PA3: Policy-Aware Agent Alignment through Chain-of-Thought
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866911533902069760 |
|---|---|
| author | Dipta, Shubhashis Roy Bis, Daniel Zhou, Kun Wang, Lichao Yao, Benjamin Z. Guo, Chenlei Sarikaya, Ruhi |
| author_facet | Dipta, Shubhashis Roy Bis, Daniel Zhou, Kun Wang, Lichao Yao, Benjamin Z. Guo, Chenlei Sarikaya, Ruhi |
| contents | 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-the-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 PolicyRecall 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_14602 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | PA3: Policy-Aware Agent Alignment through Chain-of-Thought Dipta, Shubhashis Roy Bis, Daniel Zhou, Kun Wang, Lichao Yao, Benjamin Z. Guo, Chenlei Sarikaya, Ruhi Computation and Language Artificial Intelligence Machine Learning 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-the-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 PolicyRecall 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. |
| title | PA3: Policy-Aware Agent Alignment through Chain-of-Thought |
| topic | Computation and Language Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2603.14602 |