# Marathoner: Ultra-Long-Horizon Autonomous Intelligence

arXiv: https://arxiv.org/abs/2609.34378 · HTML: https://mlsanity.com/paper/2609.34378/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 32 references that are in the map.
- First posted on arXiv 28 Sep 2026.

## Builds on (most cited first, 23 of 23)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [OpenAI GPT-5 System Card](https://mlsanity.com/paper/2601.03267/) (2025)
- [Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces](https://mlsanity.com/paper/2601.11868/) (2026)
- [AReaL: A Large-Scale Asynchronous Reinforcement Learning System for Language Reasoning](https://mlsanity.com/paper/2505.24298/) (2025)
- [Large Language Model-Based Agents for Software Engineering: A Survey](https://mlsanity.com/paper/2409.02977/) (2024)
- [Process Reward Models That Think](https://mlsanity.com/paper/2504.16828/) (2025)
- [A Survey on Code Generation with LLM-based Agents](https://mlsanity.com/paper/2508.00083/) (2025)
- [PRMBench: A Fine-grained and Challenging Benchmark for Process-Level Reward Models](https://mlsanity.com/paper/2501.03124/) (2025)
- [GenPRM: Scaling Test-Time Compute of Process Reward Models via Generative Reasoning](https://mlsanity.com/paper/2504.00891/) (2025)
- [The Programmer's Assistant: Conversational Interaction with a Large Language Model for Software Development](https://mlsanity.com/paper/2302.07080/) (2023)
- [Process Reward Model with Q-Value Rankings](https://mlsanity.com/paper/2410.11287/) (2024)
- [NL2Repo-Bench: Towards Long-Horizon Repository Generation Evaluation of Coding Agents](https://mlsanity.com/paper/2512.12730/) (2025)
- [Need Help? Designing Proactive AI Assistants for Programming](https://mlsanity.com/paper/2410.04596/) (2024)
- [A Survey of Process Reward Models: From Outcome Signals to Process Supervisions for Large Language Models](https://mlsanity.com/paper/2510.08049/) (2025)
- [AI Agentic Programming: A Survey of Techniques, Challenges, and Opportunities](https://mlsanity.com/paper/2508.11126/) (2025)
- [Unlocking Reasoning Potential in Large Langauge Models by Scaling Code-form Planning](https://mlsanity.com/paper/2409.12452/) (2024)
- [Why Reasoning Fails to Plan: A Planning-Centric Analysis of Long-Horizon Decision Making in LLM Agents](https://mlsanity.com/paper/2601.22311/) (2026)
- [SWE-Marathon: Can Agents Autonomously Complete Ultra-Long-Horizon Software Work?](https://mlsanity.com/paper/2606.07682/) (2026)
- [DreamPRM: Domain-Reweighted Process Reward Model for Multimodal Reasoning](https://mlsanity.com/paper/2505.20241/) (2025)
- [Repurposing Synthetic Data for Fine-grained Search Agent Supervision](https://mlsanity.com/paper/2510.24694/) (2025)
- [Aligning Agents via Planning: A Benchmark for Trajectory-Level Reward Modeling](https://mlsanity.com/paper/2604.08178/) (2026)

## Abstract (arXiv)

Humans naturally possess the ability to work persistently toward long-term goals. Given a challenging task, humans can continuously work for months or even years to accomplish a specific objective. In this paper, we propose Marathoner, an autonomous agentic model possessing the ability of ultra-long-horizon execution. Specifically, we propose a comprehensive post-training pipeline to instill this critical capability into base model. For Ultra-Long-Horizon Task Synthesis, we leverage major release PRs containing 1000+ lines of new code from diverse GitHub repositories as the primary source for synthesizing challenging task-level data. Additionally, we introduce Multi-Task Chaining, which chains multiple generated tasks into a single more challenging task, enabling the synthesis of tasks with frontier-level difficulty. For rejection sampling finetuning, we combine strong teacher model with diverse harnesses to generate trajectories on our synthesized tasks and conduct supervised finetuning on base model with rejection sampled trajectories. For reinforcement learning, cold-started model performs real-world execution through harnesses in independent sandboxes during rollout process, effectively facilitating the acquisition of genuine ultra-long-horizon execution capability. We further propose a novel reward strategy, Later Stage Bonus Reward, which explicitly encourages model to perform meaningful maneuvers during later stages of execution. Through extensive evaluation on 5 benchmarks containing ultra-long-horizon tasks, Marathoner achieves consistent and substantial performance improvements over base model and even surpasses performance of strong proprietary model. Further analysis shows that Marathoner can consistently work for 10+ hours and conduct 1000+ tool calls on highly challenging tasks.