{"ID":6536474,"CreatedAt":"2026-07-14T01:21:01.169441415Z","UpdatedAt":"2026-07-14T16:11:02.601666889Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.10386","arxiv_id":"2607.10386","title":"Structured Thoughts For Improved Reasoning And Context Pruning","abstract":"Large language models (LLMs) excel at generating long chains of thought, but long reasoning traces are often verbose and memory-inefficient. In this work, we introduce Structured Thoughts, a framework that organizes reasoning into alternating \u003ctry\u003e and \u003coutcome\u003e blocks: \u003ctry\u003e captures exploratory scratch work, while \u003coutcome\u003e contains the distilled conclusion of that step. We construct a dataset of structured thoughts by segmenting reasoning traces into \u003ctry\u003e blocks and prompting an LLM to summarize each step into its corresponding \u003coutcome\u003e. Fine-tuning pretrained foundation models on this reformatted data produces models that adopt the structured reasoning style, leading to performance gains of up to 8.08\\% on reasoning benchmarks compared to standard SFT. The explicit structure also enables context pruning: after each \u003ctry\u003e/\u003coutcome\u003e pair, the \u003ctry\u003e can be pruned, allowing the model to retain conclusions without keeping the full scratch work in the context. A proof-of-concept pruning implementation achieves an average of 85\\% memory / context savings with an 8.67\\% performance drop across mathematical tasks.","short_abstract":"Large language models (LLMs) excel at generating long chains of thought, but long reasoning traces are often verbose and memory-inefficient. In this work, we introduce Structured Thoughts, a framework that organizes reasoning into alternating \u003ctry\u003e and \u003coutcome\u003e blocks: \u003ctry\u003e captures exploratory scratch work, while \u003co...","url_abs":"https://arxiv.org/abs/2607.10386","url_pdf":"https://arxiv.org/pdf/2607.10386v1","authors":"[\"Zain Sarwar\",\"Supriyo Chakraborty\",\"Berkcan Kapusuzoglu\",\"Chia-Hsuan Lee\",\"Anirban Das\",\"Stephen Rawls\",\"Kartik Balasubramaniam\",\"Sambit Sahu\"]","published":"2026-07-11T16:29:51Z","proceeding":"cs.CL","tasks":"[\"cs.CL\",\"cs.AI\"]","methods":"[\"Large Language Model\",\"Language Model\",\"LoRA\",\"Generative Adversarial Network\"]","has_code":false}