Skip to main content
Back to challenges
Challenge

Build an AutoGen Multi-Agent Image Forensics Inspector with DeepSeek R1

Widespread image duplication and validation anomalies undermine published biomedical scientific papers. Develop an AutoGen multi-agent system powered by DeepSeek R1 reasoning models to audit Western blot figures, cross-check metadata, and output reproducible forensic audit reports.

Data ScienceHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Deploy an AutoGen multi-agent system using DeepSeek R1 reasoning capabilities to detect image manipulation and fraud in scientific literature.

Delivery guide

How work is evaluated

Evaluation

Tests AutoGen multi-agent system detection of spliced or altered blot image samples.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

manipulation_detection_accuracy

Ensures true positive manipulation detection score exceeds cutoff threshold.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

forensic_f1_score

F1 score across modified and pristine scientific image test sets. • target: 0.9 • range: 0-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Configure Microsoft AutoGen multi-agent workflows with specialized persona roles

  • Utilize DeepSeek R1 reasoning capabilities to analyze image duplication and feature splicing

  • Build visual contrast analysis tools to identify altered antibody validation figures

  • Generate standardized forensic inspection reports detailing manipulation likelihood scores

Resources and assets

Reference links and supporting material

Dataset notes

Dataset of 50 scientific figure panels consisting of 25 unaltered original images and 25 synthetic/spliced images with manipulation ground truth.

Challenge setup

Tests AutoGen multi-agent system detection of spliced or altered blot image samples.

Challenge input

JSON containing base64 image data or file path and figure caption text.

AutoGen

Orchestrates conversational sub-agents.

DeepSeek R1

Provides deep step-by-step reasoning for image forensics.

Evaluated output

JSON containing manipulation_detected (boolean), confidence_score (float), and flagged_regions (list).

Evaluation checks
  • Ensures true positive manipulation detection score exceeds cutoff threshold.
  • F1 score across modified and pristine scientific image test sets. • target: 0.9 • range: 0-1
Configured evaluation evidence
  • Benchmark: Biomedical Image Integrity Benchmark (BIIB)
  • Forensic F1 Score target: 0.9
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

Action Space
  • DeepSeek · Optional
  • AutoGen · Optional
Policy Serving
  • DeepSeek R1 · Required

Evaluation contract

  • manipulation_detection_accuracy · Weight 1
  • forensic_f1_score · Weight 1

Metrics

  • forensic_f1_score · Target 0.9

Benchmarks

  • Biomedical Image Integrity Benchmark (BIIB)
  • biotech

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing base64 image data or file path and figure caption text.

    Output: JSON containing manipulation_detected (boolean), confidence_score (float), and flagged_regions (list).

    Execution harness: Python

Evaluation tools

  • AutoGen · Multi-Agent Framework
  • DeepSeek R1 · Reasoning LLM

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Build an AutoGen Multi-Agent Image Forensics Inspector with DeepSeek R1