KLong: Training LLM Agent for Extremely Long-horizon Tasks

TL;DR

KLong is an open-source LLM agent designed for extremely long-horizon tasks, utilizing trajectory-splitting SFT and progressive RL to enhance long-term problem-solving capabilities.

Contribution

The paper introduces a novel training pipeline combining trajectory-splitting SFT and progressive RL, enabling LLMs to better handle very long tasks.

Findings

KLong (106B) outperforms Kimi K2 Thinking (1T) by 11.28% on PaperBench.

KLong demonstrates superior performance and generalization on multiple long-horizon benchmarks.

The proposed methods effectively preserve context and extend task-solving capabilities.

Abstract

This paper introduces KLong, an open-source LLM agent trained to solve extremely long-horizon tasks. The principle is to first cold-start the model via trajectory-splitting SFT, then scale it via progressive RL training. Specifically, we first activate basic agentic abilities of a base model with a comprehensive SFT recipe. Then, we introduce Research-Factory, an automated pipeline that generates high-quality training data by collecting research papers and constructing evaluation rubrics. Using this pipeline, we build thousands of long-horizon trajectories distilled from Claude 4.5 Sonnet (Thinking). To train with these extremely long trajectories, we propose a new trajectory-splitting SFT, which preserves early context, progressively truncates later context, and maintains overlap between sub-trajectories. In addition, to further improve long-horizon task-solving capability, we propose a novel progressive RL, which schedules training into multiple stages with progressively extended timeouts. Experiments demonstrate the superiority and generalization of KLong, as shown in Figure 1. Notably, our proposed KLong (106B) surpasses Kimi K2 Thinking (1T) by 11.28% on PaperBench, and the performance improvement generalizes to other coding benchmarks like SWE-bench Verified and MLE-bench.