Appier Advances AI Capabilities to Recognize Limits and Choose Reasoning Language

Appier, an AI-native company delivering Agentic AI as a Service (AaaS), has made significant strides in advancing the capabilities of artificial intelligence (AI). The latest research from Appier's AI Research team focuses on enabling AI to recognize when it lacks sufficient information and to choose the appropriate reasoning language for different tasks. These advancements are crucial for ensuring the reliability and global deployment of enterprise AI.

The research, published in two papers, delves into the challenge of AI recognizing information gaps and selecting the right reasoning language. The first paper, "None of the Above, Less of the Right: Parallel Patterns between Humans and LLMs on Multi-Choice Questions Answering," explores how large language models (LLMs) can identify when there is insufficient information to provide a valid answer. The study tested 28 leading LLMs and found that their accuracy significantly declined when the correct response was "none of the above." To address this, Appier's team employed training methods such as Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO), which improved model accuracy in identifying questions with no correct answer by nearly 30 percentage points.

The second paper, "Language Matters: How Do Multilingual Input and Reasoning Paths Affect Large Reasoning Models?" investigates the impact of the language used in reasoning models on their performance. The research revealed that models often default to high-resource languages such as English for reasoning, even when prompted in another language. This can lead to suboptimal results, particularly in tasks requiring cultural understanding. Appier's team found that using a "text prefilling" technique, setting an opening phrase that prompts the model to reason in a specified language, could significantly enhance performance in tasks where cultural context is crucial.

These findings highlight the need for AI systems to be capable of recognizing their limitations and to select the most appropriate reasoning approach and language for each task. In an e-commerce setting, for instance, an AI agent must be able to recognize when a product is not covered by an existing return policy and to choose the reasoning language that best captures the nuances of local markets. Appier's research underscores the importance of these capabilities in ensuring that AI systems provide accurate and reliable information, which is essential for enterprise decision-making.

Chih Han Yu, CEO and Co-founder of Appier, stated, "These two papers redefine the standard for evaluating AI. As AI moves from answering questions to making autonomous decisions, it must recognize insufficient information, adjust its actions accordingly, and select the reasoning approach best suited to each task. These capabilities will help Agentic AI evolve into a reliable decision-making system capable of navigating real-world complexity."

Appier's continued research into large language models and Agentic AI aims to enable AI to not only act autonomously but also to make more reliable judgments based on the available information, task requirements, and market context. This will help enterprises turn Agentic AI into scalable, tangible business value.