Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2504.16408 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908361500393472 |
|---|---|
| author | Yuan, Jiahao Sun, Xingzhe Yu, Xing Wang, Jingwen Du, Dehui Cui, Zhiqing Di, Zixiang |
| author_facet | Yuan, Jiahao Sun, Xingzhe Yu, Xing Wang, Jingwen Du, Dehui Cui, Zhiqing Di, Zixiang |
| contents | The LLMSR@XLLM25 formulates a low-resource structural reasoning task that challenges LLMs to generate interpretable, step-by-step rationales with minimal labeled data. We present Less is More, the third-place winning approach in the LLMSR@XLLM25, which focuses on structured reasoning from only 24 labeled examples. Our approach leverages a multi-agent framework with reverse-prompt induction, retrieval-augmented reasoning synthesis via GPT-4o, and dual-stage reward-guided filtering to distill high-quality supervision across three subtasks: question parsing, CoT parsing, and step-level verification. All modules are fine-tuned from Meta-Llama-3-8B-Instruct under a unified LoRA+ setup. By combining structure validation with reward filtering across few-shot and zero-shot prompts, our pipeline consistently improves structure reasoning quality. These results underscore the value of controllable data distillation in enhancing structured inference under low-resource constraints. Our code is available at https://github.com/JhCircle/Less-is-More. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_16408 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | LLMSR@XLLM25: Less is More: Enhancing Structured Multi-Agent Reasoning via Quality-Guided Distillation Yuan, Jiahao Sun, Xingzhe Yu, Xing Wang, Jingwen Du, Dehui Cui, Zhiqing Di, Zixiang Computation and Language The LLMSR@XLLM25 formulates a low-resource structural reasoning task that challenges LLMs to generate interpretable, step-by-step rationales with minimal labeled data. We present Less is More, the third-place winning approach in the LLMSR@XLLM25, which focuses on structured reasoning from only 24 labeled examples. Our approach leverages a multi-agent framework with reverse-prompt induction, retrieval-augmented reasoning synthesis via GPT-4o, and dual-stage reward-guided filtering to distill high-quality supervision across three subtasks: question parsing, CoT parsing, and step-level verification. All modules are fine-tuned from Meta-Llama-3-8B-Instruct under a unified LoRA+ setup. By combining structure validation with reward filtering across few-shot and zero-shot prompts, our pipeline consistently improves structure reasoning quality. These results underscore the value of controllable data distillation in enhancing structured inference under low-resource constraints. Our code is available at https://github.com/JhCircle/Less-is-More. |
| title | LLMSR@XLLM25: Less is More: Enhancing Structured Multi-Agent Reasoning via Quality-Guided Distillation |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2504.16408 |