No Task Vector Is an Island: A Comprehensive Study on the Composability of Task Vectors from On-Policy Distillation

arXiv:2609.39405 · cs.LG · Submitted 2026-09-30 · Read on arXiv

cs.LG

Submitted: 2026-09-30

Updated: 2026-09-30

Code: https://github.com/Di-viner/LLM-Fusion

Project page: https://maa.org/ma

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

Terminology

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

Sources

• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
• Task Singular Vectors: Reducing Task Interference in Model Merging
• MiniLLM: On-Policy Distillation of Large Language Models
• Rethinking On-Policy Distillation of Large Language Models: Phenomenology, Mechanism, and Recipe
• On the Geometry of On-Policy Distillation
• Nemotron-Cascade 2: Post-Training LLMs with Cascade RL and Multi-Domain On-Policy Distillation
• Dense Supervision, Sparse Updates: On the Sparsity and Geometry of On-Policy Distillation
• Instruction-Following Evaluation for Large Language Models
• CAMFT: Conflict-Aware Mergeable Fine-Tuning for Large Language Models
• Teacher Should Think Ahead: Adaptive Continuations for Reliable On-Policy Distillation

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