The new SMITH framework enables small models to approach, and even surpass, large models with fewer token, paving the way for scalable multi-agent collaboration.

Singapore, September 30th / PRNewswire / -- Appier (Tokyo Stock Exchange: 4180) is a native company that provides Agentic AI as a Service ( AaaS ). The company announced today that its latest research paper, " Joint Optimization of Tool Creation and Use for Large Language Model Agents ", has been accepted by NeurIPS. NeurIPS is a top global conference on artificial intelligence and machine learning, often referred to as the " AI Olympics". This paper proposes a SMITH ( Schema-grounded Multi-task Iterative Tool Honing ) framework, which is a reinforcement learning framework that allows AI to build tools within the same training cycle and effectively use these tools, with each tool continuously improving based on the results of solving real problems. This breakthrough addresses a key challenge in Agentic AI: models can create tools, but it is often difficult to make good use of these tools.

Research shows that small models trained using SMITH are capable of building reusable tools that can match the quality of those created by larger models, even on tasks that the smaller models have never encountered before. These tools can also be shared across different models and tasks, significantly reducing the token cost of repeated inference. The paper was accepted by NeurIPS, highlighting Appier's research strength in improving AI proxy efficiency and advancing multi-agent collaboration, as well as strengthening its position at the forefront of global Agentic AI research.

Dr. Zhou Xihan, CEO and Co-founder, stated: "Humans transform the experience of solving problems into tools, so there is no need to start from scratch every time." Agents are also evolving in the same way nowadays. Our research shows that agents can learn to build tools, continuously optimize them, and share proven tools across models of different scales, thereby making multi-agent collaboration more efficient and scalable. Accepting this paper further recognizes our future-oriented research and innovation. We will continue to bring our technology into real-world applications and deliver measurable results for businesses.

AI Agents must be able to build tools and know whether the tools are effective.

As Agentic AI takes on more and more autonomous handling of enterprise workflows, selecting and invoking appropriate external tools has become key to successful deployment. However, many AI systems still rely on engineers to pre-build API or configure fixed tools, which must be rebuilt whenever the data sources, tasks, or business requirements change. Even though AI can create their own tools, the existing methods typically assign "creation" and "usage" to different teams. As a result, the models that build these tools receive little feedback from real-world usage, and it is difficult to determine whether a tool's description is clear, whether it runs reliably, or whether it can be correctly invoked by other models.

SMITH incorporates these two capabilities into the same training loop, which is why AI is able to learn how to build good tools and use them effectively at the same time. When the tool descriptions are vague, the parameter design is poor, or the operations fail, these outcomes are fed back to the model. The processes of building, using, verifying, and optimizing form a closed loop that continuously improves the quality of the tools. Research has found that training tool creation and tool usage together is significantly more effective than training them separately.

Research scientist Lin Jieyan stated: "When SMITH uses tools to train models, it only sees the descriptions and parameter specifications of the tools, not the underlying code. This means that whether each description is clear and whether the tools can be called correctly will result in direct feedback during training. Our experiments have also confirmed that other models can use these tools to solve problems more effectively. Looking to the future, we hope to build models that can continuously interact with the environment and undertake a wider range of tasks."

Learn from 4 examples and verify on 16 new problems.

SMITH adopts a training approach that progresses from easy to difficult. AI starts by learning a method from 4 simple examples and uses it to build a tool. Subsequently, the model is tested on 16 more difficult problems that have not been seen before to see if it can generalize. Only tools that can solve new problems are retained and added to a shared tool library for use by multiple AI agents. As more high-quality tools are added, better tools will replace the weaker ones. Agents do not need to rebuild tools for each task; instead, they can build and reuse proven problem-solving capabilities.

This study presents three key findings:

• Small models can outperform large models in tool creation: A model with approximately 4 billion parameters, trained using SMITH, built tools that outperformed all other methods studied in the experiments. It even surpassed a baseline approach involving the immediate construction of tools using a model with about 30 billion parameters. Effective tool creation does not necessarily depend on larger models.
• Verified tools can be used across different model scales: Tools built with smaller models can handle new tasks well even when used by lightweight models with only about 350 million parameters. The same tools have also improved the performance of larger models, enabling agents to divide labor more flexibly and efficiently.
• Repetitive reasoning can be transformed into a reusable tool: SMITH converts repetitive reasoning into tools that can be directly called. In experiments, the average output of traditional step-by-step reasoning decreased from 3,206 token to about 100 token, representing an efficiency improvement of approximately 32 times. When facing similar problems, AI does not need to execute the entire reasoning process every time, allowing for faster reasoning while maintaining task performance and reducing computational costs.

This research has opened up a new direction for the company Agentic AI. Many daily operations are repetitive, such as converting financial indicators, processing data, querying reports, checking rules, and diverting customer service tickets. AI can transform these scattered methods into verified, shareable tools that can be called by any agent, thereby reducing the costs of repeated development and reasoning.

In the fields of advertising and marketing, agencies that handle customer data, provide personalized recommendations, offer customer service, and manage ad purchases can share validated tools and consistent business rules, thereby collaborating more efficiently. Whether entering new markets, setting up new advertisers, or starting out with limited data, companies can transform past successes into verifiable and scalable AI capabilities, helping agencies to adapt to new tasks more quickly. Appier will continue to advance Agentic AI through forward-looking research, making AI a core engine for long-term business growth.

About Appier

Appier (Tokyo Stock Exchange: 4180) is a AI native Agentic AI as Service (AaaS) company that empowers business decisions with advanced AdTech and MarTech solutions. Founded in 2012, the company's vision is to "Making AI Easy by making software intelligent", and it is committed to helping enterprises transform AI into ROI through Ad Cloud, Personalization Cloud, and Data Cloud solutions. Currently, Appier has 17 offices across the Asia-Pacific, United States, Europe, Middle East, and Africa regions, and is listed on the Tokyo Stock Exchange. For more company information, please visit www.appier.com; for more investor relations information, please visit ir.appier.com / en.