data_juicer.ops.mapper.dialog_error_recovery_mapper module

LLM: reply after user disputes a prior error (1–5).

Inspired by OpenJudge self-correction; DJ: no dispute context → neutral high.

Reference: https://agentscope-ai.github.io/OpenJudge/built_in_graders/multi_turn/ #selfcorrectiongrader

class data_juicer.ops.mapper.dialog_error_recovery_mapper.DialogErrorRecoveryMapper(*args, **kwargs)

Bases: _DialogTurnQualityMapper

When the user disputes a prior assistant mistake, is the reply corrective.

OP_NAME = 'dialog_error_recovery_mapper'

META_KEY = 'dialog_error_recovery'