Advancing Block Diffusion Language Models for Test-Time Scaling

TL;DR

This paper introduces a unified framework with adaptive decoding and block-wise generation techniques to improve the efficiency and effectiveness of block diffusion language models in test-time scaling, especially for complex reasoning tasks.

Contribution

It proposes BACD and TCCF methods for dynamic decoding and block size allocation, enabling better test-time scaling in BDLMs with significant performance gains.

Findings

Achieved 2.26x speedup on AIME24

Improved reasoning accuracy by 11.2 points

Demonstrated effectiveness of adaptive strategies in BDLMs

Abstract

Recent advances in block diffusion language models have demonstrated competitive performance and strong scalability on reasoning tasks. However, existing BDLMs have limited exploration under the test-time scaling setting and face more severe decoding challenges in long Chain-of-Thought reasoning, particularly in balancing the decoding speed and effectiveness. In this work, we propose a unified framework for test-time scaling in BDLMs that introduces adaptivity in both decoding and block-wise generation. At the decoding level, we propose Bounded Adaptive Confidence Decoding (BACD), a difficulty-aware sampling strategy that dynamically adjusts denoising based on model confidence, accelerating inference while controlling error accumulation. Beyond step-wise adaptivity, we introduce Think Coarse, Critic Fine (TCCF), a test-time scaling paradigm that allocates large block sizes to exploratory reasoning and smaller block sizes to refinement, achieving an effective efficiency-effectiveness balance. To enable efficient and effective decoding with a large block size, we adopt Progressive Block Size Extension, which mitigates performance degradation when scaling block sizes. Extensive experiments show that applying BACD and TCCF to TDAR-8B yields significant improvements over strong baselines such as TraDo-8B (2.26x speedup, +11.2 points on AIME24). These results mark an important step toward unlocking the potential of BDLMs for test-time scaling in complex reasoning tasks.