Metraipolis › archive › dc_eb65b9334ac3

STEVE: Stabilizing Textual Gradient-Based Prompt Optimization via Error-Driven Refinement and Regularized Verification — arXiv:2609.23716

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.23716
License (as archived)
CC0-1.0
Archived
2026-10-09T06:49:54.158Z
SHA-256 of the archived text
f6eeebbb184c74853b469d7d107c3c5ce1123fbab0b7b3bd795cec72bf27bbde
Stable URL of this page
https://metraipolis.com/documento/dc_eb65b9334ac3

How to cite

STEVE: Stabilizing Textual Gradient-Based Prompt Optimization via Error-Driven Refinement and Regularized Verification — arXiv:2609.23716 — arXiv (en), https://arxiv.org/abs/2609.23716 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_eb65b9334ac3 (Metraipolis archive dc_eb65b9334ac3, sha256 f6eeebbb184c74853b469d7d107c3c5ce1123fbab0b7b3bd795cec72bf27bbde)
BibTeX
@misc{dceb65b9334ac3,
  title = {STEVE: Stabilizing Textual Gradient-Based Prompt Optimization via Error-Driven Refinement and Regularized Verification — arXiv:2609.23716},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.23716},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_eb65b9334ac3; sha256 f6eeebbb184c74853b469d7d107c3c5ce1123fbab0b7b3bd795cec72bf27bbde},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.23716. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Textual-gradient methods automate prompt optimization through natural-language feedback, but their iterative updates can be unstable. We identify two sources of this instability: noisy gradients produced from already-correct examples and over-specialization to hard cases that degrades performance on simpler inputs. We introduce STEVE, a stabilization framework with two coupled mechanisms. Error-Driven Refinement generates gradients only from incorrectly handled examples, concentrating updates on informative failures. Regularized Verification treats every update as provisional and accepts it only when improvement on hard cases does not cause unacceptable regression on a preservation set. Across ten reasoning benchmarks, three evaluator/optimizer models, and established prompt-optimization baselines, STEVE reduces degradation and produces more robust prompts. Additional evaluations with gpt-5.4-mini/gpt-5.4 on symbolic reasoning, GSM8K-Platinum, and DS-1000 show that these gains persist with newer models and larger test sets. STEVE therefore provides a practical way to improve the stability and effectiveness of textual-gradient prompt optimization.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-22 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.23716 — 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 2026-10-09T06:48:45.003Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-22?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).