Metraipolis › archive › dc_186ec11ca22e
To Think or Not to Think: Pre-Decisional Reasoning Budgets for Referring Audio-Visual Segmentation — arXiv:2610.21613
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.21613
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T10:40:16.915Z
- SHA-256 of the archived text
dd857b613370df4f425243b900e57fc9b4dc502f8387a66aae114614f1bf8cf7- Stable URL of this page
https://metraipolis.com/documento/dc_186ec11ca22e
How to cite
To Think or Not to Think: Pre-Decisional Reasoning Budgets for Referring Audio-Visual Segmentation — arXiv:2610.21613 — arXiv (en), https://arxiv.org/abs/2610.21613 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_186ec11ca22e (Metraipolis archive dc_186ec11ca22e, sha256 dd857b613370df4f425243b900e57fc9b4dc502f8387a66aae114614f1bf8cf7)
BibTeX
@misc{dc186ec11ca22e,
title = {To Think or Not to Think: Pre-Decisional Reasoning Budgets for Referring Audio-Visual Segmentation — arXiv:2610.21613},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.21613},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_186ec11ca22e; sha256 dd857b613370df4f425243b900e57fc9b4dc502f8387a66aae114614f1bf8cf7},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.21613. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Multimodal reasoning systems increasingly assume that longer chain-of-thought (CoT) uniformly improves downstream grounding. We show that this assumption fails in referring audio-visual segmentation (Ref-AVS): simple queries can be harmed by excessive reasoning, creating an "overthinking trap" associated with phrase and localization drift, while genuinely ambiguous queries still benefit from multi-step reasoning. To systematically study this, we construct counterfactual reasoning-budget labels under a five-fold out-of-fold evaluation protocol. Our results recast adaptive Ref-AVS reasoning as a pre-decisional representation problem: before the first generated token, the generation-onset state contains linearly decodable information predictive of whether additional reasoning will improve segmentation. A linear probe predicts binary reasoning need with 70.1% accuracy, substantially outperforming text-only features and uncertainty-based proxy signals. As a router, the onset-state controller improves over downstream confidence routing and closely matches Short-then-Decide while using far fewer tokens. Proxy comparisons show that the onset representation contains predictive information not fully recovered by the evaluated hand-crafted features. Reusing this onset state for lightweight, single forward-pass routing reduces autoregressive token consumption by 60% while retaining about 96% of always-long performance. It further provides a cost-quality trade-off on $R^2$-AVSBench without target-label tuning, rather than an accuracy improvement over always-long. Error analysis links harmful long reasoning to phrase and localization drift, illustrating the benefit of bypassing unnecessary reasoning on simple queries. Ultimately, our results support efficient adaptive reasoning in Ref-AVS through pre-decisional internal representations. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.MM, nuovi lavori del giorno). Categorie arXiv: Multimedia (cs.MM). Pagina del lavoro: https://arxiv.org/abs/2610.21613 — 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 2026-10-16T08:53:02.654Z dalla pagina https://arxiv.org/list/cs.MM/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).