Research
R2IF: Aligning Reasoning with Decisions via Composite Rewards for Interpretable LLM Function Calling
arXiv:2604.20316v1 Announce Type: new Abstract: Function calling empowers large language models (LLMs) to interface with external tools, yet existing RL-based approaches suffer from misalignment betwe
arXiv:2604.20316v1 Announce Type: new Abstract: Function calling empowers large language models (LLMs) to interface with external tools, yet existing RL-based approaches suffer from misalignment between reasoning processes and tool-call decisions. We propose R2IF, a reasoning-aware RL framework for interpretable function calling, adopting a composite reward integrating format/correctness constraints, Chain-of-Thought Effectiveness Reward (CER), and Specification-Modification-Value (SMV) reward, optimized via GRPO. Experiments on BFCL/ACEBench show R2IF outperforms baselines by up to 34.62% (Llama3.2-3B on BFCL) with positive Average CoT Effectiveness (0.05 for Llama3.2-3B), enhancing both function-calling accuracy and interpretability for reliable tool-augmented LLM deployment.
Related
- Self-Aligned Reward: Towards Effective and Efficient Reasoners
- What If Consensus Lies? Selective-Complementary Reinforcement Learning at Test Time
- LEPO: nderline{L}atent Rnderline{e}asoning nderline{P}olicy nderline{O}ptimization for Large Language~Models
- Beyond Verifiable Rewards: Rubric-Based GRM for Reinforced Fine-Tuning SWE Agents
Source: arXiv cs.LG | 2026-04-23