analiza aiqbench
AdaSearch tackles the critical challenge of optimizing Large Language Models by using reinforcement learning to intelligently balance internal knowledge with external search queries. This approach reduces unnecessary tool calls and mitigates risks from noisy data while maintaining high accuracy, offering a more efficient path for building autonomous AI agents.
AdaSearch: Balancing Parametric Knowledge and Search in Large Language Models via Reinforcement Learning
arXiv:2512.16883v2 Announce Type: replace Abstract: Equipping large language models (LLMs) with search engines via reinforcement learning (RL) promises effective search agents. However, adaptively balancing internal parametric knowledge with external search remains a challenge, as overreliance on search introduces unnecessary cost and risks exposure to noisy or malicious content, while relying solely on parametric knowledge risks hallucination. Prior efforts mitigate search overuse through tool-call reward shaping, which requires heavy reward engineering and conflates necessary and unnecessary search. To address these limitations, we revisit the evaluation of search agents through an F1-based decision metric, revealing that prior methods often overlook readily available parametric knowledge. Motivated by this, we propose AdaSearch, a simple two-stage, outcome-driven RL framework that disentangles problem-solving from the decision to search, making the decision process explicit and interpretable. Extensive experiments demonstrate that AdaSearch significantly improves search-decision quality and reduces unnecessary search calls, with only a small trade-off in QA accuracy relative to always-search.
Abstract:Equipping large language models (LLMs) with search engines via reinforcement learning (RL) promises effective search agents. However, adaptively balancing internal parametric knowledge with external search remains a challenge, as overreliance on search introduces unnecessary cost and risks exposure to noisy or malicious content, while relying solely on parametric knowledge risks hallucination. Prior efforts mitigate search overuse through tool-call reward shaping, which requires heavy reward engineering and conflates necessary and unnecessary search. To address these limitations, we revisit the evaluation of search agents through an F1-based decision metric, revealing that prior methods often overlook readily available parametric knowledge. Motivated by this, we propose AdaSearch, a simple two-stage, outcome-driven RL framework that disentangles problem-solving from the decision to search, making the decision process explicit and interpretable. Extensive experiments demonstrate that AdaSearch significantly improves search-decision quality and reduces unnecessary search calls, with only a small trade-off in QA accuracy relative to always-search.
Accepted to EMNLP 2026. Code and artifacts will be uploaded to this https URL
Computation and Language (cs.CL)
arXiv:2512.16883 [cs.CL]
(or arXiv:2512.16883v2 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2512.16883 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 256 cuvinte
Subiecte conexe
Forgotten in Weights, Recovered by Tools: Agentic Tool Unlearning for LLM Agents
arXiv · 25 aug. 2026
ForgetBench: Benchmarking Forgetting Dynamics of Long-Term Parametric Memory in Language Models
arXiv · 31 iul. 2026
Selecting a vector store for Amazon Bedrock Knowledge Bases
AWS · 17 sept. 2026
Interacting collaborators reveal noninteracting fermions
Quantum Frontiers · 10 aug. 2026