Metraipolis › archive › dc_025e6ea18444

Amplified Does Not Mean Predictive: Reasoning Behaviors in Thinking Models — arXiv:2608.13760

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.13760
License (as archived)
CC0-1.0
Archived
2026-10-19T02:21:05.792Z
SHA-256 of the archived text
e8786f7f3535e47075d68f0b2e68b0be67904a566e33370793ccfac85b60d352
Stable URL of this page
https://metraipolis.com/documento/dc_025e6ea18444

How to cite

Amplified Does Not Mean Predictive: Reasoning Behaviors in Thinking Models — arXiv:2608.13760 — arXiv (en), https://arxiv.org/abs/2608.13760 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_025e6ea18444 (Metraipolis archive dc_025e6ea18444, sha256 e8786f7f3535e47075d68f0b2e68b0be67904a566e33370793ccfac85b60d352)
BibTeX
@misc{dc025e6ea18444,
  title = {Amplified Does Not Mean Predictive: Reasoning Behaviors in Thinking Models — arXiv:2608.13760},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.13760},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_025e6ea18444; sha256 e8786f7f3535e47075d68f0b2e68b0be67904a566e33370793ccfac85b60d352},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.13760. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Which reasoning behaviors are associated with correct answers in reasoning models, and does reasoning-oriented training amplify those behaviors? This distinction is important because reasoning-oriented training can make traces look more deliberative without amplifying the behaviors most tied to model correctness. We quantify this mismatch with Behavioral Lift, a metric that measures how much correctness changes when a behavior is present versus absent in a model's reasoning trace. Across 15 models and 6 benchmarks spanning text-only and vision-language reasoning, we annotate 15,282 traces with a taxonomy whose core behaviors are defined for both LLM and VLM traces. We find evidence for an Amplification-Lift Gap, in which thinking models strongly amplify self-correction, hypothesis testing, and uncertainty acknowledgment, while the highest-lift behaviors are confidence calibration, knowledge alignment, and self-awareness. Confidence calibration is among the strongest positive signals of correctness in both modalities, yet is barely amplified; uncertainty acknowledgment is amplified by 3--7$\times$, yet is weakly or negatively associated with correctness. We find that reasoning-oriented training does not preferentially amplify the highest-Lift behaviors, motivating process-level objectives that reward calibrated and grounded reasoning rather than surface form alone.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-17 in cs.AI).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.13760 — 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 2026-10-19T02:02:02.866Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-17?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).