When to Warn, Not What to Do: Calibrated Language Warnings for Frozen World-Action Models

Ting Wang, Yuanjie Shi, Han Wang, Yan Yan, Huan ZhangOpenReview 2027 · Submission to ICLR 2027 · Withdrawn · OpenReview

Its authors withdrew this paper, so there is no full text to read or summarize.

Abstract

World-action models (WAMs) combine future visual prediction and action generation for language-conditioned robot control. Collision safety requires anticipating unsafe contact and guiding responses that preserve task progress, yet collision records do not specify corrective actions. Learning recovery or language-feedback policies requires additional supervision or interaction. We propose risk-triggered language steering, which learns when to warn and guides a frozen WAM with a fixed prompt without corrective-action supervision. A classifier trained on collision-derived $K$-step warning labels reads risk from predicted-video and action hidden states. Unsafe-class split conformal calibration sets a threshold targeting unsafe recall under the unmodified policy. Alerts trigger regeneration of the pending chunk with the warning appended to the task instruction. Experiments with LingBot-VA on seen tasks across four SafeLIBERO simulation suites show 89.3–99.0\% AUROC and mean unsafe recall meeting nominal targets. Online, mean collision avoidance rate rises from 9.4\% to 18.8\%, versus 3.8\% with the same warning always active. Mean task success is 30.0\%, versus 29.4\% without intervention.
BibTeX
@misc{wang2027when,
  author = {Ting Wang and Yuanjie Shi and Han Wang and Yan Yan and Huan Zhang},
  title = {{When to Warn, Not What to Do: Calibrated Language Warnings for Frozen World-Action Models}},
  year = {2027},
  url = {https://openreview.net/forum?id=HIu0JHeckF},
  howpublished = {OpenReview},
  note = {Submitted to ICLR 2027},
}

Loading related content...

Figure