# SkillFactory: Self-Distillation For Learning Cognitive Behaviors

arXiv: https://arxiv.org/abs/2512.04072 · HTML: https://mlsanity.com/paper/2512.04072/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Reinforcement learning for reasoning, one of 17 directions in LLM reasoning.
- First posted on arXiv 3 Dec 2025.

## Placed in

- LLM reasoning / [Reinforcement learning for reasoning](https://mlsanity.com/lr/reinforcement-learning-for-reasoning/)
- Knowledge distillation / [On-policy and self-distillation for LLMs](https://mlsanity.com/distill/on-policy-and-self-distillation-for-llms/)

## Built on by (most cited first, 3 of 3)

- [Learning from Synthetic Data Improves Multi-hop Reasoning](https://mlsanity.com/paper/2603.02091/) (2026)
- [HyperThink: Text-to-Parameter Hypernetworks for Efficient Reasoning](https://mlsanity.com/paper/2610.03039/) (2026)
- [Visually Grounded Self-Reflection for Vision-Language Models via Reinforcement Learning](https://mlsanity.com/paper/2607.02490/) (2026)

## Abstract (arXiv)

Reasoning models leveraging long chains of thought employ various cognitive skills, such as verification of their answers, backtracking, retrying by an alternate method, and more. Previous work has shown that when a base language model exhibits these skills, training that model further with reinforcement learning (RL) can learn to leverage them. How can we get models to leverage skills that aren't exhibited by base models? Our work, SkillFactory, is a method for fine-tuning models to roughly learn these skills during a supervised fine-tuning (SFT) stage prior to RL. Our approach does not rely on distillation from a stronger model, but instead uses samples from the model itself, rearranged to provide training data in the format of those skills. These "silver" SFT traces may be imperfect, but are nevertheless effective for priming a model to acquire skills during RL. Our evaluation shows that (1) starting from SkillFactory SFT initialization helps a model to generalize to harder variants of a task post-RL, despite lower performance pre-RL;(2) cognitive skills are indeed used by the model; (3) RLed SkillFactory models are more robust to regression on out-of-domain tasks than RLed base models. Our work suggests that inductive biases learned prior to RL help models learn robust cognitive skill use.