CircuitSeer: Mining High-Quality Data by Probing Mathematical Reasoning Circuits in LLMs
This addresses the computational expense of training on massive reasoning datasets for LLMs, offering an efficient alternative to existing methods.
The paper tackled the problem of scaling LLM reasoning performance by proposing CircuitSeer, a data selection method that identifies high-quality data based on internal reasoning circuits, achieving a 1.4-point gain in Pass@1 with only 10% of the data.
Large language models (LLMs) have demonstrated impressive reasoning capabilities, but scaling their performance often relies on massive reasoning datasets that are computationally expensive to train on. Existing data selection methods aim to curate smaller, high-quality subsets but often rely on costly external models or opaque heuristics. In this work, we shift the focus from external heuristics to the model's internal mechanisms. We find that complex reasoning tasks consistently activate a sparse, specialized subset of attention heads, forming core reasoning circuits. Building on this insight, we propose CircuitSeer, a novel data selection method that quantifies the reasoning complexity of data by measuring its influence on these crucial circuits. Extensive experiments on 4 models and 9 datasets demonstrate CircuitSeer's superiority. Notably, fine-tuning Qwen2.5-Math-7B on just 10% of data selected by our method achieves a 1.4-point gain in average Pass@1 over training on the full dataset, highlighting its efficiency and effectiveness.