Large language models (LLMs) have gained much attention for their human imitation properties. These models are capable of answering questions, generating content, summarizing long textual paragraphs, and all that. Indications are essential to improve the performance of LLMs such as GPT-3.5 and GPT-4. The way prompts are created can have a big impact on an LLM’s skills in a variety of areas, including reasoning, multimodal processing, tool use, and more. These techniques, designed by researchers, have shown promise in tasks such as model distillation and simulation of agent behavior.
Manual engineering of rapid approaches raises the question of whether this procedure can be automated. By producing a set of prompts based on input and output instances of a data set, the Automatic Prompts Engineer (APE) attempted to address this issue, but APE had diminishing returns in terms of prompt quality. Researchers have suggested a method based on an evolutionary algorithm that maintains diversity for self-referential self-improvement of prompts for LLMs to overcome diminishing returns in prompt creation.
LLMs can modify their cues to improve their capabilities, just as a neural network can change its weight matrix to improve performance. Based on this comparison, LLMs can be created to improve both their own capabilities and the processes by which they improve them, thus allowing artificial intelligence to continue improving indefinitely. In response to these ideas, a team of Google DeepMind researchers introduced PromptBreeder (PB) in recent research, which is a technique for LLMs to improve themselves in a self-referential manner.
The PB requires a domain-specific problem description, a set of initial mutation cues, which are the instructions for modifying a task prompt, and thinking styles, that is, the generic cognitive heuristic in text form. By using the LLM’s ability to act as mutation operators, it generates different task prompts and mutation prompts. The suitability of these evolved task cues is evaluated in a training set, and a subset of evolutionary units comprising task cues and their associated mutation cues are selected for future generations.
The team has shared that PromptBreeder observes cues that fit the particular domain over several generations. For example, PB developed a task with explicit instructions on how to approach mathematical questions in the field of mathematics. On a variety of benchmark tasks, including common sense reasoning, arithmetic, and ethics, PB outperforms state-of-the-art cueing techniques. PP does not require parameter updates for self-referential self-improvement, suggesting a potential future in which longer and more capable LLMs may benefit from this strategy.
The working process of PromptBreeder can be summarized as follows:
- Task Prompt Mutation: Task Prompts are prompts created for certain tasks or domains. PromptBreeder starts with a population of these prompts. The task prompts then undergo mutations, resulting in variants.
- Fitness Evaluation: Using a training data set, the fitness of these modified task prompts is evaluated. This assessment measures how well the LLM responds to these variations when requested.
- Continuous evolution: Similar to biological evolution, the process of mutation and evaluation is repeated over several generations.
In summary, PromptBreeder has essentially been touted as a unique and successful technique for the autonomous evolution of indications for LLM. It attempts to improve the performance of LLMs across a variety of tasks and domains, ultimately outperforming manual cueing methods by iteratively improving both task cueing and mutation cueing.
Review theai.meta.com/research/publications/effective-long-context-scaling-of-foundation-models/”> Paper. All credit for this research goes to the researchers of this project. Also, don’t forget to join. our 31k+ ML SubReddit, Facebook community of more than 40,000 people, Discord channel, and Electronic newsletterwhere we share the latest news on ai research, interesting ai projects and more.
If you like our work, you’ll love our newsletter.
Now we are also on WhatsApp. Join our ai channel on Whatsapp.
Tanya Malhotra is a final year student of University of Petroleum and Energy Studies, Dehradun, pursuing BTech in Computer Science Engineering with specialization in artificial intelligence and Machine Learning.
She is a Data Science enthusiast with good analytical and critical thinking, along with a burning interest in acquiring new skills, leading groups and managing work in an organized manner.
<!– ai CONTENT END 2 –>
