Metraipolis › archive › dc_123fb7ff0ddd

Learning When to Think: Dual-Reference Offline Optimization for Adaptive VLM Reasoning — arXiv:2611.09154

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.09154
License (as archived)
CC0-1.0
Archived
2026-11-09T22:33:53.136Z
SHA-256 of the archived text
8b211c05587caead24e7335a8c80196de7c110e2dadc9cbb8ade43d7b4b8a67d
Stable URL of this page
https://metraipolis.com/documento/dc_123fb7ff0ddd

How to cite

Learning When to Think: Dual-Reference Offline Optimization for Adaptive VLM Reasoning — arXiv:2611.09154 — arXiv (en), https://arxiv.org/abs/2611.09154 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_123fb7ff0ddd (Metraipolis archive dc_123fb7ff0ddd, sha256 8b211c05587caead24e7335a8c80196de7c110e2dadc9cbb8ade43d7b4b8a67d)
BibTeX
@misc{dc123fb7ff0ddd,
  title = {Learning When to Think: Dual-Reference Offline Optimization for Adaptive VLM Reasoning — arXiv:2611.09154},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.09154},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_123fb7ff0ddd; sha256 8b211c05587caead24e7335a8c80196de7c110e2dadc9cbb8ade43d7b4b8a67d},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.09154. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Reasoning traces improve vision-language models on complex multimodal tasks, but forcing every input to follow a long reasoning process leads to overthinking and unnecessary inference cost. We study adaptive fast/slow thinking for VLMs, where one model answers directly for simple inputs and reasons step by step when needed. Fast answers and slow traces often come from distinct behavior policies, making a single-reference offline objective misweight one mode. We propose DUAL-REFERENCE BOLT (DR-BOLT), which distinguishes the desired fast/slow mixture reference from the sampling distribution and corrects their mismatch in reward-weighted training. Fast/slow partition competition explains routing in the optimal target; an efficient source-wise correction approximates the exact objective when response formats are well separated. Matched Qwen3-VL comparisons use identical candidate banks to isolate the objective from slow-teacher quality. DR-BOLT improves score-length trade-offs over the evaluated offline and online baselines on ScienceQA and three held-out transfer benchmarks, with a 24.1% cross-benchmark slow ratio and a 2.90-fold end-to-end training speedup over reproduced GRPO. These results support adaptive reasoning in the studied two-reference setting, while broader robustness remains open. Our code is available at this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2611.09154 — 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 2026-11-09T22:30:47.264Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).