Newest papers

Revision-Based Learning from Interactive Experience and Feedback

Jiawen Li, Chenming Ge, Alice Gao, Rishab Vempati, Yi-Hao Peng, Jason Wu · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26636 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

Os modelos de geração de código são cada vez mais usados para produzir artefatos interativos, como sites e interfaces de usuário. Muitas propriedades que determinam se essas interfaces são úteis só se tornam visíveis através de interação realista, tornando-os difíceis de incorporar no treinamento de modelo. Neste artigo, apresentamos o RELIEF (aprendizagem baseada em revisão de experiência e feedback interativos), um método para melhorar modelos de geração de código usando feedback em suas interfaces geradas. O RELIEF usa um agente de codificação para revisar essas interfaces, depois incorpora as melhorias em suas pistas de geração original para produzir supervisão para ajuste fino. Para avaliar o RELIEF, coletamos um conjunto de dados de 200 modelos de código gerados sintet

Machine translation (Meta M2M100); the English original follows.

Abstract

Code generation models are increasingly used to produce interactive artifacts, such as websites and user interfaces. Many properties that determine whether these interfaces are useful only become apparent through realistic interaction, making them difficult to incorporate into model training. In this paper, we introduce RELIEF (REvision-based Learning from Interactive Experience and Feedback), a method for improving code generation models using feedback on their generated interfaces. RELIEF uses a coding agent to revise these interfaces, then incorporates the improvements into their original generation traces to produce supervision for fine-tuning. To evaluate RELIEF, we collected a dataset of 200 synthetically generated websites paired with 355 interaction recordings and accompanying feedback from 24 users. We used critiques derived from code inspection, automated interaction, explicit human feedback, and human interaction recordings to fine-tune variants of two code generation models. Our results show that feedback effectiveness varies with base model capabilities, with critiques inferred from human interaction recordings leading to the largest usability improvements overall under our tested models.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Usability, Human–computer interaction, Code (set theory), Coding (social sciences), User experience design

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange