Hevolve AI: Self-Evolving Multimodal AI Agents

Turn your domain expertise into AI agents that keep learning. Hevolve AI lets experts build multimodal AI systems by talking to them and correcting them in real time, with no code to write.

Key Features

Quick Links

© 2024 Hevolve AI Pvt Ltd. All rights reserved.

← All research
Artificial Intelligence
arXiv (cs.LG) · July 16, 2026

On-Policy Delta Distillation

Byeongho Heo, Jaehui Hwang, Sangdoo Yun +1 more

When you train a smaller model to imitate a bigger teacher, the usual approach is to have the student copy the teacher's full output distribution token by token. This paper asks whether there is a cleaner signal to copy. Their idea, the delta signal, is the difference between the teacher and the base model it started from before it was tuned for reasoning. That difference isolates exactly what the reasoning tuning added, so the student picks up the reasoning changes instead of everything the teacher happens to output.

They call the method On-Policy Delta Distillation, and across math, science, and code benchmarks it beats standard on-policy distillation, letting reasoning models reach strong scores after only a short bout of post-training.

This is my summary of the abstract, so check the paper for the exact setup and where the delta signal helps most.

From the arXiv (cs.LG) abstract

On-policy distillation is an alternative post-training method in reinforcement learning that alleviates the constraints imposed by reward models by providing token-level supervision from a teacher model. Although on-policy distillation has been studied and applied across various settings, its fundamental design remains underexplored. In this paper, we introduce a new distillation reward, termed the delta signal, instead of directly imitating the teacher's output…


More Artificial Intelligence papers