arXiv：cs.AI· Oguzhan Gungordu, Ayse Tursucular, Siheng Xiong, Faramarz Fekri·· 16 小时前
EPOCH：以覆盖率为目标的进化搜索与经验偏好优化，用于 LLM 自动启发式设计
EPOCH: Coverage-Aware Evolution with Experiential Preference Optimization for LLM-based Automated Heuristic Design
AI 导读
EPOCH 将实例覆盖率作为 LLM 自动启发式设计（AHD）进化搜索的目标，并微调生成器以内化反思。它通过经验偏好优化让生成器从自身搜索中学习：教师 LLM 给出自然语言洞见，生成器据此产出的改进构成 DPO 偏好对，再用反思式自蒸馏（RSD）项让其更可能自行生成这些改进。在多种组合优化问题和开源 LLM 底座上，EPOCH 的启发式集合优于仅推理与微调基线，并能跨问题规模泛化。
正文
Abstract:LLM-based automated heuristic design (AHD) discovers heuristics for combinatorial optimization problems (COPs), where an LLM generates candidates in an evolutionary search. Existing frameworks freeze the generator or fine-tune it on a scalar fitness signal hiding which instances each heuristic covers. Reflection-based frameworks improve heuristics with natural-language feedback, but it stays in the prompt and never reaches the generator's weights. We propose EPOCH, which makes instance coverage the objective of the evolutionary search and fine-tunes the generator to internalize reflection. Parent selection, operator allocation, and stagnation recovery all optimize how well the heuristics jointly cover the training instances, so the search returns a complementary heuristic ensemble. The generator learns from its own search by experiential preference optimization. A teacher LLM returns a natural-language insight on the generated heuristics, and the generator produces new heuristics with this insight. These improvements join the preference pairs for direct preference optimization (DPO). Since DPO only teaches the generator to rank these improvements, a reflective self-distillation (RSD) term makes it more likely to generate them on its own, and the gains accumulate over the long horizon of the search. Across COPs and open-source LLM backbones, EPOCH's heuristic ensembles outperform inference-only and fine-tuned baselines and generalize across problem scales.
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.17379 [cs.AI] |
| (or arXiv:2610.17379v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.17379 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Oguzhan Gungordu [view email]
[v1]
Mon, 12 Oct 2026 20:36:19 UTC (821 KB)
来源：arXiv：cs.AI · arxiv.org