Papers.Data.Code
Dataset

DeliverBench

deliverbench

Contains 3 staged professional projects with source files, rubrics, references, and original-format attachments for persistent agent delivery tasks.

Hugging Face ↗Nov 2, 2026Code ↗

  • multimodal
  • agents
  • document-generation
  • benchmark
Pipeline diagram contrasting typical setup with DeliverBench's continuous execution and evaluation framework
Pipeline diagram contrasting typical setup with DeliverBench's continuous execution and evaluation framework

What it is

Benchmark of end-to-end professional delivery where one agent works through staged projects and a separate final handoff in finance, medicine, and law. 3 complete professional projects with persistent multi-stage workflows: 33 scored process stages plus 3 independent FINAL evaluations across finance, pharmacovigilance, and legal delivery.

Key results

  • 3 public projects across finance, medicine, and law
  • 33 scored process stages plus 3 FINAL units
  • Mixed document, text, and tabular files; CC BY 4.0

Useful for benchmarking agents that must revise work across stages and deliver a complete final package.

One useful ML email a week.

More from Nov 02 – Nov 08

The whole week →

Picked Nov 6

Picked Nov 5

  • Repo
    wangzan101/Mika

    Runs a cross-app Android agent entirely on-device, including background-screen execution and user handoff.

    GitHubOct 29, 2026
    →

Picked Nov 4

  • Dataset
    timmzak/ReText-Captions

    Adds text annotations to standard ReID benchmarks, making caption-conditioned and text-augmented person matching measurable on existing protocols.

    Hugging FaceOct 28, 2026
    →

Worth your time, not all of your time.

We track new ML work across papers, code, and data, then keep only the pieces worth a closer look. Selected, not collected.

Papers.Data.Code

The Weekly ML Signal

Selected, not collected.

One useful ML email a week: the papers, repos and datasets worth your attention.

Worth your time, not all of your time.