Search papers, labs, and topics across Lattice.
The paper introduces UNLOCK, a training-free framework to transfer post-trained capabilities across models of different scales by aligning latent subspaces. UNLOCK extracts a capability direction by contrasting activations between capability-present and capability-absent source models, then aligns it to a target model using a low-rank linear transformation. Experiments demonstrate significant improvements in reasoning behaviors like Chain-of-Thought and mathematical reasoning when transferring capabilities across Qwen models.
Unlock hidden potential: you can boost a model's reasoning skills by linearly transferring capabilities learned by *other* models, even across different sizes, without any training.
We investigate whether post-trained capabilities can be transferred across models without retraining, with a focus on transfer across different model scales. We propose the Master Key Hypothesis, which states that model capabilities correspond to directions in a low-dimensional latent subspace that induce specific behaviors and are transferable across models through linear alignment. Based on this hypothesis, we introduce UNLOCK, a training-free and label-free framework that extracts a capability direction by contrasting activations between capability-present and capability-absent Source variants, aligns it with a Target model through a low-rank linear transformation, and applies it at inference time to elicit the behavior. Experiments on reasoning behaviors, including Chain-of-Thought (CoT) and mathematical reasoning, demonstrate substantial improvements across model scales without training. For example, transferring CoT reasoning from Qwen1.5-14B to Qwen1.5-7B yields an accuracy gain of 12.1% on MATH, and transferring a mathematical reasoning direction from Qwen3-4B-Base to Qwen3-14B-Base improves AGIEval Math accuracy from 61.1% to 71.3%, surpassing the 67.8% achieved by the 14B post-trained model. Our analysis shows that the success of transfer depends on the capabilities learned during pre-training, and that our intervention amplifies latent capabilities by sharpening the output distribution toward successful reasoning trajectories.
