Metraipolis › archive › dc_e2a4f08eceac
Jumping the Line: Exploiting Length Predictions in LLM Scheduling — arXiv:2610.03430
- Source
- arXiv
- Original
- https://arxiv.org/abs/2610.03430
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-05T22:49:39.838Z
- SHA-256 of the archived text
fc11a544b989322fadd1efa10ea0424ac0cc64b1e213e56e9b01b84b2aa1c42a- Stable URL of this page
https://metraipolis.com/documento/dc_e2a4f08eceac
How to cite
Jumping the Line: Exploiting Length Predictions in LLM Scheduling — arXiv:2610.03430 — arXiv, https://arxiv.org/abs/2610.03430 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_e2a4f08eceac (Metraipolis archive dc_e2a4f08eceac, sha256 fc11a544b989322fadd1efa10ea0424ac0cc64b1e213e56e9b01b84b2aa1c42a)
BibTeX
@misc{dce2a4f08eceac,
title = {Jumping the Line: Exploiting Length Predictions in LLM Scheduling — arXiv:2610.03430},
howpublished = {arXiv},
url = {https://arxiv.org/abs/2610.03430},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_e2a4f08eceac; sha256 fc11a544b989322fadd1efa10ea0424ac0cc64b1e213e56e9b01b84b2aa1c42a},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.03430. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Efficient request scheduling is increasingly important for reducing completion time in large language model (LLM) serving. Size-based policies such as Shortest Job First prioritize shorter requests, but output lengths are unknown before generation, so practical schedulers rely on predicted lengths. We introduce JIL, an attack on prediction-based LLM schedulers that manipulates the scheduling signal to obtain higher priority and reduce completion time. Using TRAIL as a case study, JIL optimizes an adversarial suffix that causes a lightweight output-length probe to underestimate a request's length. We evaluate JIL on two datasets and four LLMs across varied request profiles and deployment configurations. JIL reduces predicted output lengths by up to 83.4 percent, and adversarial requests complete up to 1.53 times faster on average in end-to-end serving experiments. The reduction in predicted length is substantially larger than the change in actual output length, revealing a mismatch between the scheduler's estimate and the request's realized size. Response utility varies across models and tasks, exposing a trade-off between scheduling advantage and response quality. We also evaluate scheduler-side defenses and find that grouping length predictions into coarse intervals reduces JIL's scheduling advantage and mitigates delays to benign requests. 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/2610.03430 — 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 2026-10-05T22:45:55.700Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).