Recursive Language Models

arxiv:2512.24601

Original source · Paper JSON

LLMs: architectures, context, training · 2025-12-31

Cites (1) · Cited by (5)

Complete cleaned Markdown · Source Markdown including bibliography

Read in parts

1. Part 1
2. Part 2
3. Part 3
4. Part 4
5. Part 5

Contents

• Recursive Language Models
• Abstract
• 1 Introduction
• 2 Recursive Language Models
• 3 Scaling Long Context Tasks
• 3.1 Tasks
• 3.2 Methods and Baselines
• 4 Results and Discussion
• 5 Analyses of RLM Trajectories
• 6 Related Works
• 7 Limitations and Future Work
• 8 Conclusion
• Appendix A Additional Training Details
• Appendix B Negative Results: Things We Tried That Did Not Work.
• Appendix C Additional Methods and Baseline Details
• C.1 Prompts for Experiments
• C.2 Summary agent baseline
• C.3 LongCoT-mini experiment.
• Appendix D Additional Benchmark Details
• D.1 OOLONG-Pairs Benchmark
• D.2 Scaling Huge Document Corpora in BrowseComp+
• Appendix E Additional RLM Trajectories
• E.1 RLM(GPT-5) on BrowseComp-Plus-Query_74
• E.2 RLM(Qwen3-Coder) on OOLONG-Pairs-Query_3
• E.3 RLM(Qwen3-Coder) on OOLONG-Query_212
• E.4 RLM(GPT-5) on CodeQA-Query_44
• Appendix F Additional Quantitative Results
• F.1 Additional Quantitative Analysis of Main Results
• F.2 Additional Runtime and Cost Analysis of RLMs
• Instructions for reporting errors