AutoGen & Cerebras Inference Sale-Leaseback High-Speed Multi-Agent Simulator
Institutional investors negotiating sale-leasebacks need fast, multi-turn financial negotiation simulations under dynamic yield constraints. Construct an AutoGen multi-agent system powered by ultra-fast Cerebras Inference to simulate landlord-tenant lease terms in sub-second roundtrips.
What you are building
The core problem, expected build, and operating context for this challenge.
Build an AutoGen multi-agent negotiation simulator powered by Cerebras Inference for high-speed institutional sale-leaseback modeling.
How work is evaluated
Evaluates speed and financial equity of negotiated sale-leaseback agreements.
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.
boundary_constraint_check
Ensures final rent does not exceed tenant max and cap rate satisfies landlord min.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
latency_ms_per_turn
Average millisecond inference latency per agent response turn. • target: 300 • range: 0-1500
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Configure AutoGen UserProxyAgent and AssistantAgent pairs for lease negotiations
Integrate Cerebras Inference API endpoint for low-latency multi-turn conversations
Implement lease agreement convergence logic (Cap rate, escalation rate, lease term)
Measure total conversation completion latency and financial target optimization
Reference links and supporting material
50 commercial property financial structures used to seed buyer-seller negotiation parameters.
Challenge setup
Evaluates speed and financial equity of negotiated sale-leaseback agreements.
Challenge input
JSON containing property asset value, minimum landlord cap rate, maximum tenant rent
AutoGen
Provides multi-agent conversation framework and group chat orchestration.
Cerebras Inference
Offers ultra-low latency token generation for real-time agent loops.
Evaluated output
JSON with agreed_cap_rate, initial_rent, escalation_rate, total_latency_ms
- Ensures final rent does not exceed tenant max and cap rate satisfies landlord min.
- Average millisecond inference latency per agent response turn. • target: 300 • range: 0-1500
- Benchmark: CRENegotiationBench
- Latency Ms Per Turn target: 300
- 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
- Vijil · Optional
- Cerebras Inference · Optional
Evaluation contract
- boundary_constraint_check · Weight 1
- latency_ms_per_turn · Weight 1
Metrics
- latency_ms_per_turn · Target 300
Benchmarks
- CRENegotiationBench
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing property asset value, minimum landlord cap rate, maximum tenant rent
Output: JSON with agreed_cap_rate, initial_rent, escalation_rate, total_latency_ms
Execution harness: Python
Evaluation tools
- AutoGen · Agent Framework
- Cerebras Inference · Inference Infrastructure
Recipe freeze marker
No recipe freeze time is recorded.