HiringBranch, an AI soft skills-based hiring platform, announced a study showing its AI is 32%-50% more accurate at measuring soft skills than a majority of assessment technology on the market.

In a recent study, HiringBranch proved how much more accurate its dataset is at creating a language model that can measure soft skills—even next to Google’s large language model (LLM), the Flan T5-Large dataset. Accurately measuring soft skills is essential in a skills-based hiring economy.

The study tested open-and-closed sourced models, using rigorous iteration techniques for preparation, pre-processing, training and tuning to maximize accuracy metrics. Models were tested with inputs ranging from 66 million to 7.3 billion parameters to understand performance between small and large datasets.

Assaf Bar Moshe, PhD Scientist and Chief Research and Development Officer at HiringBranch says, “What we learned from our experiments was that LLMs do not work well enough for measuring soft skills. Overall there is a lot of hype around LLMs, and they are good for general problems but for specific problems, a dedicated corpora is needed.”

Bigger isn’t necessarily better when it comes to the dataset used for training AI algorithms. Smaller datasets can have more accurate results than larger ones with enough of the right data, Bar Moshe explains.

The implications of precise soft skill measurement are significant for candidates and employers. This study cements HiringBranch’s value in the AI-hiring marketplace, effectively demonstrating a new scientific level of soft skill measurement is possible. From this perspective, HiringBranch is unlocking unprecedented quality in soft skill measurement for the global marketplace.

“HiringBranch is unlocking unprecedented quality in soft skill measurement for the global marketplace”

As skills become the new resume, this study proves to employers that superior candidate evaluation is possible. Accurate soft skill measurement can lead to better quality throughput, better job matching, and improved hiring performance. Already more than one million candidates have taken the HiringBranch assessment, benefiting from a fairer hiring process, and yielding more effective hiring results for recruiters. The HiringBranch team is continuously testing and improving the accuracy of its AI algorithms using the most advanced data science techniques available.

Explore HRTech News for the latest Tech Trends in Human Resources Technology.