# On the Role of Reasoning Patterns in the Generalization Discrepancy of Long Chain-of-Thought Supervised Fine-Tuning

arXiv: https://arxiv.org/abs/2604.01702 · HTML: https://mlsanity.com/paper/2604.01702/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning.
- Builds on 26 of its 51 references that are in the map.
- First posted on arXiv 2 Apr 2026.

## Placed in

- LLM reasoning / [Test-time compute and efficient thinking](https://mlsanity.com/lr/test-time-compute-and-efficient-thinking/)

## Builds on (most cited first, 26 of 26)

- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [gpt-oss-120b & gpt-oss-20b Model Card](https://mlsanity.com/paper/2508.10925/) (2025)
- [OpenAI o1 System Card](https://mlsanity.com/paper/2412.16720/) (2024)
- [DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models](https://mlsanity.com/paper/2512.02556/) (2025)
- [Kimi K2.5: Visual Agentic Intelligence](https://mlsanity.com/paper/2602.02276/) (2026)
- [Phi-4-reasoning Technical Report](https://mlsanity.com/paper/2504.21318/) (2025)
- [Big-Math: A Large-Scale, High-Quality Math Dataset for Reinforcement Learning in Language Models](https://mlsanity.com/paper/2502.17387/) (2025)
- [On the Interplay of Pre-Training, Mid-Training, and RL on Reasoning Language Models](https://mlsanity.com/paper/2512.07783/) (2025)
- [Stabilizing Reinforcement Learning with LLMs: Formulation and Practices](https://mlsanity.com/paper/2512.01374/) (2025)
- [Not All Correct Answers Are Equal: Why Your Distillation Source Matters](https://mlsanity.com/paper/2505.14464/) (2025)
- [Attention Illuminates LLM Reasoning: The Preplan-and-Anchor Rhythm Enables Fine-Grained Policy Optimization](https://mlsanity.com/paper/2510.13554/) (2025)
- [Think Deep, Not Just Long: Measuring LLM Reasoning Effort via Deep-Thinking Tokens](https://mlsanity.com/paper/2602.13517/) (2026)
- [Which Reasoning Trajectories Teach Students to Reason Better? A Simple Metric of Informative Alignment](https://mlsanity.com/paper/2601.14249/) (2026)
- [Shape of Thought: When Distribution Matters More than Correctness in Reasoning Tasks](https://mlsanity.com/paper/2512.22255/) (2025)
- [Data Repetition Beats Data Scaling in Long-CoT Supervised Fine-Tuning](https://mlsanity.com/paper/2602.11149/) (2026)
- [Probing the Trajectories of Reasoning Traces in Large Language Models](https://mlsanity.com/paper/2601.23163/) (2026)
- [Your thoughts tell who you are: Characterize the reasoning patterns of LRMs](https://mlsanity.com/paper/2509.24147/) (2025)
- [The Molecular Structure of Thought: Mapping the Topology of Long Chain-of-Thought Reasoning](https://mlsanity.com/paper/2601.06002/) (2026)
- [Learning to Reason: Training LLMs with GPT-OSS or DeepSeek R1 Reasoning Traces](https://mlsanity.com/paper/2511.19333/) (2025)
- [Effective Reasoning Chains Reduce Intrinsic Dimensionality](https://mlsanity.com/paper/2602.09276/) (2026)
- [CoTJudger: A Graph-Driven Framework for Automatic Evaluation of Chain-of-Thought Efficiency and Redundancy in LRMs](https://mlsanity.com/paper/2603.07078/) (2026)
- [Motif-2-12.7B-Reasoning: A Practitioner's Guide to RL Training Recipes](https://mlsanity.com/paper/2512.11463/) (2025)
- [Understanding Chain-of-Thought in Large Language Models via Topological Data Analysis](https://mlsanity.com/paper/2512.19135/) (2025)

## Built on by (most cited first, 4 of 4)

- [Demystifying On-Policy Distillation: Roles, Pathologies, and Regulations](https://mlsanity.com/paper/2607.13399/) (2026)
- [Rethinking Generalization in Reasoning SFT: A Conditional Analysis on Optimization, Data, and Model Capability](https://mlsanity.com/paper/2604.06628/) (2026)
- [Every Coin Has Two Sides: On the Dual Nature of Generalization in On-Policy Distillation of Large Language Models](https://mlsanity.com/paper/2608.16647/) (2026)
- [Diagnosing Harmful Continuation in Answer-Correct Long-CoT Training Traces](https://mlsanity.com/paper/2605.29288/) (2026)

## Abstract (arXiv)

Supervised Fine-Tuning (SFT) on long Chain-of-Thought (CoT) trajectories has become a pivotal phase in building large reasoning models. However, how CoT trajectories from different sources influence the generalization performance of models remains an open question. In this paper, we conduct a comparative study using two sources of verified CoT trajectories generated by two competing models, textttDeepSeek-R1-0528 and textttgpt-oss-120b, with their problem sets controlled to be identical. Despite their comparable performance, we uncover a striking paradox: lower training loss does not translate to better generalization. SFT on textttDeepSeek-R1-0528 data achieves remarkably lower training loss, yet exhibits significantly worse generalization performance on reasoning benchmarks compared to those trained on textttgpt-oss-120b. To understand this paradox, we perform a multi-faceted analysis probing token-level SFT loss and step-level reasoning behaviors. Our analysis reveals a difference in reasoning patterns. textttgpt-oss-120b exhibits highly convergent and deductive trajectories, whereas textttDeepSeek-R1-0528 favors a divergent and branch-heavy exploration pattern. Consequently, models trained with textttDeepSeek-R1 data inherit inefficient exploration behaviors, often getting trapped in redundant exploratory branches that hinder them from reaching correct solutions. Building upon this insight, we propose a simple yet effective remedy of filtering out frequently branching trajectories to improve the generalization of SFT. Experiments show that training on selected textttDeepSeek-R1-0528 subsets surprisingly improves reasoning performance by up to 5.1% on AIME25, 5.5% on BeyondAIME, and on average 3.6% on five benchmarks.