Get the latest tech news

DeepMind’s new inference-time scaling technique improves planning accuracy in LLMs


With "Mind Evolution" LLMs can use search and genetic algorithms to generate and combine different solutions and find the optimal one.

In its latest research paper, Google DeepMind introduced the concept of “ Mind Evolution,” a technique that optimizes responses of large language models (LLMs) for planning and reasoning tasks. For example, Gemini 1.5 Flash and o1-preview achieve a success rate of only 5.6% and 11.7% on TravelPlanner, a benchmark that simulates organizing a trip plan based on user preferences and constraints expressed in natural language. On the Trip Planning benchmark, which involves creating an itinerary of cities to visit with a number of days in each, Mind Evolution achieved 94.1% on the test instances while other methods reached a maximum of 77% success rate.

Get the Android app

Or read this on Venture Beat

Read more on:

Photo of DeepMind

DeepMind

Photo of LLMs

LLMs

Photo of planning accuracy

planning accuracy

Related news:

News photo

Show HN: BrowserAI – Run LLMs directly in browser using WebGPU (open source)

News photo

Europe accelerates AI drug discovery as DeepMind spinoff targets trials this year

News photo

LLMs Demonstrate Behavioral Self-Awareness [pdf]