Metraipolis › archive › dc_fbdfc5cacf76

Acquire, Repair, Preserve: A Diagnosis-Guided Post-Training Recipe for Small-Model Dialogue Game Agents — arXiv:2608.28458

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.28458
License (as archived)
CC0-1.0
Archived
2026-10-17T10:56:37.767Z
SHA-256 of the archived text
dfd7c99b3c7c64b94f2055d72d17e2bbe2789233e6f2e7db1d6e91b984590f03
Stable URL of this page
https://metraipolis.com/documento/dc_fbdfc5cacf76

How to cite

Acquire, Repair, Preserve: A Diagnosis-Guided Post-Training Recipe for Small-Model Dialogue Game Agents — arXiv:2608.28458 — arXiv (en), https://arxiv.org/abs/2608.28458 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_fbdfc5cacf76 (Metraipolis archive dc_fbdfc5cacf76, sha256 dfd7c99b3c7c64b94f2055d72d17e2bbe2789233e6f2e7db1d6e91b984590f03)
BibTeX
@misc{dcfbdfc5cacf76,
  title = {Acquire, Repair, Preserve: A Diagnosis-Guided Post-Training Recipe for Small-Model Dialogue Game Agents — arXiv:2608.28458},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.28458},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_fbdfc5cacf76; sha256 dfd7c99b3c7c64b94f2055d72d17e2bbe2789233e6f2e7db1d6e91b984590f03},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.28458. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Interactive dialogue games test a capability that static benchmarks largely leave implicit: a model must carry state across turns, interpret feedback, and choose valid actions under changing constraints. We study this setting in the LM Playschool Challenge with a 2B open-weight model, and find that many failures are not only broad knowledge failures but also local decision failures: repeated guesses, malformed actions, and violations of feedback that the model has just seen. These diagnostics motivate a training recipe organized around three steps: acquire broad game participation through supervised fine-tuning, repair mechanically verifiable failures within one targeted dialogue-game family using turn-local preference pairs, and preserve general capabilities beyond these dialogue games. In the official final evaluation, our submission improves public clemscore from 10.67 to 38.92 and closed in-domain score from 13.41 to 41.17, while approximately preserving aggregate static performance (44.14 vs. 44.24 for the baseline). Out-of-domain clemscore remains low at 7.88, with the largest gains concentrated in unseen variants of the targeted family. Our results suggest that broad SFT brings most of the model's capability improvement; turn-local supervision can be effective when failure detection is precise, with observed transfer concentrated primarily within-family.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-31 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.28458 — 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 2026-10-17T10:47:33.140Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-31?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).