.08120
InfantAgent-Next: A Multimodal Generalist Agent for Automated Computer Interaction `lei2025infantagent` · Lei, Bin; Kang, Weitai; Zhang, Zijian; Chen, Winson; Xie, Xi; Zuo, Shan; Xie, Mimi; Payani, Ali; Hong, Mingyi; Yan, Yan; others	2025	arXiv preprint arXiv:2505.10887	Validation / testing	Tool Code Memory	Tool-centric	—	arXiv:2505.10887
InfCode: Adversarial Iterative Refinement of Tests and Patches for Reliable Software Issue Resolution `li2025infcode` · Li, KeFan; Wang, Mengfei; Zhang, Hengzhi; Li, Zhichao; Yuan, Yuan; Li, Mu; Gao, Xiang; Sun, Hailong; Hu, Chunming; Lv, Weifeng	2025	arXiv preprint arXiv:2511.16004	Validation / testing	Tool Code	Tool-centric	—	arXiv:2511.16004
PatchPilot: A Cost-Efficient Software Engineering Agent with Early Attempts on Formal Verification `li2025patchpilot` · Li, Hongwei; Tang, Yuheng; Wang, Shiqi; Guo, Wenbo	2025	arXiv preprint arXiv:2502.02747	History-aware reasoning	Tool Code Memory History	Tool-centric	—	arXiv:2502.02747
Swe-debate: Competitive multi-agent debate for software issue resolution `li2025swe` · Li, Han; Shi, Yuling; Lin, Shaoxin; Gu, Xiaodong; Lian, Heng; Wang, Xin; Jia, Yantao; Huang, Tao; Wang, Qianxiang	2025	arXiv preprint arXiv:2507.23348	Validation / testing	Arch Tool Code	Multi-agent	—	arXiv:2507.23348
Se-agent: Self-evolution trajectory optimization in multi-step reasoning with llm-based agents `lin2025se` · Lin, Jiaye; Guo, Yifu; Han, Yuzhen; Hu, Sen; Ni, Ziyi; Wang, Licheng; Chen, Mingguang; Liu, Hongzhang; Chen, Ronghao; He, Yangfan; others	2025	arXiv preprint arXiv:2508.02085	Validation / testing	Tool Code Memory	Memory-based	—	arXiv:2508.02085
The Complexity Trap: Simple Observation Masking Is as Efficient as LLM Summarization for Agent Context Management `lindenbauer2025complexity` · Lindenbauer, Tobias; Slinko, Igor; Felder, Ludwig; Bogomolov, Egor; Zharov, Yaroslav	2025	arXiv preprint arXiv:2508.21433	History-aware reasoning	Tool Code Memory History	Tool-centric	—	arXiv:2508.21433
Context as a tool: Context management for long-horizon swe-agents `liu2025context` · Liu, Shukai; Yang, Jian; Jiang, Bo; Li, Yizhi; Guo, Jinyang; Liu, Xianglong; Dai, Bryan	2025	arXiv preprint arXiv:2512.22087	Training / post-training	Tool Code Memory	Training-based / RL	—	arXiv:2512.22087
GraphLocator: Graph-guided Causal Reasoning for Issue Localization `liu2025graphlocator` · Liu, Wei; Peng, Chao; Gao, Pengfei; Liu, Aofan; Zhang, Wei; Zhao, Haiyan; Jin, Zhi	2025	arXiv preprint arXiv:2512.22469	Localization	Tool Code	Graph-based	—	arXiv:2512.22469
Alibaba lingmaagent: Improving automated issue resolution via comprehensive repository exploration `ma2025alibaba` · Ma, Yingwei; Yang, Qingping; Cao, Rongyu; Li, Binhua; Huang, Fei; Li, Yongbin	2025	Proceedings of the 33rd ACM International Conference on the Foundations of Software Engineering	Validation / testing	Tool Code Memory	Graph-based	—	Scholar
Sorft: Issue resolving with subtask-oriented reinforced fine-tuning `ma2025sorft` · Ma, Zexiong; Peng, Chao; Gao, Pengfei; Meng, Xiangxin; Zou, Yanzhen; Xie, Bing	2025	arXiv preprint arXiv:2502.20127	History-aware reasoning	Code History	Training-based / RL	—	arXiv:2502.20127
Thinking longer, not larger: Enhancing software engineering agents via scaling test-time compute `ma2025thinking` · Ma, Yingwei; Li, Yongbin; Dong, Yihong; Jiang, Xue; Li, Yanhao; Liu, Yue; Cao, Rongyu; Chen, Jue; Huang, Fei; Li, Binhua	2025	2025 40th IEEE/ACM International Conference on Automated Software Engineering (ASE)	History-aware reasoning	Tool Code Memory History	Training-based / RL	—	Scholar
Tool-integrated reinforcement learning for repo deep search `ma2025tool` · Ma, Zexiong; Peng, Chao; Zeng, Qunhong; Gao, Pengfei; Zou, Yanzhen; Xie, Bing	2025	arXiv preprint arXiv:2508.03012	History-aware reasoning	Tool Code Memory History	Training-based / RL	—	arXiv:2508.03012
EXPEREPAIR: Dual-Memory Enhanced LLM-based Repository-Level Program Repair `mu2025experepair` · Mu, Fangwen; Wang, Junjie; Shi, Lin; Wang, Song; Li, Shoubin; Wang, Qing	2025	arXiv preprint arXiv:2506.10484	Memory / experience	Tool Code Memory	Memory-based	—	arXiv:2506.10484
Issue2test: Generating reproducing test cases from issue reports `nashid2025issue2test` · Nashid, Noor; Bouzenia, Islem; Pradel, Michael; Mesbah, Ali	2025	arXiv preprint arXiv:2503.16320	Validation / testing	Arch Tool Code	Tool-centric	—	arXiv:2503.16320
Reasoningbank: Scaling agent self-evolving with reasoning memory `ouyang2025reasoningbank` · Ouyang, Siru; Yan, Jun; Hsu, I; Chen, Yanfei; Jiang, Ke; Wang, Zifeng; Han, Rujun; Le, Long T; Daruki, Samira; Tang, Xiangru; others	2025	arXiv preprint arXiv:2509.25140	Memory / experience	Tool Code Memory	Memory-based	—	arXiv:2509.25140
SemAgent: A Semantics Aware Program Repair Agent `pabba2025semagent` · Pabba, Anvith; Mathai, Alex; Chakraborty, Anindya; Ray, Baishakhi	2025	arXiv preprint arXiv:2506.16650	Validation / testing	Tool Code	Tool-centric	—	arXiv:2506.16650
Codecor: An llm-based self-reflective multi-agent framework for code generation `pan2025codecor` · Pan, Ruwei; Zhang, Hongyu; Liu, Chao	2025	arXiv preprint arXiv:2501.07811	Validation / testing	Tool	Multi-agent	—	arXiv:2501.07811
Forecasting Frontier Language Model Agent Capabilities `pimpale2025forecasting` · Pimpale, Govind; H{\o}jmark, Axel; Scheurer, J{\'e}r{\'e}my; Hobbhahn, Marius	2025	arXiv preprint arXiv:2502.15850	Validation / testing	Tool Code Memory	Tool-centric	—	arXiv:2502.15850
Devstral: Fine-tuning Language Models for Coding Agent Applications `rastogi2025devstral` · Rastogi, Abhinav; Yang, Adam; Jiang, Albert Q; Liu, Alexander H; Sablayrolles, Alexandre; H{\'e}liou, Am{\'e}lie; Martin, Am{\'e}lie; Agarwal, Anmol; Ehrenberg, Andy; Lo, Andy; others	2025	arXiv preprint arXiv:2509.25193	Training / post-training	Tool Code Memory	Training-based / RL	—	arXiv:2509.25193
SweRank: Software Issue Localization with Code Ranking `reddy2025swerank` · Reddy, Revanth Gangi; Suresh, Tarun; Doo, JaeHyeok; Liu, Ye; Nguyen, Xuan Phi; Zhou, Yingbo; Yavuz, Semih; Xiong, Caiming; Ji, Heng; Joty, Shafiq	2025	arXiv preprint arXiv:2505.07849	History-aware reasoning	Code History	Training-based / RL	—	arXiv:2505.07849
SweRank+: Multilingual, Multi-Turn Code Ranking for Software Issue Localization `reddy2025swerank+` · Reddy, Revanth Gangi; Liu, Ye; Zhao, Wenting; Doo, JaeHyeok; Suresh, Tarun; Lee, Daniel; Xiong, Caiming; Zhou, Yingbo; Yavuz, Semih; Joty, Shafiq	2025	arXiv preprint arXiv:2512.20482	History-aware reasoning	Code Memory Lang History	Training-based / RL	Language-specific	arXiv:2512.20482
Specrover: Code intent extraction via llms `ruan2025specrover` · Ruan, Haifeng; Zhang, Yuntong; Roychoudhury, Abhik	2025	2025 IEEE/ACM 47th International Conference on Software Engineering (ICSE)	Validation / testing	Tool Code	Tool-centric	—	Scholar
Aime: Towards Fully-Autonomous Multi-Agent Framework `shi2025aime` · Shi, Yexuan; Wang, Mingyu; Cao, Yunxiang; Lai, Hongjie; Lan, Junjian; Han, Xin; Wang, Yu; Geng, Jie; Li, Zhenan; Xia, Zihao; others	2025	arXiv preprint arXiv:2507.11988	Patching	Tool Code	Multi-agent	—	arXiv:2507.11988
SWE-RM: Execution-free Feedback For Software Engineering Agents `shum2025swe` · Shum, KaShun; Hui, Binyuan; Chen, Jiawei; Zhang, Lei; Yang, Jiaxi; Huang, Yuzhen; Lin, Junyang; He, Junxian; others	2025	arXiv preprint arXiv:2512.21919	Training / post-training	Tool Code Memory	Training-based / RL	—	arXiv:2512.21919
Nemotron-cortexa: Enhancing llm agents for software engineering tasks via improved localization and solution diversity `sohrabizadeh2025nemotron` · Sohrabizadeh, Atefeh; Song, Jialin; Liu, Mingjie; Roy, Rajarshi; Lee, Chankyu; Raiman, Jonathan; Catanzaro, Bryan	2025	Forty-second International Conference on Machine Learning	History-aware reasoning	Tool Code History	Graph-based	—	Scholar
Coding Agents with Multimodal Browsing are Generalist Problem Solvers `soni2025coding` · Soni, Aditya Bharat; Li, Boxuan; Wang, Xingyao; Chen, Valerie; Neubig, Graham	2025	arXiv preprint arXiv:2506.03011	Validation / testing	Tool Code	Tool-centric	—	arXiv:2506.03011
Bugpilot: Complex bug generation for efficient learning of swe skills `sonwane2025bugpilot` · Sonwane, Atharv; White, Isadora; Lee, Hyunji; Pereira, Matheus; Caccia, Lucas; Kim, Minseon; Shi, Zhengyan; Singh, Chinmay; Sordoni, Alessandro; C{\^o}t{\'e}, Marc-Alexandre; others	2025	arXiv preprint arXiv:2510.19898	Training / post-training	Tool Code Memory	Training-based / RL	—	arXiv:2510.19898
Learn-by-interact: A data-centric framework for self-adaptive agents in realistic environments `su2025learn` · Su, Hongjin; Sun, Ruoxi; Yoon, Jinsung; Yin, Pengcheng; Yu, Tao; Ar{\i}k, Sercan {\"O}	2025	arXiv preprint arXiv:2501.10893	Training / post-training	Memory	Training-based / RL	—	arXiv:2501.10893
Scaling long-horizon llm agent via context-folding `sun2025scaling` · Sun, Weiwei; Lu, Miao; Ling, Zhan; Liu, Kang; Yao, Xuesong; Yang, Yiming; Chen, Jiecao	2025	arXiv preprint arXiv:2510.11967	Training / post-training	Tool Code	Training-based / RL	—	arXiv:2510.11967
Agent kb: Leveraging cross-domain experience for agentic problem solving `tang2025agent` · Tang, Xiangru; Qin, Tianrui; Peng, Tianhao; Zhou, Ziyang; Shao, Daniel; Du, Tingting; Wei, Xinming; Xia, Peng; Wu, Fang; Zhu, He; others	2025	arXiv preprint arXiv:2507.06229	Memory / experience	Memory	RAG-based	—	arXiv:2507.06229
Boosting Open-Source LLMs for Program Repair via Reasoning Transfer and LLM-Guided Reinforcement Learning `tang2025boosting` · Tang, Xunzhu; Klein, Jacques; Bissyand{\'e}, Tegawend{\'e} F	2025	arXiv preprint arXiv:2506.03921	Training / post-training	Tool Code Memory	Training-based / RL	—	arXiv:2506.03921
Co-PatcheR: Collaborative Software Patching with Component (s)-specific Small Reasoning Models `tang2025co` · Tang, Yuheng; Li, Hongwei; Zhu, Kaijie; Yang, Michael; Ding, Yangruibo; Guo, Wenbo	2025	arXiv preprint arXiv:2505.18955	History-aware reasoning	Tool Code Memory History	Training-based / RL	—	arXiv:2505.18955
SynFix: Dependency-aware program repair via RelationGraph analysis `tang2025synfix` · Tang, Xunzhu; Gao, Jiechao; Xu, Jin; Sun, Tiezhu; Song, Yewei; Ezzini, Saad; Ou{\'e}draogo, Wendk{\^u}uni C; Klein, Jacques; Bissyand{\'e}, Tegawend{\'e} F	2025	Findings of the Association for Computational Linguistics: ACL 2025	Validation / testing	Tool Code	Graph-based	—	Scholar
Code Graph Model (CGM): A Graph-Integrated Large Language Model for Repository-Level Software Engineering Tasks `tao2025code` · Tao, Hongyuan; Zhang, Ying; Tang, Zhenhao; Peng, Hongen; Zhu, Xukun; Liu, Bingchang; Yang, Yingguang; Zhang, Ziyin; Xu, Zhaogui; Zhang, Haipeng; others	2025	arXiv preprint arXiv:2505.16901	History-aware reasoning	Code History	Graph-based	—	arXiv:2505.16901
Meta-RAG on Large Codebases Using Code Summarization `tawosi2025meta` · Tawosi, Vali; Alamir, Salwa; Liu, Xiaomo; Veloso, Manuela	2025	arXiv preprint arXiv:2508.02611	Patching	Code	RAG-based	—	arXiv:2508.02611
CoreThink: A Symbolic Reasoning Layer to reason over Long Horizon Tasks with LLMs `vaghasiya2025corethink` · Vaghasiya, Jay; Ghugarkar, Omkar; Bhat, Vishvesh; Dholaria, Vipul; McAuley, Julian	2025	arXiv preprint arXiv:2509.00971	Validation / testing	Tool Code	Tool-centric	—	arXiv:2509.00971
Repeton: Structured Bug Repair with ReAct-Guided Patch-and-Test Cycles `vinh2025repeton` · Vinh, Nguyen Phu; Hoang, Anh Chung; Ngo, Chris; Hy, Truong-Son	2025	arXiv preprint arXiv:2506.08173	Validation / testing	Tool Code	Tool-centric	—	arXiv:2506.08173
Aegis: An agent-based framework for bug reproduction from issue descriptions `wang2025aegis` · Wang, Xinchen; Gao, Pengfei; Meng, Xiangxin; Peng, Chao; Hu, Ruida; Lin, Yun; Gao, Cuiyun	2025	Proceedings of the 33rd ACM International Conference on the Foundations of Software Engineering	Validation / testing	Tool Code Memory	Tool-centric	—	Scholar
Extracting Conceptual Knowledge to Locate Software Issues `wang2025extracting` · Wang, Ying; Mao, Wenjun; Wang, Chong; Zhou, Zhenhao; Zhou, Yicheng; Zhao, Wenyun; Lou, Yiling; Peng, Xin	2025	arXiv preprint arXiv:2509.21427	Localization	Code	RAG-based	—	arXiv:2509.21427
Improving Code Localization with Repository Memory `wang2025improving` · Wang, Boshi; Xu, Weijian; Li, Yunsheng; Gao, Mei; Xie, Yujia; Sun, Huan; Chen, Dongdong	2025	arXiv preprint arXiv:2510.01003	Memory / experience	Code History	Graph-based	—	arXiv:2510.01003
Mcts-refined cot: High-quality fine-tuning data for llm-based repository issue resolution `wang2025mcts` · Wang, Yibo; Peng, Zhihao; Wang, Ying; Wei, Zhao; Yu, Hai; Zhu, Zhiliang	2025	arXiv preprint arXiv:2506.12728	History-aware reasoning	Code History	Training-based / RL	—	arXiv:2506.12728
A Practitioner's Guide to Multi-turn Agentic Reinforcement Learning `wang2025practitioner` · Wang, Ruiyi; Ammanabrolu, Prithviraj	2025	arXiv preprint arXiv:2510.01132	Training / post-training	Tool Code Memory	Training-based / RL	—	arXiv:2510.01132
SeamlessFlow: A Trainer Agent Isolation RL Framework Achieving Bubble-Free Pipelines via Tag Scheduling `wang2025seamlessflow` · Wang, Jinghui; Wang, Shaojie; Cui, Yinghan; Chen, Xuxing; Wang, Chao; Zhang, Xiaojiang; Zhang, Minglei; Zhang, Jiarong; Zhuang, Wenhao; Cao, Yuchen; others	2025	arXiv preprint arXiv:2508.11553	History-aware reasoning	Tool Code History	Training-based / RL	—	arXiv:2508.11553
SWE-Dev: Building Software Engineering Agents with Training and Inference Scaling `wang2025swe` · Wang, Haoran; Hou, Zhenyu; Wei, Yao; Tang, Jie; Dong, Yuxiao	2025	arXiv preprint arXiv:2506.07636	History-aware reasoning	Tool Code Memory History	Training-based / RL	—	arXiv:2506.07636
Swe-rl: Advancing llm reasoning via reinforcement learning on open software evolution `wei2025swe` · Wei, Yuxiang; Duchenne, Olivier; Copet, Jade; Carbonneaux, Quentin; Zhang, Lingming; Fried, Daniel; Synnaeve, Gabriel; Singh, Rishabh; Wang, Sida I	2025	arXiv preprint arXiv:2502.18449	History-aware reasoning	Code History	Training-based / RL	—	arXiv:2502.18449
Toward training superintelligent software agents through self-play swe-rl `wei2025toward` · Wei, Yuxiang; Sun, Zhiqing; McMilin, Emily; Gehring, Jonas; Zhang, David; Synnaeve, Gabriel; Fried, Daniel; Zhang, Lingming; Wang, Sida	2025	arXiv preprint arXiv:2512.18552	History-aware reasoning	Tool Code Memory History	Training-based / RL	—	arXiv:2512.18552
Confucius Code Agent: Scalable Agent Scaffolding for Real-World Codebases `wong2025confucius` · Wong, Sherman; Qi, Zhenting; Wang, Zhaodong; Hu, Nathan; Lin, Samuel; Ge, Jun; Gao, Erwin; Chen, Wenlin; Du, Yilun; Yu, Minlan; others	2025	arXiv preprint arXiv:2512.10398	Validation / testing	Tool Code Memory	Tool-centric	—	arXiv:2512.10398
Git context controller: Manage the context of llm-based agents like git `wu2025git` · Wu, Junde	2025	arXiv preprint arXiv:2508.00031	Validation / testing	Arch Tool Code Memory	Tool-centric	—	arXiv:2508.00031
Demystifying llm-based software engineering agents `xia2025demystifying` · Xia, Chunqiu Steven; Deng, Yinlin; Dunn, Soren; Zhang, Lingming	2025	Proceedings of the ACM on Software Engineering	Validation / testing	Tool Code	Tool-centric	—	Scholar
Live-SWE-agent: Can Software Engineering Agents Self-Evolve on the Fly? `xia2025live` · Xia, Chunqiu Steven; Wang, Zhe; Yang, Yan; Wei, Yuxiang; Zhang, Lingming	2025	arXiv preprint arXiv:2511.13646	Validation / testing	Tool Code Memory	Tool-centric	—	arXiv:2511.13646
Improving the efficiency of LLM agent systems through trajectory reduction `xiao2025improving` · Xiao, Yuan-An; Gao, Pengfei; Peng, Chao; Xiong, Yingfei	2025	arXiv preprint arXiv:2509.23586	Validation / testing	Tool Code Memory	Tool-centric	—	arXiv:2509.23586
Swe-fixer: Training open-source llms for effective and efficient github issue resolution `xie2025swe` · Xie, Chengxing; Li, Bowen; Gao, Chang; Du, He; Lam, Wai; Zou, Difan; Chen, Kai	2025	arXiv preprint arXiv:2501.05040	History-aware reasoning	Tool Code Memory History	Training-based / RL	—	arXiv:2501.05040
Think-Search-Patch: A Retrieval-Augmented Reasoning Framework for Repository-Level Code Repair `xiong2025think` · Xiong, Bojian; Lei, Yikun; Liu, Xikai; Zhang, Shaowei; Zhu, Pengyun; Liu, Yan; Leng, Yongqi; Shi, Ling; Zhong, Meizhi; Zhang, Yurong; others	2025	Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track	Benchmark / infrastructure	Code	Training-based / RL	—	Scholar
Enhancing repository-level software repair via repository-aware knowledge graphs `yang2025enhancing` · Yang, Boyang; Tian, Haoye; Ren, Jiadong; Jin, Shunfu; Liu, Yang; Liu, Feng; Le, Bach	2025	arXiv preprint arXiv:2503.21710	History-aware reasoning	Tool Code History	Graph-based	—	arXiv:2503.21710
KGCompass: Knowledge Graph Enhanced Repository-Level Software Repair `yang2025kgcompass` · YANG, BOYANG; REN, JIADONG; JIN, SHUNFU; LIU, YANG; LIU, FENG; LE, BACH; TIAN, HAOYE	2025	N/A	History-aware reasoning	Tool Code History	Graph-based	—	Scholar
Kimi-dev: Agentless training as skill prior for swe-agents `yang2025kimi` · Yang, Zonghan; Wang, Shengjie; Fu, Kelin; He, Wenyang; Xiong, Weimin; Liu, Yibo; Miao, Yibo; Gao, Bofei; Wang, Yejie; Ma, Yingwei; others	2025	arXiv preprint arXiv:2509.23045	History-aware reasoning	Tool Code Memory History	Training-based / RL	—	arXiv:2509.23045
Lingxi: Repository-Level Issue Resolution Framework Enhanced by Procedural Knowledge Guided Scaling `yang2025lingxi` · Yang, Xu; Zhou, Jiayuan; Pacheco, Michael; Zhu, Wenhan; He, Pengfei; Wang, Shaowei; Liu, Kui; Pan, Ruiqi	2025	arXiv preprint arXiv:2510.11838	History-aware reasoning	Arch Tool Code Memory	Tool-centric	—	arXiv:2510.11838
Swe-smith: Scaling data for software engineering agents `yang2025swe` · Yang, John; Lieret, Kilian; Jimenez, Carlos E; Wettig, Alexander; Khandpur, Kabir; Zhang, Yanzhe; Hui, Binyuan; Press, Ofir; Schmidt, Ludwig; Yang, Diyi	2025	arXiv preprint arXiv:2504.21798	History-aware reasoning	Tool Code Memory History	Training-based / RL	—	arXiv:2504.21798
Building Coding Agents via Entropy-Enhanced Multi-Turn Preference Optimization `yu2025building` · Yu, Jiahao; Cheng, Zelei; Wu, Xian; Xing, Xinyu	2025	arXiv preprint arXiv:2509.12434	Training / post-training	Tool Code Memory	Training-based / RL	—	arXiv:2509.12434
Orcaloca: An llm agent framework for software issue localization `yu2025orcaloca` · Yu, Zhongming; Zhang, Hejia; Zhao, Yujie; Huang, Hanxian; Yao, Matrix; Ding, Ke; Zhao, Jishen	2025	arXiv preprint arXiv:2502.00350	Validation / testing	Tool Code Memory	Graph-based	—	arXiv:2502.00350
Utboost: Rigorous evaluation of coding agents on swe-bench `yu2025utboost` · Yu, Boxi; Zhu, Yuxuan; He, Pinjia; Kang, Daniel	2025	arXiv preprint arXiv:2506.09289	Benchmark / infrastructure	Tool Code History	Tool-centric	SWE-bench	arXiv:2506.09289
Guided Search Strategies in Non-Serializable Environments with Applications to Software Engineering Agents `zainullina2025guided` · Zainullina, Karina; Golubev, Alexander; Trofimova, Maria; Polezhaev, Sergei; Badertdinov, Ibragim; Litvintseva, Daria; Karasik, Simon; Fisin, Filipp; Skvortsov, Sergei; Nekrashevich, Maksim; others	2025	arXiv preprint arXiv:2505.13652	History-aware reasoning	Tool Code Memory History	Training-based / RL	—	arXiv:2505.13652
Satori-SWE: Evolutionary Test-Time Scaling for Sample-Efficient Software Engineering `zeng2025satori` · Zeng, Guangtao; Shen, Maohao; Chen, Delin; Qi, Zhenting; Das, Subhro; Gutfreund, Dan; Cox, David; Wornell, Gregory; Lu, Wei; Hong, Zhang-Wei; others	2025	arXiv preprint arXiv:2505.23604	History-aware reasoning	Tool Code Memory History	Training-based / RL	—	arXiv:2505.23604
Skywork-SWE: Unveiling Data Scaling Laws for Software Engineering in LLMs `zeng2025skywork` · Zeng, Liang; Li, Yongcong; Xiao, Yuzhen; Li, Changshi; Liu, Chris Yuhao; Yan, Rui; Wei, Tianwen; He, Jujie; Song, Xuchen; Liu, Yang; others	2025	arXiv preprint arXiv:2506.19290	History-aware reasoning	Tool Code Memory History	Training-based / RL	—	arXiv:2506.19290
cast: Enhancing code retrieval-augmented generation with structural chunking via abstract syntax tree `zhang2025cast` · Zhang, Yilin; Zhao, Xinran; Wang, Zora Zhiruo; Yang, Chenyang; Wei, Jiayi; Wu, Tongshuang	2025	arXiv preprint arXiv:2506.15655	Benchmark / infrastructure	Code	RAG-based	—	arXiv:2506.15655
Darwin Godel Machine: Open-Ended Evolution of Self-Improving Agents `zhang2025darwin` · Zhang, Jenny; Hu, Shengran; Lu, Cong; Lange, Robert; Clune, Jeff	2025	arXiv preprint arXiv:2505.22954	Validation / testing	Tool Code Memory	Tool-centric	—	arXiv:2505.22954
One Tool Is Enough: Reinforcement Learning for Repository-Level LLM Agents `zhang2025one` · Zhang, Zhaoxi; Duan, Yitong; Zhang, Yanzhi; Xu, Yiming; Wang, Zhixiang; Liang, Kun; Li, Yang; Liang, Jiahui; Xia, Deguo; Huang, Jizhou; others	2025	arXiv preprint arXiv:2512.20957	Training / post-training	Tool Code Memory	Training-based / RL	—	arXiv:2512.20957
Sealign: Alignment training for software engineering agent `zhang2025sealign` · Zhang, Kechi; Zhang, Huangzhao; Li, Ge; You, Jinliang; Li, Jia; Zhao, Yunfei; Jin, Zhi	2025	arXiv preprint arXiv:2503.18455	Training / post-training	Tool Memory	Graph-based	—	arXiv:2503.18455
Tom-swe: User mental modeling for software engineering agents `zhou2025tom` · Zhou, Xuhui; Chen, Valerie; Wang, Zora Zhiruo; Neubig, Graham; Sap, Maarten; Wang, Xingyao	2025	arXiv preprint arXiv:2510.21903	Validation / testing	Arch Tool Code Memory	Tool-centric	—	arXiv:2510.21903
Training Versatile Coding Agents in Synthetic Environments `zhu2025training` · Zhu, Yiqi; Gandhi, Apurva; Neubig, Graham	2025	arXiv preprint arXiv:2512.12216	Training / post-training	Arch Tool Code Memory	Training-based / RL	—	arXiv:2512.12216
Swe-search: Enhancing software agents with monte carlo tree search and iterative refinement `antoniades2024swe` · Antoniades, Antonis; {\"O}rwall, Albert; Zhang, Kexun; Xie, Yuxi; Goyal, Anirudh; Wang, William	2024	arXiv preprint arXiv:2410.20285	Validation / testing	Tool Code Memory	Tool-centric	—	arXiv:2410.20285
Masai: Modular architecture for software-engineering ai agents `arora2024masai` · Arora, Daman; Sonwane, Atharv; Wadhwa, Nalin; Mehrotra, Abhav; Utpala, Saiteja; Bairi, Ramakrishna; Kanade, Aditya; Natarajan, Nagarajan	2024	arXiv preprint arXiv:2406.11638	Architecture / constraint-aware	Tool Code	Multi-agent	—	arXiv:2406.11638
Large language monkeys: Scaling inference compute with repeated sampling `brown2024large` · Brown, Bradley; Juravsky, Jordan; Ehrlich, Ryan; Clark, Ronald; Le, Quoc V; R{\'e}, Christopher; Mirhoseini, Azalia	2024	arXiv preprint arXiv:2407.21787	Validation / testing	Tool Code	Tool-centric	—	arXiv:2407.21787
Coder: Issue resolving with multi-agent and task graphs `chen2024coder` · Chen, Dong; Lin, Shaoxin; Zeng, Muhan; Zan, Daoguang; Wang, Jian-Gang; Cheshkov, Anton; Sun, Jun; Yu, Hao; Dong, Guoliang; Aliev, Artem; others	2024	arXiv preprint arXiv:2406.01304	History-aware reasoning	Tool Code Memory History	Multi-agent	—	arXiv:2406.01304
Exploring the Potential of Conversational Test Suite Based Program Repair on SWE-bench `cheshkov2024exploring` · Cheshkov, Anton; Zadorozhny, Pavel; Levichev, Rodion; Maslov, Evgeny; Jaldin, Ronaldo Franco	2024	arXiv preprint arXiv:2410.04485	Validation / testing	Tool Code	Tool-centric	SWE-bench	arXiv:2410.04485
SuperCoder2. 0: Technical Report on Exploring the feasibility of LLMs as Autonomous Programmer `gautam2024supercoder2` · Gautam, Anmol; Kumar, Kishore; Jha, Adarsh; NS, Mukunda; Bhola, Ishaan	2024	arXiv preprint arXiv:2409.11190	Validation / testing	Tool Code	Tool-centric	—	arXiv:2409.11190
Infant agent: A tool-integrated, logic-driven agent with cost-effective api usage `lei2024infant` · Lei, Bin; Li, Yuchen; Zeng, Yiming; Ren, Tao; Luo, Yi; Shi, Tianyu; Gao, Zitian; Hu, Zeyu; Kang, Weitai; Chen, Qiuwu	2024	arXiv preprint arXiv:2411.01114	Validation / testing	Tool Code Memory	Tool-centric	—	arXiv:2411.01114
Codetree: Agent-guided tree search for code generation with large language models `li2024codetree` · Li, Jierui; Le, Hung; Zhou, Yingbo; Xiong, Caiming; Savarese, Silvio; Sahoo, Doyen	2024	arXiv preprint arXiv:2411.04329	Patching	Code	RAG-based	—	arXiv:2411.04329
Llms as continuous learners: Improving the reproduction of defective code in software issues `lin2024llms` · Lin, Yalan; Ma, Yingwei; Cao, Rongyu; Li, Binhua; Huang, Fei; Gu, Xiaodong; Li, Yongbin	2024	arXiv preprint arXiv:2411.13941	Validation / testing	Tool Code Memory	Tool-centric	—	arXiv:2411.13941
Codexembed: A generalist embedding model family for multiligual and multi-task code retrieval `liu2024codexembed` · Liu, Ye; Meng, Rui; Joty, Shafiq; Savarese, Silvio; Xiong, Caiming; Zhou, Yingbo; Yavuz, Semih	2024	arXiv preprint arXiv:2411.12644	Benchmark / infrastructure	Code History	Training-based / RL	—	arXiv:2411.12644
Codexgraph: Bridging large language models and code repositories via code graph databases `liu2024codexgraph` · Liu, Xiangyan; Lan, Bo; Hu, Zhiyuan; Liu, Yang; Zhang, Zhicheng; Wang, Fei; Shieh, Michael; Zhou, Wenmeng	2024	arXiv preprint arXiv:2408.03910	Patching	Code	Graph-based	—	arXiv:2408.03910
Marscode agent: Ai-native automated bug fixing `liu2024marscode` · Liu, Yizhou; Gao, Pengfei; Wang, Xinchen; Liu, Jie; Shi, Yexuan; Zhang, Zhao; Peng, Chao	2024	arXiv preprint arXiv:2409.00899	Validation / testing	Tool Code	Graph-based	—	arXiv:2409.00899
Lingma swe-gpt: An open development-process-centric language model for automated software improvement `ma2024lingma` · Ma, Yingwei; Cao, Rongyu; Cao, Yongchang; Zhang, Yue; Chen, Jue; Liu, Yibo; Liu, Yuchen; Li, Binhua; Huang, Fei; Li, Yongbin	2024	arXiv preprint arXiv:2411.00622	History-aware reasoning	Tool Code Memory History	Training-based / RL	—	arXiv:2411.00622
Repository Structure-Aware Training Makes SLMs Better Issue Resolver `ma2024repository` · Ma, Zexiong; An, Shengnan; Lin, Zeqi; Zou, Yanzhen; Xie, Bing	2024	arXiv preprint arXiv:2412.19031	History-aware reasoning	Code History	Training-based / RL	—	arXiv:2412.19031
Repograph: Enhancing ai software engineering with repository-level code graph `ouyang2024repograph` · Ouyang, Siru; Yu, Wenhao; Ma, Kaixin; Xiao, Zilin; Zhang, Zhihan; Jia, Mengzhao; Han, Jiawei; Zhang, Hongming; Yu, Dong	2024	arXiv preprint arXiv:2410.14684	Patching	Code	Graph-based	—	arXiv:2410.14684
Training software engineering agents and verifiers with swe-gym `pan2024training` · Pan, Jiayi; Wang, Xingyao; Neubig, Graham; Jaitly, Navdeep; Ji, Heng; Suhr, Alane; Zhang, Yizhe	2024	arXiv preprint arXiv:2412.21139	Benchmark / infrastructure	Tool Code Memory Stack	Training-based / RL	Training environment	arXiv:2412.21139
Hyperagent: Generalist software engineering agents to solve coding tasks at scale `phan2024hyperagent` · Phan, Huy Nhat; Nguyen, Tien N; Nguyen, Phong X; Bui, Nghi DQ	2024	arXiv preprint arXiv:2409.16299	Validation / testing	Tool Code Memory	Tool-centric	—	arXiv:2409.16299
Watson: A Cognitive Observability Framework for the Reasoning of LLM-Powered Agents `rombaut2024watson` · Rombaut, Benjamin; Masoumzadeh, Sogol; Vasilevski, Kirill; Lin, Dayi; Hassan, Ahmed E	2024	arXiv preprint arXiv:2411.03455	Localization	Code Memory	Memory-based	—	arXiv:2411.03455
CoRNStack: High-quality contrastive data for better code retrieval and reranking `suresh2024cornstack` · Suresh, Tarun; Reddy, Revanth Gangi; Xu, Yifei; Nussbaum, Zach; Mulyar, Andriy; Duderstadt, Brandon; Ji, Heng	2024	arXiv preprint arXiv:2412.01007	Benchmark / infrastructure	Code	Training-based / RL	—	arXiv:2412.01007
Magis: Llm-based multi-agent framework for github issue resolution `tao2024magis` · Tao, Wei; Zhou, Yucheng; Wang, Yanlin; Zhang, Wenqiang; Zhang, Hongyu; Cheng, Yu	2024	Advances in Neural Information Processing Systems	History-aware reasoning	Tool Code History	Multi-agent	—	Scholar
Openhands: An open platform for ai software developers as generalist agents `wang2024openhands` · Wang, Xingyao; Li, Boxuan; Song, Yufan; Xu, Frank F; Tang, Xiangru; Zhuge, Mingchen; Pan, Jiayi; Song, Yueqi; Li, Bowen; Singh, Jaskirat; others	2024	arXiv preprint arXiv:2407.16741	Validation / testing	Tool Code Memory	Tool-centric	—	arXiv:2407.16741
Swe-agent: Agent-computer interfaces enable automated software engineering `yang2024swe` · Yang, John; Jimenez, Carlos E; Wettig, Alexander; Lieret, Kilian; Yao, Shunyu; Narasimhan, Karthik; Press, Ofir	2024	Advances in Neural Information Processing Systems	Validation / testing	Tool Code Memory	Tool-centric	—	Scholar
Autocoderover: Autonomous program improvement `zhang2024autocoderover` · Zhang, Yuntong; Ruan, Haifeng; Fan, Zhiyu; Roychoudhury, Abhik	2024	Proceedings of the 33rd ACM SIGSOFT International Symposium on Software Testing and Analysis	Validation / testing	Tool Code	Tool-centric	—	Scholar
CodeV: Issue Resolving with Visual Data `zhang2024codev` · Zhang, Linhao; Zan, Daoguang; Yang, Quanshun; Huang, Zhirong; Chen, Dong; Shen, Bo; Liu, Tianyu; Gong, Yongshun; Huang, Pengjie; Lu, Xudong; others	2024	arXiv preprint arXiv:2412.17315	Patching	Code	Pipeline	—	arXiv:2412.17315
Diversity empowers intelligence: Integrating expertise of software engineering agents `zhang2024diversity` · Zhang, Kexun; Yao, Weiran; Liu, Zuxin; Feng, Yihao; Liu, Zhiwei; Murthy, Rithesh; Lan, Tian; Li, Lei; Lou, Renze; Xu, Jiacheng; others	2024	arXiv preprint arXiv:2408.07060	Validation / testing	Code Memory	Agentic	—	arXiv:2408.07060
No papers match the current filters.

Compare Methods¶

Method Comparison Table¶

This table is designed for related-work comparison and baseline selection, so it foregrounds strategy differences rather than the full coding schema.

Paper	Agent Type	Main Knowledge	Retrieval Style	Feedback Loop	Memory / History	Benchmark / Setting	Best Use Case
Swe-agent: Agent-computer interfaces enable automated software engineering	Agentic runtime	Tool, Code, Memory	Active + tool-centric	Strong	Weak explicit memory	—	Classic repository-level baseline
EXPEREPAIR: Dual-Memory Enhanced LLM-based Repository-Level Program Repair	Memory-based repair	Tool, Code, Memory	Retrieval + memory	Medium	Strong experience memory	—	Experience reuse
KGCompass: Knowledge Graph Enhanced Repository-Level Software Repair	Graph-guided retrieval	Tool, Code, History	Structured historical retrieval	Medium	Strong history	—	History-aware localization and repair
R2e-gym: Procedural environments and hybrid verifiers for scaling open-weights swe agents	Training environment	Tool, Code, Memory, History	Tool-mediated	Strong	Medium	Training environment	Agent training infrastructure
ArkEval: Benchmarking and Evaluating Automated CodeRepair for ArkTS	Benchmark / evaluation	Tool, Code, Lang, History	RAG + repo context	Medium	Weak	ArkTS / HarmonyOS	Language- and platform-specific evaluation
Agentic Rubrics as Contextual Verifiers for SWE Agents	Verifier-centered	Arch, Tool, Code	Structured contextual retrieval	Strong	Constraint-aware	—	Architecture / validation constraints

Focused Cross-Dimension Tables¶

Only the most action-oriented cross-dimension views are kept here. They answer concrete navigation and comparison questions instead of repeating every RQ matrix from the manuscript.

Source × Extraction Method¶

Source	Rule-based Extraction	Model-based Extraction	Manual Extraction
Codebase	125 Knowledge: Existing Code Knowledge, Development Tool Knowledge Examples: BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Replay: Efficient Test-Time Scaling for Software Engineering Agents	88 Knowledge: Existing Code Knowledge, Development Tool Knowledge Examples: BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Prot$\backslash$'eg$\backslash$'e: Learning to Selectively Collaborate With an Expert Unlocks Small Language Models as Software Engineering Agents	38 Knowledge: Existing Code Knowledge, Development Tool Knowledge Examples: BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Prot$\backslash$'eg$\backslash$'e: Learning to Selectively Collaborate With an Expert Unlocks Small Language Models as Software Engineering Agents
Dynamic Execution Results	101 Knowledge: Development Tool Knowledge, Existing Code Knowledge Examples: BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Replay: Efficient Test-Time Scaling for Software Engineering Agents	74 Knowledge: Existing Code Knowledge, Development Tool Knowledge Examples: BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Prot$\backslash$'eg$\backslash$'e: Learning to Selectively Collaborate With an Expert Unlocks Small Language Models as Software Engineering Agents	35 Knowledge: Development Tool Knowledge, Existing Code Knowledge Examples: BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Prot$\backslash$'eg$\backslash$'e: Learning to Selectively Collaborate With an Expert Unlocks Small Language Models as Software Engineering Agents
Expert	42 Knowledge: Existing Code Knowledge, Development Tool Knowledge Examples: BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Prot$\backslash$'eg$\backslash$'e: Learning to Selectively Collaborate With an Expert Unlocks Small Language Models as Software Engineering Agents	30 Knowledge: Existing Code Knowledge, Development Tool Knowledge Examples: BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Prot$\backslash$'eg$\backslash$'e: Learning to Selectively Collaborate With an Expert Unlocks Small Language Models as Software Engineering Agents	35 Knowledge: Existing Code Knowledge, Development Tool Knowledge Examples: BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Prot$\backslash$'eg$\backslash$'e: Learning to Selectively Collaborate With an Expert Unlocks Small Language Models as Software Engineering Agents
Repository History	43 Knowledge: Existing Code Knowledge, Repository Evolution Knowledge Examples: Closing the Loop: Universal Repository Representation with RPG-Encoder SWE-Tester: Training Open-Source LLMs for Issue Reproduction in Real-World Repositories	31 Knowledge: Existing Code Knowledge, Repository Evolution Knowledge Examples: Closing the Loop: Universal Repository Representation with RPG-Encoder SWE-Tester: Training Open-Source LLMs for Issue Reproduction in Real-World Repositories	11 Knowledge: Existing Code Knowledge, Repository Evolution Knowledge Examples: Closing the Loop: Universal Repository Representation with RPG-Encoder Swe-lego: Pushing the limits of supervised fine-tuning for software issue resolving
Document	9 Knowledge: Existing Code Knowledge, Development Tool Knowledge Examples: BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Prot$\backslash$'eg$\backslash$'e: Learning to Selectively Collaborate With an Expert Unlocks Small Language Models as Software Engineering Agents	9 Knowledge: Existing Code Knowledge, Development Tool Knowledge Examples: BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Prot$\backslash$'eg$\backslash$'e: Learning to Selectively Collaborate With an Expert Unlocks Small Language Models as Software Engineering Agents	7 Knowledge: Development Tool Knowledge, Existing Code Knowledge Examples: BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Prot$\backslash$'eg$\backslash$'e: Learning to Selectively Collaborate With an Expert Unlocks Small Language Models as Software Engineering Agents

Stage × Dominant Knowledge¶

Stage	Dominant Knowledge	Representative Works	Typical Artifacts
Localization	Existing Code Knowledge, Development Tool Knowledge, Experiential Knowledge	BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Replay: Efficient Test-Time Scaling for Software Engineering Agents SWE-Prot$\backslash$'eg$\backslash$'e: Learning to Selectively Collaborate With an Expert Unlocks Small Language Models as Software Engineering Agents	Suspicious files, functions, modules, or ranked locations.
Patch Generation	Existing Code Knowledge, Development Tool Knowledge, Experiential Knowledge	BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Replay: Efficient Test-Time Scaling for Software Engineering Agents SWE-Prot$\backslash$'eg$\backslash$'e: Learning to Selectively Collaborate With an Expert Unlocks Small Language Models as Software Engineering Agents	Candidate patches or edit plans.
Patch Validation	Existing Code Knowledge, Development Tool Knowledge, Experiential Knowledge	BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Replay: Efficient Test-Time Scaling for Software Engineering Agents Architecture-Aware Multi-Design Generation for Repository-Level Feature Addition	Test results, rubric decisions, and patch-selection signals.
Model Training	Existing Code Knowledge, Development Tool Knowledge, Experiential Knowledge	SWE-Prot$\backslash$'eg$\backslash$'e: Learning to Selectively Collaborate With an Expert Unlocks Small Language Models as Software Engineering Agents Agentic Rubrics as Contextual Verifiers for SWE Agents SWE-Master: Unleashing the Potential of Software Engineering Agents via Post-Training	Training trajectories, preference data, or reward signals.
Reproduction Test Generation	Existing Code Knowledge, Development Tool Knowledge, Experiential Knowledge	BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? Dynamic analysis enhances issue resolution SWE-Tester: Training Open-Source LLMs for Issue Reproduction in Real-World Repositories	Failing tests, reproduction scripts, or issue reproductions.
Task Planning	Development Tool Knowledge, Existing Code Knowledge, Experiential Knowledge	BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? SWE-Replay: Efficient Test-Time Scaling for Software Engineering Agents Architecture-Aware Multi-Design Generation for Repository-Level Feature Addition	Plans, decomposition traces, or retrieval strategies.
Environment Building	Development Tool Knowledge, Existing Code Knowledge, Experiential Knowledge	BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? Closing the Loop: Universal Repository Representation with RPG-Encoder Evaluating and Improving Automated Repository-Level Rust Issue Resolution with LLM-based Agents	Sandbox setup, test environment, executable workspace.

RQ6 Trends¶

These lightweight views are retained as advanced analytics. They summarize temporal emphasis and recurring knowledge pairings, but the primary workflow for finding concrete papers remains the master table and method-comparison sections above.

Temporal Trend¶

Year	Included Papers
2024	25
2025	92
2026	16

Latest-Half Knowledge Frequency¶

Knowledge Type	Count
Existing Code Knowledge	65
Development Tool Knowledge	49
Experiential Knowledge	36
Repository Evolution Knowledge	23
Design Architecture Knowledge	4
Technology Stack Knowledge	1

Pairwise Knowledge Combinations¶

Knowledge A	Knowledge B	Count
Development Tool Knowledge	Existing Code Knowledge	101
Existing Code Knowledge	Experiential Knowledge	65
Development Tool Knowledge	Experiential Knowledge	62
Existing Code Knowledge	Repository Evolution Knowledge	43
Development Tool Knowledge	Repository Evolution Knowledge	31
Experiential Knowledge	Repository Evolution Knowledge	21
Design Architecture Knowledge	Existing Code Knowledge	13
Design Architecture Knowledge	Development Tool Knowledge	11
Design Architecture Knowledge	Experiential Knowledge	6
Development Tool Knowledge	Technology Stack Knowledge	4
Existing Code Knowledge	Technology Stack Knowledge	4
Design Architecture Knowledge	Repository Evolution Knowledge	4
Experiential Knowledge	Technology Stack Knowledge	3
Existing Code Knowledge	Programming Language Knowledge	3
Development Tool Knowledge	Programming Language Knowledge	2
Programming Language Knowledge	Repository Evolution Knowledge	2
Repository Evolution Knowledge	Technology Stack Knowledge	2
Design Architecture Knowledge	Domain-specific Knowledge	1
Design Architecture Knowledge	Technology Stack Knowledge	1
Development Tool Knowledge	Domain-specific Knowledge	1
Domain-specific Knowledge	Existing Code Knowledge	1
Domain-specific Knowledge	Experiential Knowledge	1
Domain-specific Knowledge	Technology Stack Knowledge	1
Programming Language Knowledge	Technology Stack Knowledge	1
Experiential Knowledge	Programming Language Knowledge	1

Resources¶

Benchmark Index¶

Benchmark	Scope	Language	Task	Execution	Good For
BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing?	Cross-repository issue resolution	Multi / domain-heavy	Benchmark	Mixed	Domain and stack constraints
ArkEval: Benchmarking and Evaluating Automated CodeRepair for ArkTS	ArkTS repair	ArkTS	Benchmark	Yes	Low-resource language evaluation
R2e-gym: Procedural environments and hybrid verifiers for scaling open-weights swe agents	Training environment	Multi	Gym / environment	Yes	Procedural training and verification
Training software engineering agents and verifiers with swe-gym	SWE-Gym	Repo-level	Training + evaluation	Yes	Agent training baselines
SWE-Factory: Your Automated Factory for Issue Resolution Training Data and Evaluation Benchmarks	Data / benchmark factory	Repo-level	Infrastructure	Yes	Benchmark construction

Tool / Framework Index¶

Tool / Framework	Type	Main Stage	Open Style	Strength
Swe-agent: Agent-computer interfaces enable automated software engineering	Agent runtime	Full pipeline	Paper-linked	Classic agent-computer interface baseline
Openhands: An open platform for ai software developers as generalist agents	Generalist platform	Full pipeline	Paper-linked	Sandbox-based software agent runtime
Autocoderover: Autonomous program improvement	Repair agent	Localization + patching	Paper-linked	Repository exploration and autonomous improvement
KGCompass: Knowledge Graph Enhanced Repository-Level Software Repair	Graph retrieval system	Localization + patching	Paper-linked	History-aware repository retrieval

Knowledge Graph / Memory Index¶

Resource	Category	Emphasis
KGCompass: Knowledge Graph Enhanced Repository-Level Software Repair	Repository history graph	History + code alignment
CoSIL: Software Issue Localization via LLM-Driven Code Repository Graph Searching	Repository graph search	Localization via structure-aware navigation
Codexgraph: Bridging large language models and code repositories via code graph databases	Code graph database	Graph-backed repository reasoning
EXPEREPAIR: Dual-Memory Enhanced LLM-based Repository-Level Program Repair	Experience memory	Dual-memory repair guidance
Reasoningbank: Scaling agent self-evolving with reasoning memory	Reasoning memory	Reusable strategy bank

Representative Works Profiles¶

BeyondSWE [10]

Domain-specific Knowledge Technology Stack Knowledge Design Architecture Knowledge

Cross-repository, domain-specific, and dependency-migration aware. Tests agents beyond single-repo bug fixing with domain knowledge and technology stack constraints.

BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing?

BibTeX: `chen2026beyondswe` · Year: `2026`

arXiv:2603.03194 Open in Table

ArkEval [111]

Programming Language Knowledge Technology Stack Knowledge Domain-specific Knowledge

Low-resource language (ArkTS) + platform (HarmonyOS) constraints. Benchmarks code repair under language-specific and ecosystem constraints.

ArkEval: Benchmarking and Evaluating Automated CodeRepair for ArkTS

BibTeX: `xie2026arkeval` · Year: `2026`

arXiv:2602.08866 Open in Table

R2E-Gym [28]

Development Tool Knowledge Technology Stack Knowledge

Procedural environments and hybrid verifiers for scaling open-weights SWE agents. Executable environments with deterministic and LLM-based verification.

R2e-gym: Procedural environments and hybrid verifiers for scaling open-weights swe agents

BibTeX: `jain2025r2e` · Year: `2025`

arXiv:2504.07164 Open in Table

Lingxi [118]

Existing Code Knowledge Development Tool Knowledge Experiential Knowledge

Extracts procedural knowledge from historical data, demonstrating how repository history can inform current issue resolution.

Lingxi: Repository-Level Issue Resolution Framework Enhanced by Procedural Knowledge Guided Scaling

BibTeX: `yang2025lingxi` · Year: `2025`

arXiv:2510.11838 Open in Table

KGCompass [114]

Existing Code Knowledge Repository Evolution Knowledge

Repository-aware knowledge graph connecting issues, PRs, files, classes, and functions. Structured history retrieval for localization and repair.