Metraipolis › archive › dc_085bc39520df

PROSE: Preservation-Constrained Selective Editing for Guilty Bias in Legal Judgment Prediction — arXiv:2611.04921

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.04921
License (as archived)
CC0-1.0
Archived
2026-11-04T21:43:40.622Z
SHA-256 of the archived text
23fdd9f274743030032da7544be550887346020dccdcc447f15d867e667b5ae6
Stable URL of this page
https://metraipolis.com/documento/dc_085bc39520df

How to cite

PROSE: Preservation-Constrained Selective Editing for Guilty Bias in Legal Judgment Prediction — arXiv:2611.04921 — arXiv (en), https://arxiv.org/abs/2611.04921 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_085bc39520df (Metraipolis archive dc_085bc39520df, sha256 23fdd9f274743030032da7544be550887346020dccdcc447f15d867e667b5ae6)
BibTeX
@misc{dc085bc39520df,
  title = {PROSE: Preservation-Constrained Selective Editing for Guilty Bias in Legal Judgment Prediction — arXiv:2611.04921},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.04921},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_085bc39520df; sha256 23fdd9f274743030032da7544be550887346020dccdcc447f15d867e667b5ae6},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.04921. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Legal judgment prediction models may retain a guilty verdict after a legally decisive condition has been altered such that the specified charge is no longer established. We study this guilty bias via selective repair over matched charge-established and charge-not-established cases, where the unsupported verdict should change while its supported counterpart remains stable. We construct CAIL-MC, a legally reviewed dataset of 11,478 matched pairs covering 261 Chinese criminal charges. Held-out activation interventions show that contrasts learned from correctly resolved pairs transfer to false-guilty cases, but target-reaching interventions can also damage matched supported-guilty decisions. Motivated by this observation, we introduce PROSE, which reconstructs case-specific targets from a training-derived residual geometry, aligns them with a local margin-descent signal, attenuates supported-guilty-sensitive components, and realizes bounded targets through a preservation-regularized low-rank parameter update. Across Qwen3-8B and GLM4-9B, PROSE achieves stronger pairwise safe repair and evidence sensitivity, while preserving standard single-pass inference.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2611.04921 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-11-04T21:36:04.435Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).