Metraipolis › archive › dc_d3725080fa64
DeliverBench: Benchmarking AI Agents on Delivering Realistic Multi-Stage Professional Projects — arXiv:2611.03412
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.03412
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-03T18:15:39.247Z
- SHA-256 of the archived text
b593d5e2d0e9d378f975a8fdea29d980d009ba9e5ec1abe5c1a0a748c11cc7f2- Stable URL of this page
https://metraipolis.com/documento/dc_d3725080fa64
How to cite
DeliverBench: Benchmarking AI Agents on Delivering Realistic Multi-Stage Professional Projects — arXiv:2611.03412 — arXiv (en), https://arxiv.org/abs/2611.03412 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d3725080fa64 (Metraipolis archive dc_d3725080fa64, sha256 b593d5e2d0e9d378f975a8fdea29d980d009ba9e5ec1abe5c1a0a748c11cc7f2)
BibTeX
@misc{dcd3725080fa64,
title = {DeliverBench: Benchmarking AI Agents on Delivering Realistic Multi-Stage Professional Projects — arXiv:2611.03412},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.03412},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d3725080fa64; sha256 b593d5e2d0e9d378f975a8fdea29d980d009ba9e5ec1abe5c1a0a748c11cc7f2},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.03412. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
LLM agents are judged by what they deliver, and in professional work what gets delivered is a project, not a task. A project pays off only if the whole package holds, and one weak part can limit the value of the rest. Yet existing benchmarks mostly test agents on one task at a time, leaving a gap between the capability we measure and the outcome that makes agents useful. To address this gap, we introduce DeliverBench, which tests whether agents can carry a project from brief to final delivery. Its 11 projects, spanning financial due diligence, legal contract drafting, and drug safety reporting, are reconstructed with 13 domain experts from real materials and chain 150 subtasks through 2-15 rounds of new materials and client feedback. Because no reference outputs are injected, agents inherit their own errors from one task to the next. Expert rubrics then score every checkpoint as well as the final package. Across seven model-harness systems, even the best system scores only 65.0 out of 100 on final delivery and passes 50.0% of subtasks at a 90-point threshold. The failures are often failures of carrying work forward: in one audit project, all seven systems addressed the review findings, but only one preserved traceability. Together, these results suggest that task-level capability does not simply compose into delivery, which also depends on keeping the work coherent as it evolves. 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/2611.03412 — 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 2026-11-03T18:09:53.458Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).