Generator-Assistant Stepwise Rollback Framework for Large Language Model Agent
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866909807590506496 |
|---|---|
| author | Li, Xingzuo Chen, Kehai Long, Yunfei Bai, Xuefeng Xu, Yong Zhang, Min |
| author_facet | Li, Xingzuo Chen, Kehai Long, Yunfei Bai, Xuefeng Xu, Yong Zhang, Min |
| contents | Large language model (LLM) agents typically adopt a step-by-step reasoning framework, in which they interleave the processes of thinking and acting to accomplish the given task. However, this paradigm faces a deep-rooted one-pass issue whereby each generated intermediate thought is plugged into the trajectory regardless of its correctness, which can cause irreversible error propagation. To address the issue, this paper proposes a novel framework called Generator-Assistant Stepwise Rollback (GA-Rollback) to induce better decision-making for LLM agents. Particularly, GA-Rollback utilizes a generator to interact with the environment and an assistant to examine each action produced by the generator, where the assistant triggers a rollback operation upon detection of incorrect actions. Moreover, we introduce two additional strategies tailored for the rollback scenario to further improve its effectiveness. Extensive experiments show that GA-Rollback achieves significant improvements over several strong baselines on three widely used benchmarks. Our analysis further reveals that GA-Rollback can function as a robust plug-and-play module, integrating seamlessly with other methods. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_02519 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Generator-Assistant Stepwise Rollback Framework for Large Language Model Agent Li, Xingzuo Chen, Kehai Long, Yunfei Bai, Xuefeng Xu, Yong Zhang, Min Computation and Language Large language model (LLM) agents typically adopt a step-by-step reasoning framework, in which they interleave the processes of thinking and acting to accomplish the given task. However, this paradigm faces a deep-rooted one-pass issue whereby each generated intermediate thought is plugged into the trajectory regardless of its correctness, which can cause irreversible error propagation. To address the issue, this paper proposes a novel framework called Generator-Assistant Stepwise Rollback (GA-Rollback) to induce better decision-making for LLM agents. Particularly, GA-Rollback utilizes a generator to interact with the environment and an assistant to examine each action produced by the generator, where the assistant triggers a rollback operation upon detection of incorrect actions. Moreover, we introduce two additional strategies tailored for the rollback scenario to further improve its effectiveness. Extensive experiments show that GA-Rollback achieves significant improvements over several strong baselines on three widely used benchmarks. Our analysis further reveals that GA-Rollback can function as a robust plug-and-play module, integrating seamlessly with other methods. |
| title | Generator-Assistant Stepwise Rollback Framework for Large Language Model Agent |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2503.02519 |