- 论文信息与摘要
- 1 Introduction
- 2 Pre-training
- 3 Post-Training
- 4 Evaluations
- 5 Limitations
- 6 Conclusions
- 7 Acknowledgments
- References
- Appendix
- Appendix A Contributions
- Appendix B Token Template of Tool Calling
- Appendix C Evaluation Details
- Appendix D QK-Clip Does Not Impair Model Quality
- Appendix E Why Muon is More Prone to Logit Explosion
- Appendix F K2 Critic Rubrics for General RL
- Appendix G Engine Switching Pipeline for RL Training
