人工智慧中的遞迴式自我改進：從有界自我精煉到自主研究迴圈

摘要

AI系統日益參與自身的改進：修正其輸出、在部署過程中調整其框架、針對自身生成的資料進行訓練，並日漸涉足AI研究本身。相關文獻常以「自我精煉」、「自我獎勵」、「自我對弈」、「自我演化」等詞彙描述，但這些術語混淆了根本不同的目標。我們針對1,250篇arXiv論文（2024-2026年）進行調查，依兩個維度進行分類：系統改進的對象——部署時的行為、透過訓練調整的策略、評估機制，或研究流程本身；以及迴路閉合的程度（從人類參與迴路到完全閉合）。此分類法將有界限的自我精煉——收斂、可評估且已成為工業實踐的類型——與開放式的遞迴自我改進區分開來，後者在所有可測量的維度上仍受限於基礎需求、崩塌動態以及計算限制。其獨特之處在於為自我評估設立了專門類別：每一條改進迴路皆隱含著「某種信號能替代人類判斷」的主張。我們調查了評估機制的設計空間——評判者、過程獎勵模型、驗證器、評分量表、後設評估——並將這些信號依驗證層級排序，從形式驗證器（最強）到內在自我評估（最弱）。我們觀察到，已證實的自我改進強度與此層級一致，其失敗模式（自我確認迴圈、模型崩塌、多樣性崩塌）源於對該層級的違反，而「研究方向設定」此一使人類保留在迴路中的瓶頸，正處於該層級的頂端。我們將技術文獻連結至遞迴自我改進極限的理論，以及前沿實驗室迴路閉合報告所引發的安全與治理問題，並指出衡量自我改進的治理級評量，是此領域最缺乏研究的環節。

English

AI systems increasingly participate in their own improvement: revising their outputs, adapting their own harnesses during deployment, training on data they generate, and, increasingly, conducting AI research itself. This literature is described under a vocabulary ("self-refine," "self-reward," "self-play," "self-evolve") that conflates fundamentally different ambitions. We survey 1,250 arXiv papers (2024-2026) along two axes: what the system improves -- its behavior in deployment, its policy through training, its evaluator, or the research process itself -- and the degree of loop closure (human-in-the-loop to fully closed). The taxonomy separates bounded self-refinement -- convergent, evaluable, and already industrial practice -- from open-ended recursive self-improvement (RSI), which remains bounded by grounding requirements, collapse dynamics, and compute constraints on every measured axis. Its distinctive feature is a dedicated category for self-evaluation: every improvement loop is a claim that some signal can substitute for human judgment. We survey the evaluator design space -- judges, process reward models, verifiers, rubrics, meta-evaluation -- order the signals into a verification hierarchy from formal verifiers (strongest) to intrinsic self-assessment (weakest), and observe that demonstrated self-improvement strength tracks this hierarchy, that its failure modes (self-confirming loops, model collapse, diversity collapse) follow from its violations, and that the "research direction-setting" bottleneck keeping humans in the loop sits at the top of that hierarchy. We connect the technical literature to the theory of RSI limits and to the safety and governance questions raised by frontier-lab accounts of closing the loop, and identify governance-grade measurement of self-improvement as the field's most underpopulated niche.

人工智慧中的遞迴式自我改進：從有界自我精煉到自主研究迴圈

Recursive Self-Improvement in AI: From Bounded Self-Refinement to Autonomous Research Loops

摘要

Support