Metrics for one convergence attempt. More...

Public Attributes
bool	converged = false
	Whether the final harness converged (compile+validate passed).

int	final_warnings = 0
	Final validator warning count.

bool	meets_spec = false
	Whether the final harness meets all expected properties.

FeedbackMode	mode = FeedbackMode::FULL_DIAGNOSTIC
	Which feedback mode was used.

std::string	task_name
	Task that was being solved.

int	total_errors = 0
	Number of compile+validate errors across all turns.

int	total_estimated_tokens = 0
	Estimated total tokens consumed across all turns.

int	turns = 0
	How many LLM turns were needed to converge (or max).

Detailed Description

Metrics for one convergence attempt.

Definition at line 92 of file harness_bench.h.

The documentation for this struct was generated from the following file:

Public Attributes