This study re-evaluates the use of input rewriting to improve frozen downstream models for dialogue discourse parsing under realistic deployment conditions where no clarification supervision is available. The authors find that last-utterance clarification is far less reliable than suggested by supervised settings, as parser-agnostic rewriting often introduces more regressions than repairs.
- Across three Segmented Discourse Representation Theory (SDRT) datasets and multiple parsers, the analysis reveals that edits enabling fixes frequently disrupt discourse cues relied upon by the parser.
- A best-of-8 rewriting analysis shows a practical ceiling where a large fraction of errors are not repairable through input rewriting alone.
- A parser-aware clarifier trained with GRPO reduces regressions by up to 37% by learning conservative abstention, yet still fails to produce selectivity-aware clarifications that consistently improve parsing.
The findings recast clarification as a selective intervention problem and identify rewritability prediction as the key missing capability for input-side optimization of frozen discourse parsers.