analiza aiqbench
This research addresses the latency and context window constraints of LLMs in enterprise settings by training agents to recall specific business policies during chain-of-thought reasoning rather than relying on full in-context prompts. By introducing a novel reward function combining policy recall metrics with hallucination penalties, the method significantly reduces token usage while improving adherence to complex rules compared to standard baselines. The work represents a practical advancement in making large language models more efficient and reliable for regulated business applications.
PA3: Policy-Aware Agent Alignment through Chain-of-Thought
arXiv:2603.14602v3 Announce Type: replace-cross Abstract: Conversational assistants powered by large language models (LLMs) excel at tool-use tasks but struggle with adhering to complex, business-specific rules. While models can reason over business rules provided in context, including all policies for every query introduces high latency and wastes compute. Furthermore, these lengthy prompts lead to long contexts, harming overall performance due to the 'needle-in-a-haystack' problem. To address these challenges, we propose a multi-stage alignment method that teaches models to recall and apply relevant business policies during chain-of-thought reasoning at inference time, without including the full business policy in-context. Furthermore, we introduce a novel Policy Recall reward based on the Jaccard score and a Hallucination Penalty for GRPO training. Altogether, our best model outperforms the baseline by 16 points and surpasses comparable in-context baselines of similar model size by 3 points, while using 40% fewer words.
Abstract:Conversational assistants powered by large language models (LLMs) excel at tool-use tasks but struggle with adhering to complex, business-specific rules. While models can reason over business rules provided in context, including all policies for every query introduces high latency and wastes compute. Furthermore, these lengthy prompts lead to long contexts, harming overall performance due to the 'needle-in-a-haystack' problem. To address these challenges, we propose a multi-stage alignment method that teaches models to recall and apply relevant business policies during chain-of-thought reasoning at inference time, without including the full business policy in-context. Furthermore, we introduce a novel Policy Recall reward based on the Jaccard score and a Hallucination Penalty for GRPO training. Altogether, our best model outperforms the baseline by 16 points and surpasses comparable in-context baselines of similar model size by 3 points, while using 40% fewer words.
Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
arXiv:2603.14602 [cs.CL]
(or arXiv:2603.14602v3 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2603.14602 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 · 237 cuvinte
Subiecte conexe
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · 17 sept. 2026
Construct validity of Claude Opus 4.8's System Card – A commentary
LessWrong · 12 iun. 2026
The J-Space Debate, Agent Swarms, and Pacing Frontier AI - Digital Minds Newsletter #4
LessWrong · 18 sept. 2026
GLM-5.3 Sometimes Considers How Train-Time Actions Shape Future Behavior
LessWrong · 15 oct. 2026