Search

Hugging Face Showcases How Test-Time Compute Scaling Can Help SLMs Outperform Larger AI Models

The researchers were able to improve the capabilities of open AI models using Google DeepMind’s study.

Advertisement
Highlights
  • Hugging Face was able to make the Llama 3B model outperform the 70B model
  • Test-time compute scaling allows models to “think longer” on problems
  • The researchers reverse-engineered closed models to develop the technique
Hugging Face Showcases How Test-Time Compute Scaling Can Help SLMs Outperform Larger AI Models

Reasoning models such as OpenAI’s o1 use test-time scaling to improve their output

Photo Credit: Hugging Face

Hugging Face shared a new case study last week showcasing how small language models (SLMs) can outperform larger models. In the post, the platform's researchers claimed that instead of increasing the training time of artificial intelligence (AI) models, focusing on the test-time compute can show enhanced results for AI models. The latter is an inference strategy that allows AI models to spend more time on solving a problem and offers different approaches such as self-refinement and searching against a verifier that can improve their efficiency.

How Test-Time Compute Scaling Works

In a post, Hugging Face highlighted that the traditional approach to improving the capabilities of an AI model can often be resource-intensive and extremely expensive. Typically, a technique dubbed train-time compute is used where the pretraining data and algorithms are used to improve the way a foundation model breaks down a query and gets to the solution.

Alternatively, the researchers claimed that focusing on test-time compute scaling, a technique where AI models are allowed to spend more time solving a problem and letting them correct themselves can show similar results.

Highlighting the example of OpenAI's o1 reasoning-focused model, which uses test-time compute, the researchers stated that this technique can let AI models display enhanced capabilities despite making no changes to the training data or pretraining methods. However, there was one problem. Since most reasoning models are closed, there is no way to know the strategies that are being used.

The researchers used a study by Google DeepMind and reverse engineering techniques to unravel how exactly LLM developers can scale test-time compute in the post-training phase. As per the case study, just increasing the processing time does not show significant improvement in outputs for complex queries.

Instead, the researchers recommend using a self-refinement algorithm that allows AI models to assess the responses in subsequent iterations and identify and correct potential errors. Additionally, using a verifier that models can search against can further improve the responses. Such verifiers can be a learned reward model or hard-coded heuristics.

More advanced techniques would involve a best-of-N approach where a model generates multiple responses per problem and assigns a score to judge which would be better suited. Such approaches can be paired with a reward model. Beam search, which prioritises step-by-step reasoning and assigning scores for each step, is another strategy highlighted by researchers.

By using the abovementioned strategies, the Hugging Face researchers were able to use the Llama 3B SLM and make it outperform Llama 70B, a much larger model, on the MATH-500 benchmark.

For the latest tech news and reviews, follow Gadgets 360 on X, Facebook, WhatsApp, Threads and Google News. For the latest videos on gadgets and tech, subscribe to our YouTube channel. If you want to know everything about top influencers, follow our in-house Who'sThat360 on Instagram and YouTube.

 
Show Full Article
Please wait...
Advertisement

Related Stories

Popular Mobile Brands
  1. Amazon Great Freedom Festival 2025 Sale to Kick Off in India on This Date
  2. iQOO Z10 Turbo+ Officially Confirmed to Run on This Chipset
  3. Huawei Band 10 Review
  4. Oppo Reno 14 5G Now Available in India in New Mint Green Colourway
  1. Meta Names ChatGPT Co-Creator Shengjia Zhao as Chief Scientist of Superintelligence Lab
  2. Who-Fi: An AI-Powered Wi-Fi Technology That Can Identify and Track Individuals Without Cameras
  3. NASA’s X-59 Moves Closer to First Flight with Advanced Taxi Tests and Augmented Vision
  4. Unusual Plasma Waves Above Jupiter’s North Pole Can Possibly Be Explained
  5. NASA to Live Stream SpaceX Crew-11 Launch Docking, Know How to Watch Online
  6. Apple Expands App Store Age Rating System With More Granular Categories
  7. Amazon Kindle Colorsoft Kids With 7-Inch Display and a Kid-Friendly Cover Launched
  8. Dying Light: The Beast Delayed to September for 'Extra Polishing Work'
  9. Flipkart Freedom Sale 2025 Date Revealed; Plus, VIP Members To Get Early Access
  10. Huawei Mate XT 2 Colour Options Leaked; Could Be World’s Second Mass-Produced Tri-Fold Phone
Gadgets 360 is available in
Download Our Apps
App Store App Store
Available in Hindi
App Store
© Copyright Red Pixels Ventures Limited 2025. All rights reserved.
Trending Products »
Latest Tech News »