# LegalDelta: Enhancing Legal Reasoning in LLMs via Reinforcement Learning with Chain-of-Thought Guided Information Gain

arXiv: https://arxiv.org/abs/2508.12281 · HTML: https://mlsanity.com/paper/2508.12281/

## Key facts

- Cited by 12 papers in the map (data through 14 Oct 2026).
- Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning.
- Builds on 17 of its 38 references that are in the map.
- First posted on arXiv 17 Aug 2025.

## Placed in

- LLM reasoning / [Reasoning for agents, planning and code](https://mlsanity.com/lr/reasoning-for-agents-planning-and-code/)

## Builds on (most cited first, 17 of 17)

- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [Don't Stop Pretraining: Adapt Language Models to Domains and Tasks](https://mlsanity.com/paper/2004.10964/) (2020)
- [An Empirical Study of Catastrophic Forgetting in Large Language Models During Continual Fine-tuning](https://mlsanity.com/paper/2308.08747/) (2023)
- [Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRs](https://mlsanity.com/paper/2503.01307/) (2025)
- [Large Language Models for Mathematical Reasoning: Progresses and Challenges](https://mlsanity.com/paper/2402.00157/) (2024)
- [CAIL2018: A Large-Scale Legal Dataset for Judgment Prediction](https://mlsanity.com/paper/1807.02478/) (2018)
- [Chatlaw: A Multi-Agent Legal Assistant based on a Role-Aligned Mixture-of-Experts Architecture](https://mlsanity.com/paper/2306.16092/) (2023)
- [DISC-LawLLM: Fine-tuning Large Language Models for Intelligent Legal Services](https://mlsanity.com/paper/2309.11325/) (2023)
- [Competitive Programming with Large Reasoning Models](https://mlsanity.com/paper/2502.06807/) (2025)
- [JEC-QA: A Legal-Domain Question Answering Dataset](https://mlsanity.com/paper/1911.12011/) (2019)
- [Reinforcement Learning with Verifiable Rewards: GRPO's Effective Loss, Dynamics, and Success Amplification](https://mlsanity.com/paper/2503.06639/) (2025)
- [Chinese Open Instruction Generalist: A Preliminary Release](https://mlsanity.com/paper/2304.07987/) (2023)
- [Generalist Reward Models: Found Inside Large Language Models](https://mlsanity.com/paper/2506.23235/) (2025)
- [ReasonMed: A 370K Multi-Agent Generated Dataset for Advancing Medical Reasoning](https://mlsanity.com/paper/2506.09513/) (2025)
- [Legal Mathematical Reasoning with LLMs: Procedural Alignment through Two-Stage Reinforcement Learning](https://mlsanity.com/paper/2504.02590/) (2025)
- [LegalDuet: Learning Fine-grained Representations for Legal Judgment Prediction via a Dual-View Contrastive Learning](https://mlsanity.com/paper/2401.15371/) (2024)

## Built on by (most cited first, 12 of 12)

- [LRAS: Advanced Legal Reasoning with Agentic Search](https://mlsanity.com/paper/2601.07296/) (2026)
- [LegalOne: A Family of Foundation Models for Reliable Legal Reasoning](https://mlsanity.com/paper/2602.00642/) (2026)
- [LegalGraphRAG: Multi-Agent Graph Retrieval-Augmented Generation for Reliable Legal Reasoning](https://mlsanity.com/paper/2605.28120/) (2026)
- [Can LLMs Time Travel? Enhancing Temporal Consistency in Legal Agentic Search through Reinforcement Learning](https://mlsanity.com/paper/2605.25920/) (2026)
- [The Cases LJP Never Sees: Prosecution Decision Prediction for More Complete Criminal Liability Assessment](https://mlsanity.com/paper/2605.28464/) (2026)
- [LegalDrill: Diagnosis-Driven Synthesis for Legal Reasoning in Small Language Models](https://mlsanity.com/paper/2604.23809/) (2026)
- [Relation Reasoning with LLMs in Expensive Optimization](https://mlsanity.com/paper/2605.02933/) (2026)
- [Pen-Strategist: A Reasoning Framework for Penetration Testing Strategy Formation and Analysis](https://mlsanity.com/paper/2605.04499/) (2026)
- [JPO: Juris Policy Optimization for Structured Legal Reasoning in Criminal Judgment Prediction](https://mlsanity.com/paper/2608.29616/) (2026)
- [LexReward: A Taxonomy-Driven Reward Framework for Legal Language Models](https://mlsanity.com/paper/2609.39071/) (2026)
- [Cited but Not Consulted: A Counterfactual Audit of Legal Chain-of-Thought Faithfulness](https://mlsanity.com/paper/2610.12361/) (2026)
- [LegalRewardBench: Benchmarking Reward Models for Legal Preference Evaluation](https://mlsanity.com/paper/2610.18299/) (2026)

## Abstract (arXiv)

Legal Artificial Intelligence (LegalAI) has achieved notable advances in automating judicial decision-making with the support of Large Language Models (LLMs). However, existing legal LLMs still struggle to generate reliable and interpretable reasoning processes. They often default to fast-thinking behavior by producing direct answers without explicit multi-step reasoning, limiting their effectiveness in complex legal scenarios that demand rigorous justification. To address this challenge, we propose LegalDelta, a reinforcement learning framework designed to enhance legal reasoning through chain-of-thought guided information gain. During training, LegalDelta employs a dual-mode input setup-comprising direct answer and reasoning-augmented modes-and maximizes the information gain between them. This encourages the model to acquire meaningful reasoning patterns rather than generating superficial or redundant explanations. LegalDelta follows a two-stage approach: (1) distilling latent reasoning capabilities from a powerful Large Reasoning Model (LRM), DeepSeek-R1, and (2) refining reasoning quality via differential comparisons, combined with a multidimensional reward mechanism that assesses both structural coherence and legal-domain specificity. Experimental results on multiple legal reasoning tasks demonstrate that LegalDelta outperforms strong baselines in both accuracy and interpretability. It consistently produces more robust and trustworthy legal judgments without relying on labeled preference data. All code and data will be released at https://github.com/NEUIR/LegalDelta.