# Lessons from Training Grounded LLMs with Verifiable Rewards

arXiv: https://arxiv.org/abs/2506.15522 · HTML: https://mlsanity.com/paper/2506.15522/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Builds on 18 of its 22 references that are in the map.
- First posted on arXiv 18 Jun 2025.

## Builds on (most cited first, 18 of 18)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://mlsanity.com/paper/2005.11401/) (2020)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [Retrieval-Augmented Generation for Large Language Models: A Survey](https://mlsanity.com/paper/2312.10997/) (2023)
- [Survey of Hallucination in Natural Language Generation](https://mlsanity.com/paper/2202.03629/) (2022)
- [Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection](https://mlsanity.com/paper/2310.11511/) (2023)
- [Enabling Large Language Models to Generate Text with Citations](https://mlsanity.com/paper/2305.14627/) (2023)
- [Making Retrieval-Augmented Language Models Robust to Irrelevant Context](https://mlsanity.com/paper/2310.01558/) (2023)
- [RECOMP: Improving Retrieval-Augmented LMs with Compression and Selective Augmentation](https://mlsanity.com/paper/2310.04408/) (2023)
- [Attribute First, then Generate: Locally-attributable Grounded Text Generation](https://mlsanity.com/paper/2403.17104/) (2024)
- [Effective Large Language Model Adaptation for Improved Grounding and Citation Generation](https://mlsanity.com/paper/2311.09533/) (2023)
- [Training Language Models to Generate Text with Citations via Fine-grained Rewards](https://mlsanity.com/paper/2402.04315/) (2024)
- [Measuring and Enhancing Trustworthiness of LLMs in RAG through Grounded Attributions and Learning to Refuse](https://mlsanity.com/paper/2409.11242/) (2024)
- [Citation-Enhanced Generation for LLM-based Chatbots](https://mlsanity.com/paper/2402.16063/) (2024)
- [Improving Attributed Text Generation of Large Language Models via Preference Learning](https://mlsanity.com/paper/2403.18381/) (2024)
- [Ground Every Sentence: Improving Retrieval-Augmented LLMs with Interleaved Reference-Claim Generation](https://mlsanity.com/paper/2407.01796/) (2024)

## Built on by (most cited first, 8 of 8)

- [Hybrid Reinforcement: When Reward Is Sparse, It's Better to Be Dense](https://mlsanity.com/paper/2510.07242/) (2025)
- [Position: The Hidden Costs and Measurement Gaps of Reinforcement Learning with Verifiable Rewards](https://mlsanity.com/paper/2509.21882/) (2025)
- [Causally-Enhanced Reinforcement Policy Optimization](https://mlsanity.com/paper/2509.23095/) (2025)
- [CAGE: Cognitive Attribution Graphs for Faithful Inline Citation Generation in Long-Form Question Answering](https://mlsanity.com/paper/2607.24236/) (2026)
- [C2-Cite: Contextual-Aware Citation Generation for Attributed Large Language Models](https://mlsanity.com/paper/2602.00004/) (2025)
- [ProcessThinker: Enhancing Multi-modal Large Language Models Reasoning via Rollout-based Process Reward](https://mlsanity.com/paper/2606.11209/) (2026)
- [Do LLM Attribution Metrics Transfer? Auditing Retrieval-Augmented Generation Evaluation Across Datasets and Constructs](https://mlsanity.com/paper/2606.23915/) (2026)
- [Overlap, Unique and Conflict: Can LLMs Extract What They Can Recognize?](https://mlsanity.com/paper/2609.38799/) (2026)

## Abstract (arXiv)

Generating grounded and trustworthy responses remains a key challenge for large language models (LLMs). While retrieval-augmented generation (RAG) with citation-based grounding holds promise, instruction-tuned models frequently fail even in straightforward scenarios: missing explicitly stated answers, citing incorrectly, or refusing when evidence is available. In this work, we explore how reinforcement learning (RL) and internal reasoning can enhance grounding in LLMs. We use the GRPO (Group Relative Policy Optimization) method to train models using verifiable outcome-based rewards targeting answer correctness, citation sufficiency, and refusal quality, without requiring gold reasoning traces or expensive annotations. Through comprehensive experiments across ASQA, QAMPARI, ELI5, and ExpertQA we show that reasoning-augmented models significantly outperform instruction-only variants, especially in handling unanswerable queries and generating well-cited responses. A two-stage training setup, first optimizing answer and citation behavior and then refusal, further improves grounding by stabilizing the learning signal. Additionally, we revisit instruction tuning via GPT-4 distillation and find that combining it with GRPO enhances performance on long-form, generative QA tasks. Overall, our findings highlight the value of reasoning, stage-wise optimization, and outcome-driven RL for building more verifiable and reliable LLMs.