@inproceedings{yin-etal-2026-deliberative,
    title = "Deliberative Searcher: Improving {LLM} Reliability via Reinforcement Learning with Constraints",
    author = "Yin, Zhenyun  and
      Wang, Shujie  and
      Wang, Xuhong  and
      Ma, Xingjun  and
      Wang, Yingchun",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.acl-long.199/",
    pages = "4340--4354",
    ISBN = "979-8-89176-390-6",
    abstract = "Large language models with search capabilities frequently exhibit miscalibrated confidence, producing incorrect answers with high certainty. We present Deliberative Searcher, a reasoning-primary framework that integrates search operations into chain-of-thought generation while maintaining explicit confidence calibration. Our method employs constrained reinforcement learning with adaptive Lagrangian multipliers to jointly optimize correctness and reliability. Experiments across five benchmarks demonstrate substantial improvements: our 7B model reduces average false-certain rates from 54{\%} in baselines to 2{\%}, while our 72B variant achieves competitive accuracy with closed-source models and reduces false-certain rates to 9{\%}. The well-calibrated confidence scores also enable more efficient test-time compute: instead of standard majority voting, we use confidence-weighted aggregation and match the performance of 16-sample majority voting with only 4 samples, a $4\times$ reduction in inference compute. These results establish calibrated confidence as a foundation for both trustworthy outputs and adaptive test-time compute, demonstrating the value of the proposed constrained RL framework in search-augmented language models."
}