Genomic Patent Prior Art and Infringement Classifier with AutoGen
IP disputes in single-cell sequencing require rigorous multi-agent analysis of patent claims against scientific publications. Build a conversational multi-agent audit team using Microsoft AutoGen to parse claim charts and highlight overlapping method steps. Achieve 90% precision on patent claim overlap extraction.
What you are building
The core problem, expected build, and operating context for this challenge.
Orchestrate a multi-agent patent audit system using Microsoft AutoGen to evaluate genomic tech claims against prior art.
How work is evaluated
Evaluates accuracy in detecting matching elements between single-cell patent claims and prior art protocols.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
risk_level_format
Risk level must be HIGH, MEDIUM, or LOW.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
claim_overlap_precision
Precision of identified claim limitations matching prior art text • target: 0.9 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Implement multi-agent conversation topologies using Microsoft AutoGen framework
Extract structured patent claim elements from USPTO documentation
Cross-examine genomic library preparation protocol steps against prior art literature
Generate evidence reports highlighting potential claim infringement and anticipation risks
Reference links and supporting material
Dataset of 50 paired single-cell RNA-seq patent claims (e.g. 10x Genomics vs Parse) and corresponding research paper methods.
Challenge setup
Evaluates accuracy in detecting matching elements between single-cell patent claims and prior art protocols.
Challenge input
JSON containing patent_claim_text and prior_art_method_text.
AutoGen
Framework specified for multi-agent group chat consensus.
Evaluated output
JSON object with overlapping_elements array, infringement_risk_level, and consensus_reasoning.
- Risk level must be HIGH, MEDIUM, or LOW.
- Precision of identified claim limitations matching prior art text • target: 0.9 • range: 0-1
- Benchmark: Biotech Patent Claim Infringement Benchmark
- Claim Overlap Precision target: 0.9
- 1 public reference case
- 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
- AutoGen · Required
- HeyBoss AI · Optional
- Qwen2.5-72B · Optional
Evaluation contract
- risk_level_format · Weight 1
- claim_overlap_precision · Weight 1
Metrics
- claim_overlap_precision · Target 0.9
Benchmarks
- Biotech Patent Claim Infringement Benchmark
- biotech
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing patent_claim_text and prior_art_method_text.
Output: JSON object with overlapping_elements array, infringement_risk_level, and consensus_reasoning.
Execution harness: Python
Evaluation tools
- AutoGen · Multi-agent Conversational Framework
Recipe freeze marker
No recipe freeze time is recorded.