SWE-Marathon: Can Agents Autonomously Complete Ultra-Long-Horizon Software Work?

• URL: http://arxiv.org/abs/2606.07682v1
• Date: Fri, 05 Jun 2026 00:39:44 GMT
• Title: SWE-Marathon: Can Agents Autonomously Complete Ultra-Long-Horizon Software Work?
• Abstract summary: SWE-Marathon is a benchmark of 20 long-horizon tasks spanning software engineering and adjacent technical domains.<n>Each task consists of a unique executable environment, a human-written reference solution, and a multi-layer verification suite.<n>Logged agent attempts average 27.2M total tokens, making SWE-Marathon substantially longer-horizon than existing SWE and command-line agent benchmarks.
• Score: 11.093716663660203
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: AI agents are increasingly expected to complete long-horizon workflows that require sustained progress over hours, millions of tokens, and complex environments. Yet current agent benchmarks largely evaluate short-form tasks, such as single pull requests, small tickets, or 5-10 minute exercises, limiting our ability to measure agents' capabilities in planning, long-context understanding, and memory use. We introduce SWE-Marathon, a benchmark of 20 long-horizon tasks spanning software engineering and adjacent technical domains. Each task consists of a unique executable environment, a human-written reference solution, and a multi-layer verification suite. Logged agent attempts average 27.2M total tokens, making SWE-Marathon substantially longer-horizon than existing SWE and command-line agent benchmarks. Current frontier coding agents solve fewer than 30% of tasks. Failures often arise from poor self-verification, self-reported infeasibility, and premature termination. We also observe reward-hacking behavior in 13.8% of rollouts, where agents attempt to exploit the environment or verifier to bypass the intended workflow. SWE-Marathon includes adversarial review of test suites and execution environments, as well as multi-layer checks designed to prevent shortcut solutions. We release SWE-Marathon, evaluation code, and agent trajectories at https://swe-marathon.org/.

Related papers

• Long-Horizon-Terminal-Bench: Testing the Limits of Agents on Long-Horizon Terminal Tasks with Dense Reward-Based Grading [46.911171278867585]
  Long-Horizon-Terminal-Bench is a terminal benchmark of 46 long-horizon tasks spanning nine categories.<n>Each task follows a Terminal-Bench-style setup with a reference solution or simulation engine, but is further graded into fine-grained sub-tasks.<n>We evaluate 15 frontier models and find that agents consume on average 9.9M tokens per task, with roughly 231 episodes and 85.3 minutes of execution time per run.
  arXiv  Detail & Related papers  (2026-07-09T21:56:37Z)
• SWE-INTERACT: Reimagining SWE Benchmarks as User-Driven Long-Horizon Coding Sessions [3.67653051806469]
  SWE-Interact places agents in a realistic developer workflow.<n>It starts with vague or incomplete instructions, progressively reveals requirements.<n>It provides targeted feedback, revisions, and new constraints until the full task goal has been handed off.
  arXiv  Detail & Related papers  (2026-06-29T17:17:45Z)
• SentinelBench: A Benchmark for Long-Running Monitoring Agents [20.69666656998877]
  SentinelBench is an open-source benchmark for time-evolving monitoring tasks.<n>It contains 100 tasks across 10 synthetic web environments, including email, calendars, finance, professional networking, and entertainment.<n>It measures task completion, reaction time, and resource use, exposing the tradeoff between responsiveness and cost.
  arXiv  Detail & Related papers  (2026-06-03T18:32:00Z)
• GTA: Generating Long-Horizon Tasks for Web Agents at Scale [82.43869456830664]
  We introduce a scalable framework, GTA, that integrates crawling, retrieval-based seeding, in-context generation, and automated quality control.<n>We instantiate the pipeline on over 50 websites covering e-commerce, government, forums, and news, with multilingual and multi-hop coverage.<n>Our contributions are three-fold: (i) formalizing multi-hop web-agent task generation, (ii) proposing an efficient and validated pipeline for automatic data creation, and (iii) releasing a dynamic benchmark with reproducible evaluation.
  arXiv  Detail & Related papers  (2026-05-28T01:05:50Z)
• SaaS-Bench: Can Computer-Use Agents Leverage Real-World SaaS to Solve Professional Workflows? [29.5539656241093]
  Computer-Using Agents (CUAs) are rapidly extending large language models (LLMs) beyond text-based reasoning toward action execution in more complex environments.<n>Existing benchmarks often rely on simplified settings, isolated tasks, or short-horizon interactions.<n>We introduce a benchmark built on 23 deployable systems across six professional domains, containing 106 tasks grounded in realistic work scenarios.<n>Experiments show that representative LLM-based agents struggle on-Bench, with even the strongest model completing fewer than 4% of tasks end-to-end.
  arXiv  Detail & Related papers  (2026-05-15T09:35:15Z)
• Odysseys: Benchmarking Web Agents on Realistic Long Horizon Tasks [67.44219836008348]
  We introduce Odysseys: a benchmark of 200 long-horizon web tasks derived from real world browsing sessions evaluated on the live Internet.<n>We find that binary pass/fail evaluation is inadequate for long-horizon settings and introduce a rubric-based evaluation, annotating each Odysseys task with an average of 6.1 graded rubrics.<n>We find that the strongest models achieve a success rate of 44.5%, which leaves substantial room for future improvements.
  arXiv  Detail & Related papers  (2026-04-27T20:05:41Z)
• LongCLI-Bench: A Preliminary Benchmark and Study for Long-horizon Agentic Programming in Command-Line Interfaces [65.11019654023978]
  LongCLI-Bench is a benchmark designed to evaluate agentic capabilities across long-horizon, realistic tasks.<n>We curated 20 high-quality, long-horizon tasks from over 1,000 computer science assignments and real-world tasks.<n>Experiments reveal that even state-of-the-art agents achieve pass rates below 20% in LongCLI-Bench.
  arXiv  Detail & Related papers  (2026-02-15T23:12:57Z)
• Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces [126.23612941699565]
  Terminal-Bench 2.0 is a benchmark composed of 89 tasks in computer terminal environments inspired by problems from real world.<n>We show that frontier models and agents score less than 65% on the benchmark.<n>We publish the dataset and evaluation harness to assist developers and researchers in future work at https://www.tbench.ai/.
  arXiv  Detail & Related papers  (2026-01-17T01:29:30Z)
• TOM-SWE: User Mental Modeling For Software Engineering Agents [75.28749912645127]
  ToM-SWE is a dual-agent architecture that pairs a primary software-engineering (SWE) agent with a lightweight theory-of-mind (ToM) partner agent.<n>ToM-SWE infers user goals, constraints, and preferences from instructions and interaction history.<n>In two software engineering benchmarks, ToM-SWE improves task success rates and user satisfaction.
  arXiv  Detail & Related papers  (2025-10-24T16:09:51Z)
• UltraHorizon: Benchmarking Agent Capabilities in Ultra Long-Horizon Scenarios [63.67884284105684]
  We introduce textbfUltraHorizon, a novel benchmark that measures the foundational capabilities essential for complex real-world challenges.<n>Agents are designed in long-horizon discovery tasks where they must iteratively uncover hidden rules.<n>Our experiments reveal that LLM-agents consistently underperform in these settings, whereas human participants achieve higher scores.
  arXiv  Detail & Related papers  (2025-09-26T02:04:00Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.