ReDAct: Uncertainty-Aware Deferral for LLM Agents
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911576174362624 |
|---|---|
| author | Piatrashyn, Dzianis Kotelevskii, Nikita Grishchenkov, Kirill Glazkov, Nikita Nasonov, Ivan Makarov, Ilya Baldwin, Timothy Nakov, Preslav Vashurin, Roman Panov, Maxim |
| author_facet | Piatrashyn, Dzianis Kotelevskii, Nikita Grishchenkov, Kirill Glazkov, Nikita Nasonov, Ivan Makarov, Ilya Baldwin, Timothy Nakov, Preslav Vashurin, Roman Panov, Maxim |
| contents | Recently, LLM-based agents have become increasingly popular across many applications, including complex sequential decision-making problems. However, they inherit the tendency of LLMs to hallucinate, leading to incorrect decisions. In sequential settings, even a single mistake can irreversibly degrade the trajectory, making hallucinations an even bigger problem. Although larger LLMs hallucinate less, they incur a significantly higher per-token cost. In this paper, we address this tradeoff by proposing ReDAct (Reason-Defer-Act). In ReDAct, an agent is equipped with two LLMs: a small, cheap model used by default, and a large, more reliable but expensive model. When the predictive uncertainty of the small model exceeds a calibrated threshold, the decision is deferred to the large model. We evaluate our approach in text-based embodied environments such as ALFWorld and MiniGrid and show that deferring only about 15% of decisions to the large model can match the quality of using it exclusively, while significantly reducing inference costs. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_07036 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | ReDAct: Uncertainty-Aware Deferral for LLM Agents Piatrashyn, Dzianis Kotelevskii, Nikita Grishchenkov, Kirill Glazkov, Nikita Nasonov, Ivan Makarov, Ilya Baldwin, Timothy Nakov, Preslav Vashurin, Roman Panov, Maxim Computation and Language Machine Learning Multiagent Systems Recently, LLM-based agents have become increasingly popular across many applications, including complex sequential decision-making problems. However, they inherit the tendency of LLMs to hallucinate, leading to incorrect decisions. In sequential settings, even a single mistake can irreversibly degrade the trajectory, making hallucinations an even bigger problem. Although larger LLMs hallucinate less, they incur a significantly higher per-token cost. In this paper, we address this tradeoff by proposing ReDAct (Reason-Defer-Act). In ReDAct, an agent is equipped with two LLMs: a small, cheap model used by default, and a large, more reliable but expensive model. When the predictive uncertainty of the small model exceeds a calibrated threshold, the decision is deferred to the large model. We evaluate our approach in text-based embodied environments such as ALFWorld and MiniGrid and show that deferring only about 15% of decisions to the large model can match the quality of using it exclusively, while significantly reducing inference costs. |
| title | ReDAct: Uncertainty-Aware Deferral for LLM Agents |
| topic | Computation and Language Machine Learning Multiagent Systems |
| url | https://arxiv.org/abs/2604.07036 |