Metraipolis › archive › dc_d4ba54fd686c

Geometry of Divergence: Tracking Hidden-State Trajectories for Adaptive Multi-Turn Reasoning — arXiv:2608.30650

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.30650
License (as archived)
CC0-1.0
Archived
2026-10-17T07:44:47.138Z
SHA-256 of the archived text
9cc5f31c405d3c90b34293c0a14f64321f37e7de70aa65cf4c9cde878b85dd80
Stable URL of this page
https://metraipolis.com/documento/dc_d4ba54fd686c

How to cite

Geometry of Divergence: Tracking Hidden-State Trajectories for Adaptive Multi-Turn Reasoning — arXiv:2608.30650 — arXiv (en), https://arxiv.org/abs/2608.30650 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d4ba54fd686c (Metraipolis archive dc_d4ba54fd686c, sha256 9cc5f31c405d3c90b34293c0a14f64321f37e7de70aa65cf4c9cde878b85dd80)
BibTeX
@misc{dcd4ba54fd686c,
  title = {Geometry of Divergence: Tracking Hidden-State Trajectories for Adaptive Multi-Turn Reasoning — arXiv:2608.30650},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.30650},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d4ba54fd686c; sha256 9cc5f31c405d3c90b34293c0a14f64321f37e7de70aa65cf4c9cde878b85dd80},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.30650. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

LLM agents need to sustain goal-consistent reasoning across long multi-turn interactions under strict resource constraints. However, as the multi-turn context accumulates, it can destabilize the underlying LLM's internal representation of task-relevant information from earlier turns, blurring the boundary between constructive reasoning and representation drift. We formulate multi-turn reasoning as a hidden-state trajectory of the underlying LLM that is characterized via two complementary signals: temporal curvature that captures the directional consistency of turn-to-turn updates, and variance slope which measures the expansion or contraction of the exploration space. Across four tasks and three underlying LLMs, we observed that these geometric signals distinguish between correct and incorrect episodes prior to completion. We further decompose each episode into three-action chains formed from four actions (Read, Write, Respond, Transfer) and show that separability is action-dependent, with different signals distinguishing various chain patterns. Our experiments demonstrate that trajectory geometry can identify critical turns in the reasoning process, increasing task success rates on $\tau$-Bench from 24.1% to 39.6% while reducing token cost by 11.2%.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2608.30650 — 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 2026-10-17T07:28:26.617Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).