arXiv cs.AI· Kyoko Ogawa, Naoki Mori·· 2 天前AI 评分22
T-GADE：热力学生成式 AI 驱动的 LLM 产物进化框架
T-GADE: Thermodynamical Generative-AI-Driven Evolution of LLM Artifacts
AI 导读
研究者提出 T-GADE，一个用 LLM 扩展的进化计算框架，将 LLM 生成的结构化产物作为个体，交叉与变异由 LLM 算子实现，并引入可在评估后自我修复个体的 HoxLM。
正文 · 原文
Abstract:Introducing large language models (LLMs) into evolutionary computation enables semantic generation of candidates, but the classical problem of maintaining population diversity remains. Inspired by the thermodynamical genetic algorithm (TDGA), we propose T-GADE (Thermodynamical Generative-AI-Driven Evolution of LLM Artifacts), an evolutionary computation framework extended through LLMs. T-GADE keeps the notions of a population and genetic operators, takes structured artifacts generated by an LLM as individuals, and implements crossover and mutation as LLM operators. It further introduces the HoxLM, an LLM that attempts self-repair of an individual after evaluation and before survival selection; two update schemes, generational and steady-state; and Fermi-type and Bose-type occupancy rules that define solution identity and whether duplicates are permitted. Survival selection is unified as minimization of the free energy $F_T=\langle E\rangle-TH$, and the selection temperature $T$ explicitly controls the balance between quality and diversity. We tested an implementation example aligned with Evolution of Heuristics (EoH) on the online bin-packing task, where the excess is the bin count in excess of the volume lower bound, relative to that bound. The HoxLM-equipped configuration reached lower excess with fewer total LLM calls than EoH. In the preregistered comparison at 400 total calls it won 9 of 10 seed-paired runs, with a median sealed excess of 0.836% against 3.912%, and it was also better in the mean and in the worst case. Temperature controls the spread of the retained population under both update schemes, and a small positive temperature improves training quality over the collapsed $T=0$ point in the generational Bose-type scheme. The evolved heuristics are shown as readable scoring rules.
| Comments: | v2: 12 pages, 4 figures, 6 tables. Presents T-GADE as a general diversity-aware framework for LLM-based evolutionary computation with an EoH-aligned implementation example, reports a preregistered seed-paired comparison in which the host-equipped configuration reaches low excess with fewer total LLM calls than EoH, and treats the selection temperature as an explicit diversity control |
| Subjects: | Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE) |
| Cite as: | arXiv:2609.12286 [cs.AI] |
| (or arXiv:2609.12286v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2609.12286 arXiv-issued DOI via DataCite |
Submission history
From: Kyoko Ogawa [view email]
[v1]
Thu, 10 Sep 2026 23:33:37 UTC (90 KB)
[v2]
Wed, 14 Oct 2026 22:40:12 UTC (28 KB)
来源：arXiv cs.AI · arxiv.org