الملخص
Training long-horizon agents to solve complex tasks requires effective supervision over extended interaction sequences. However, sparse terminal rewards obscure intermediate contributions, while on-policy distillation can lose informative teacher guidance as student-generated histories grow. To address this problem, we introduce SCAD, which organizes interactions into planning and bounded subtask execution, distills execution in local contexts, and refines planning credit through cross-rollout subtask prefix trees, with planning receiving full terminal credit and execution receiving positive terminal credit and teacher guidance. Across all evaluated benchmarks, SCAD improves macro-average accuracy over the strongest training baseline by 4.48 percentage points for text tasks and 4.19 points for multimodal tasks. SCAD effectively combines outcome-based credit assignment with teacher-guided distillation to improve planning and execution in long-horizon agents.
الكلمات المفتاحية
الموضوع
بيانات النشر
- المجلة
- غير متاح
- وصول مفتوح
- وصول مفتوح أخضر
اقتبس هذه المقالة
APA 7
Wu, S., Zhao, S., Zhu, Z., Wu, J., Luu, A. T., & Luo, H. (2026). SCAD: Structured Credit Assignment and Distillation for Long-Horizon Agents. https://omanscience.com/ar/articles/scad-structured-credit-assignment-and-distillation-for-long-horizon-agents
MLA 9
Wu, Shangyang, et al. "SCAD: Structured Credit Assignment and Distillation for Long-Horizon Agents." https://omanscience.com/ar/articles/scad-structured-credit-assignment-and-distillation-for-long-horizon-agents.
شيكاغو (المؤلف–التاريخ)
Wu, Shangyang, Shuai Zhao, Ziyue Zhu, Jinyang Wu, Anh Tuan Luu, and Haoran Luo. 2026. "SCAD: Structured Credit Assignment and Distillation for Long-Horizon Agents." https://omanscience.com/ar/articles/scad-structured-credit-assignment-and-distillation-for-long-horizon-agents.
هارفارد
Wu, S., Zhao, S., Zhu, Z., Wu, J., Luu, A. T. and Luo, H. (2026) 'SCAD: Structured Credit Assignment and Distillation for Long-Horizon Agents', Available at: https://omanscience.com/ar/articles/scad-structured-credit-assignment-and-distillation-for-long-horizon-agents.
فانكوفر
Wu S, Zhao S, Zhu Z, Wu J, Luu AT, Luo H. SCAD: Structured Credit Assignment and Distillation for Long-Horizon Agents. https://omanscience.com/ar/articles/scad-structured-credit-assignment-and-distillation-for-long-horizon-agents
IEEE
S. Wu, S. Zhao, Z. Zhu, J. Wu, A. T. Luu, and H. Luo, "SCAD: Structured Credit Assignment and Distillation for Long-Horizon Agents," https://omanscience.com/ar/articles/scad-structured-credit-assignment-and-distillation-for-long-horizon-agents.