Using Contrastive Learning to Improve Two-Way Reasoning in Large Language Models: The Obfuscation Task as a Case Study

2025-09-05

AI Summary

This work investigates whether large language models possess genuine conceptual understanding, proposing “bidirectional reasoning”—generalization between forward and inverse tasks without reverse fine-tuning—as a core diagnostic criterion. We empirically discover, for the first time, that standard forward fine-tuning induces “cognitive specialization”: while forward-task performance improves, inverse reasoning degrades significantly. To address this, we introduce Contrastive Fine-Tuning (CFT), a framework that jointly optimizes semantic-preserving positive pairs, semantically divergent negative pairs, and forward-confusing examples—thereby implicitly modeling bidirectional mapping relationships. Experiments demonstrate that CFT substantially enhances inverse reasoning capability without compromising forward-task accuracy, enabling bidirectional reasoning to emerge naturally. This work establishes a novel, scalable benchmark and methodology for rigorously evaluating conceptual understanding in language models.

Technology Category

• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning
• Natural Language Processing: (Large) Language Models
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

This research addresses a fundamental question in AI: whether large language models truly understand concepts or simply recognize patterns. The authors propose bidirectional reasoning,the ability to apply transformations in both directions without being explicitly trained on the reverse direction, as a test for genuine understanding. They argue that true comprehension should naturally allow reversibility. For example, a model that can change a variable name like userIndex to i should also be able to infer that i represents a user index without reverse training. The researchers tested current language models and discovered what they term cognitive specialization: when models are fine-tuned on forward tasks, their performance on those tasks improves, but their ability to reason bidirectionally becomes significantly worse. To address this issue, they developed Contrastive Fine-Tuning (CFT), which trains models using three types of examples: positive examples that maintain semantic meaning, negative examples with different semantics, and forward-direction obfuscation examples. This approach aims to develop deeper understanding rather than surface-level pattern recognition and allows reverse capabilities to develop naturally without explicit reverse training. Their experiments demonstrated that CFT successfully achieved bidirectional reasoning, enabling strong reverse performance while maintaining forward task capabilities. The authors conclude that bidirectional reasoning serves both as a theoretical framework for assessing genuine understanding and as a practical training approach for developing more capable AI systems.

Problem

• Testing whether language models truly understand concepts versus pattern recognition
• Addressing performance degradation in bidirectional reasoning after fine-tuning
• Developing methods to achieve reversible transformations without explicit reverse training

Innovation

• Contrastive Fine-Tuning (CFT) method
• Bidirectional reasoning without reverse training
• Positive-negative contrastive learning examples