4-bit量化会引发递归推理模型误差累积，导致精度崩溃。
Quantizing Recursive Reasoning Models
- 采用块级缩放的MXInt4量化方法，缓解误差积累问题。
- 4比特整数量化使数独解题准确率从84.1%降至0.0%。
- 适用于高深度递归模型，尤其适合需要长期推理的任务。
递归推理模型通过多次复用紧凑的权重共享模块来解决难题。由于这些模块被反复使用，量化误差在每一步都会累积。虽然8比特量化（整数或浮点）能保持精度，但采用全局4比特格式会导致系统性偏差积累，使数独任务的精确解准确率从84.1%骤降至0.0%（仅约25%单元正确）。本文揭示该崩溃源于激活缩放粒度而非位宽或数值格式。关键发现是改用块级缩放可完全恢复性能。为此，我们提出MXInt4——一种基于块的整数激活格式，应用于递归推理模型，在任务上表现媲美块级浮点格式，同时保留整数元素与2的幂次块尺度。此外，递归深度与模块重用程度影响量化敏感性，最深模型（EqR均衡模型）最为敏感，但块级缩放仍保持鲁棒性，并成功迁移至开放式ARC-AGI基准测试。
原文摘要 · Abstract (English)
Recursive reasoning models solve hard puzzles by applying compact, weight-tied blocks over many refinement steps. Because these blocks are reused many times, quantizing them creates a unique dynamical problem: the quantization error is incurred at every step. While 8-bit quantization (integer or float) preserves accuracy, moving to a per-tensor 4-bit format causes a systematic bias to accumulate. The ensuing drift catastrophically degrades exact-solution accuracy on Sudoku from 84.1% to 0.0% (only ~25% of cells correct). In this work, we show that this collapse is caused by activation-scaling granularity rather than bit-width or number format. Crucially, moving to per-block scaling completely restores the transition. To implement this, we apply MXInt4, a blockwise integer activation format, to recursive reasoning models. It is competitive with blockwise float formats on our tasks, while keeping integer elements and power-of-two block scales. Finally, recursion depth and reuse modulate quantization sensitivity, with the deepest architecture we test (the EqR equilibrium model) the most sensitive. Yet blockwise scaling overcomes this vulnerability, staying robust across these architectures and transferring to the open-ended ARC-AGI benchmark.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。