Menu
Home
People
Places
Arts
History
Plants & Animals
Science
Life & Culture
Technology
Reference.org
Reasoning language model
open-in-new
<p>Reasoning language models (RLMs) are <a href="/facts/Large_language_model/WnogWVJY">large language models</a> that have been further trained to solve multi-step <a href="/facts/Reasoning/tcBImN8J">reasoning</a> tasks. These models perform better on logical, mathematical or programmatic tasks than traditional autoregressive LLMs, have the ability to <a href="/facts/Backtracking/Jm9vDUk8">backtrack</a>, and employ test-time compute as an additional <a href="/facts/Neural_scaling_law/B372rC8f">scaling axis</a> beyond <a href="/facts/Training%2c_validation%2c_and_test_data_sets/V3mMQb2G">training examples</a>, parameter count, and train-time compute. </p>