reasoning_effort() creates a dials parameter for the reasoning effort of
reasoning models such as gpt-6-luna, with the values "low", "medium"
and "high". Use it instead of temperature() when tuning a reasoning
model.
See also
Other integrations:
as_dsprrr_metric(),
as_dsprrr_traces(),
as_ellmer_tool(),
as_vitals_cost(),
as_vitals_samples(),
as_vitals_solver(),
as_vitals_task(),
create_search_tool(),
llm_predict(),
ragnar_tool(),
register_dsprrr_engine(),
summarize_traces_df(),
temperature(),
top_p(),
use_dsprrr_template(),
validate_workflow(),
vitals_metrics
Examples
reasoning_effort()
#> Reasoning Effort (qualitative)
#> 3 possible values include:
#> 'low', 'medium', and 'high'
dials::grid_regular(reasoning_effort())
#> # A tibble: 3 × 1
#> reasoning_effort
#> <chr>
#> 1 low
#> 2 medium
#> 3 high