IS-CoT: Breaking the Long-form Generation Collapse via Interleaved Structural Thinking
Addresses the length collapse problem in long-form generation for LLMs, enabling coherent output at extended lengths.
LLMs suffer from severe length collapse in open-ended writing beyond 2,000 words. The proposed IS-CoT framework with a dynamic Plan-Write-Reflect cycle trains IS-Writer-8B, achieving +3.08 over DeepSeek-V3.2 on LongBench-Write and robust length compliance.
Generating coherent and controllable long-form content remains a persistent challenge for Large Language Models (LLMs). While reasoning-enhanced models have demonstrated success in logic-intensive domains, our evaluation reveals that they suffer from a severe length collapse in open-ended writing, where performance degrades sharply as target lengths exceed 2,000 words. We attribute this failure to the limitation of static hierarchical planning, which struggles to provide dynamic guidance over extended contexts. To bridge this gap, we introduce the Interleaved Structural Chain-of-Thought (IS-CoT) framework. Unlike external agentic workflows, IS-CoT embeds a dynamic Plan-Write-Reflect cycle into the generation process, enabling continuous strategy adaptation and global alignment without additional assistance. Based on this framework, we construct a high-quality dataset of interleaved reasoning traces via a multi-teacher pipeline and train IS-Writer-8B. Experiments demonstrate that IS-Writer-8B achieves state-of-the-art performance on challenging long-form benchmarks (e.g., +3.08 vs. DeepSeek-V3.2 on LongBench-Write), exhibiting robust length compliance and coherence competitive with significantly larger proprietary models.