CAST: Credit Alignment for Cross-Tokenizer On-Policy Distillation

arXiv:2610.13257 · cs.LG, cs.CL · Submitted 2026-10-09 · Read on arXiv

cs.LG, cs.CL

Submitted: 2026-10-09

Updated: 2026-10-09

Comments: 22 pages, 8 figures, 11 tables

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Alignment — 1× in this paper · explained in 1 episode
• Distillation — 1× in this paper · explained in 2 episodes
• On-Policy Distillation — 1× in this paper · explained in 3 episodes

Sources

• Beyond Benchmarks: MathArena as an Evaluation Platform for Mathematics with LLMs
• The Llama 3 Herd of Models
• Breaking the Tokenizer Barrier: On-Policy Distillation across Model Families
• KDRL: Post-Training Reasoning LLMs via Unified Knowledge Distillation and Reinforcement Learning
• Qwen3 Technical Report

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks