papers / latency-budget

Latency Budget: Grounded Answers Under Explicit Deadlines

Pooria Arab · orcid 0009-0005-9964-6841 · 2026-10-29 · Published on Zenodo

abstract

An answer can be grounded and still arrive too late to be useful. Latency Budget measures this distinction on 36 synthetic directory, calendar, and document tasks. Credit requires a single correct candidate, replayed evidence, and completion within an explicit deadline. We run 1656 local episodes across 23 repeated conditions without paid inference. The privileged oracle obtains 100.0 [90.4, 100.0] percent deadline credit (95% Wilson interval); each blind control obtains 0.0 [0.0, 9.6] percent. An additional simulated factorial experiment varies declared thinking time and redundant tool calls while holding task deadlines fixed. At 80 ms per turn, one redundant call reduces deadline credit to 75.0 [58.9, 86.2] percent, while grounded correctness remains 100.0 [90.4, 100.0] percent. The experiment demonstrates sensitivity to wasted actions under the chosen cost model. It does not measure learned-agent performance, network latency, or human deadline preferences.

benchmark: benchmark-latency-budget. submitted to: none recorded.

citation

Arab, Pooria (2026). Latency Budget: Grounded Answers Under Explicit Deadlines. https://doi.org/10.5281/zenodo.23752409

bibtex

@misc{latency-budget,
  author = {Arab, Pooria},
  title = {Latency Budget: Grounded Answers Under Explicit Deadlines},
  year = {2026},
  doi = {10.5281/zenodo.23752409},
  url = {https://pooriaarab.com/papers/latency-budget},
}

venues, counts and the software records live on papers.