Zipping the Thought: When and How Compressed Reasoning Data Works in LLM Post-Training

• URL: http://arxiv.org/abs/2605.28008v1
• Date: Wed, 27 May 2026 06:02:41 GMT
• Title: Zipping the Thought: When and How Compressed Reasoning Data Works in LLM Post-Training
• Abstract summary: Large language models (LLMs) can now solve complex problems through long chain-of-thought (CoT) reasoning.<n>The trade-off between performance and token cost remains a central challenge.<n>We propose a taxonomy of CoT consisting of Explicit CoT, Composed CoT, and Implicit CoT.
• Score: 41.976487058248345
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language models (LLMs) can now solve complex problems through long chain-of-thought (CoT) reasoning, but the trade-off between performance and token cost remains a central challenge. To address this issue, supervised fine-tuning (SFT) often uses compressed reasoning data, where CoT traces are shortened into compact forms. However, the effect of such compressed reasoning data on post-training remains poorly understood. In this paper, we propose a taxonomy of CoT consisting of Explicit CoT, which outputs all operations without aggregation, Composed CoT, which combines multiple operations into a single step, and Implicit CoT, which omits intermediate operations. We construct a synthetic compositional reasoning task that allows controlled variation of difficulty, compression granularity, and data size, and conducted a comprehensive set of experiments across different model families and sizes. Notably, we find that (i) coarser CoT requires more SFT data, (ii) compared with Explicit CoT, Composed CoT and Implicit CoT benefit more from data scaling, while Composed CoT benefits from data repetition and Implicit CoT tends to lead to memorization, (iii) unlike SFT, subsequent reinforcement learning (RL) with verifiable rewards (RLVR) decomposes compressed steps learned during SFT, and (iv) unidirectional CoT ordering shows stronger generalization on longer sequential tasks. Our findings provide implications for CoT design under data resource constraints and offer important insights into the mechanisms of SFT and RL in LLM post-training.

Related papers

• Revisiting the Necessity of Lengthy Chain-of-Thought in Vision-centric Reasoning Generalization [55.6995787502694]
  We study how different Chain-of-language (CoT) designs affect the acquisition of the generalizable visual reasoning ability.<n>We compare three representative CoT formats: Language CoT, Grounding CoT, and Visual CoT.<n>Our experiments reveal that visual and longer CoT mainly accelerate convergence but do not lift the final performance ceiling.
  arXiv  Detail & Related papers  (2025-11-27T16:19:34Z)
• Learning Composable Chains-of-Thought [57.73731224510169]
  We train large language models (LLMs) to reason on chain-of-thought (CoT) traces of in-distribution reasoning problems.<n>We take a step towards compositional generalization of reasoning skills when addressing a target compositional task that has no labeled CoT data.<n>We can train "atomic CoT" models on the atomic tasks with Composable CoT data and combine them with multitask learning or model merging for better zero-shot performance on the target compositional task.
  arXiv  Detail & Related papers  (2025-05-28T17:51:10Z)
• Beyond In-Distribution Success: Scaling Curves of CoT Granularity for Language Model Generalization [35.16980045900664]
  Generalization to novel compound tasks under distribution shift is important for deploying transformer-based language models (LMs)<n>This work investigates Chain-of-Thought (CoT) reasoning as a means to enhance OOD generalization.
  arXiv  Detail & Related papers  (2025-02-25T15:04:17Z)
• When More is Less: Understanding Chain-of-Thought Length in LLMs [51.631483479081645]
  Large Language Models (LLMs) employ Chain-of-Thought (CoT) reasoning to deconstruct complex problems.<n>This paper argues that longer CoTs are often presumed superior, arguing that longer is not always better.
  arXiv  Detail & Related papers  (2025-02-11T05:28:59Z)
• C3oT: Generating Shorter Chain-of-Thought without Compromising Effectiveness [18.073777359647515]
  Chain-of-Thought (CoT) before deriving the answer can improve the reasoning capabilities of large language models (LLMs)<n>However, the length of the generated CoT is much longer than the desired final answer, which results in additional decoding costs.<n>This paper presents a CoT compression framework that involves a compressor to compress an original longer CoT into a shorter CoT.
  arXiv  Detail & Related papers  (2024-12-16T11:12:45Z)
• From Sparse Dependence to Sparse Attention: Unveiling How Chain-of-Thought Enhances Transformer Sample Efficiency [17.612497960364916]
  Chain-of-thought (CoT) significantly enhances the reasoning performance of large language models (LLM)<n>We demonstrate that CoT can substantially improve sample efficiency even when representation power is sufficient.<n>We show that CoT simplifies the learning process by introducing sparse dependencies among input tokens, and leads to a sparse and interpretable attention.
  arXiv  Detail & Related papers  (2024-10-07T19:45:09Z)
• Training Nonlinear Transformers for Chain-of-Thought Inference: A Theoretical Generalization Analysis [82.51626700527835]
  Chain-of-shift (CoT) is an efficient method that enables the reasoning ability of large language models by augmenting the query using examples with multiple intermediate steps.<n>We show that despite the theoretical success of CoT, it fails to provide an accurate generalization when CoT does.
  arXiv  Detail & Related papers  (2024-10-03T03:12:51Z)
• Expediting and Elevating Large Language Model Reasoning via Hidden Chain-of-Thought Decoding [14.175444025026508]
  Large language models (LLMs) have demonstrated remarkable capabilities in tasks requiring chain-of-thought (CoT) prompting. generating the full CoT process results in significantly longer output sequences, leading to increased computational costs and latency during inference. We propose a novel approach to compress the CoT process through semantic alignment, enabling more efficient decoding while preserving the benefits of CoT reasoning.
  arXiv  Detail & Related papers  (2024-09-13T06:29:20Z)
• ChainLM: Empowering Large Language Models with Improved Chain-of-Thought Prompting [124.69672273754144]
  Chain-of-Thought (CoT) prompting can enhance the reasoning capabilities of large language models (LLMs) Existing CoT approaches usually focus on simpler reasoning tasks and thus result in low-quality and inconsistent CoT prompts. We introduce CoTGenius, a novel framework designed for the automatic generation of superior CoT prompts.
  arXiv  Detail & Related papers  (2024-03-21T11:34:26Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.