ImpRIF: Stronger Implicit Reasoning Leads to Better Complex Instruction Following

2026-02-04

arXiv.org

AI Summary

This work addresses the limitations of large language models in understanding instructions requiring implicit reasoning, complex logical structures, and multiple interdependent constraints. To overcome these challenges, the authors propose a novel training framework based on reasoning graphs, which formalizes complex instructions into verifiable graph representations. By integrating graph-driven chain-of-thought fine-tuning with reinforcement learning, the approach explicitly guides the model to perform logical inference along structured reasoning paths. The method leverages synthetically generated large-scale datasets comprising both single-turn and multi-turn instructions and achieves substantial performance gains over existing baselines across five challenging instruction-following benchmarks. These results demonstrate that explicitly modeling the latent structure of implicit reasoning significantly enhances the model’s capacity for deep instruction comprehension.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

As applications of large language models (LLMs) become increasingly complex, the demand for robust complex instruction following capabilities is growing accordingly. We argue that a thorough understanding of the instruction itself, especially the latent reasoning structure embedded between the lines, is crucial for improving instruction following. Therefore we target complex instructions that involve implicit reasoning, intricate logical relations, and multi-constraint dependencies. We propose ImpRIF, a method to enhance LLMs'understanding of implicit reasoning instructions, thereby improving its ability to follow complex instructions. We formalize such instructions as verifiable reasoning graphs, enabling programmatic verification and graph-driven chain-of-thought reasoning. Based on this formulation, we synthesize large-scale single- and multi-turn data, propose fine-tuning with graph reasoning, and apply reinforcement learning to explicitly train models to reason along the graph. On five complex instruction following benchmarks, our models substantially outperform their base models. These results demonstrate that enhancing implicit reasoning capabilities can significantly improve complex instruction following. This project will be open-sourced in the near future.

Problem

• complex instruction following
• implicit reasoning
• large language models
• logical relations
• multi-constraint dependencies

Innovation

• implicit reasoning
• reasoning graphs
• complex instruction following
• graph-driven chain-of-thought
• reinforcement learning