R2E: Turning any Github Repository into a Programming Agent Environment
By Naman Jain, Manish Shetty, Tianjun Zhang, King Han, Koushik Sen, Ion Stoica. 2024
57 citations recorded by the source corpus.
Subfields: LLM Agents & Tool Use, Code Generation & Program Synthesis
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Code Generation & Program Synthesis
Citations in the corpus
- From Crowdsourced Data to High-Quality Benchmarks: Arena-Hard and BenchBuilder Pipeline (2024, methodology): “Recent works LiveCodeBench [19], R2E [20], as well as the community based live evaluation, Chatbot Arena [9].”
- BigCodeBench: Benchmarking Code Generation with Diverse Function Calls and Complex Instructions (2024, methodology): “For test case construction, the criteria are as follows: (1) Test cases should be encapsulated by the unittest framework, making them pytest -compatible; and (2) Test cases need to be deterministic, asserting expected program behaviors based on the inputs [27].”
- Neurosymbolic Repo-level Code Localization (2026, unknown): “End-to-end issue resolution benchmarks, such as SWE-Bench [12] and its extensions, assess full issue-solving capabilities [2–4, 6, 8, 9, 12, 18–20, 31, 33, 37], primarily focus on bug-fix tasks and limited programming languages.”
- Agentic Software Issue Resolution with Large Language Models: A Survey (2025, unknown): “Size Type SWE-bench(Train) [76] 2023-10 19,008 ✗ - Real R2E [71] 2024-07 246 ✓ 270GBs Synthetic SWE-Gym [124] 2024-12 2,438 ✓ 6TBs Real SWE-bench-extra [16] 2024-12 6,411 ✓ Not Released Real SWE-Fixer [166] 2025-01 110k ✗ - Real R2E-Gym(Subset) [72] 2025-04 4,578 ✓ 4TBs Synthetic SWE-Synth [125]…”
- VeriCoder: Enhancing LLM-Based RTL Code Generation through Functional Correctness Validation (2025, unknown): “However, practical hardware development often involves large, repository-level code-bases with intricate cross-file dependencies and requirements for long-range context [40]–[42].”
- CodeARC: Benchmarking Reasoning Capabilities of LLM Agents for Inductive Program Synthesis (2025, unknown): “…much of the existing work focuses on programming tasks guided by natural language (Chen et al., 2021; Austin et al., 2021; Jimenez et al., 2023; Jain et al., 2024b), we study a fundamentally different problem: inductive program synthesis, where the objective is to infer the target program…”