test-time compute
PulseAugur coverage of test-time compute — every cluster mentioning test-time compute across labs, papers, and developer communities, ranked by signal.
-
Test-time compute boosts LLM accuracy via majority vote, verifiers, and sequential reasoning
Test-time compute strategies allow for improved accuracy in language models by increasing computational resources during inference, rather than training larger models. Methods like majority vote (self-consistency) and b…
-
AI models improve via interaction scaling, breaking compute limits
Researchers have explored a novel approach to improving AI model performance by introducing interaction scaling, a method that goes beyond traditional reasoning or sampling techniques. This third axis of compute involve…
-
Interaction Scaling: AI's Third Axis for Enhanced Test-Time Compute
Researchers have introduced a novel method called "Interaction Scaling" that enhances AI model performance by incorporating external feedback loops. Unlike traditional methods that rely solely on internal reasoning or s…
-
OpenAI model disproves 80-year-old math conjecture
An OpenAI model has successfully disproven an 80-year-old mathematical conjecture posed by Paul Erdős, known as the unit distance conjecture. The model, utilizing a "test-time compute" mechanism, explored multiple reaso…