Paper data

Knowledge Plane: university firewall

§3.2 Fault Diagnosis, cause #1 (Figure 1)

Knowledge Plane: stale DNS

§3.2 Fault Diagnosis, cause #2 (Figure 1)

Knowledge Plane: overloaded server

§3.2 Fault Diagnosis, cause #3 (Figure 1)

Celer Bridge BGP hijack

§3.3 Reasoning amidst conflicting information (Figure 2)

Billing-aware load balancing

§3.4 Agentic policy-based routing (Figure 3)

Inputs

What the agents were given: prompts, topologies, the experiment config, and the code defining the shared system prompt, tools, and each scenario.

• code/billing_policy.py
• code/kp_why_fix.py
• code/network_agent.py
• code/routing_policies.py
• code/routing_policies_converged.py
• paper_experiments.toml
• prompts/celer_bridge/AWS.txt
• prompts/celer_bridge/ISP.txt
• prompts/celer_bridge/LegitAS.txt
• prompts/celer_bridge/Tier1.txt
• prompts/celer_bridge/TinyInc.txt
• prompts/celer_bridge/TransitAS.txt
• prompts/celer_bridge/final-report.txt
• prompts/knowledge_plane/ACM.txt
• prompts/knowledge_plane/AS1.txt
• prompts/knowledge_plane/AS2.txt
• prompts/knowledge_plane/EveLink.txt
• prompts/knowledge_plane/Uni.txt
• prompts/knowledge_plane/User.txt
• prompts/knowledge_plane/Web.txt
• prompts/knowledge_plane/final-report.txt
• prompts/optimizing_billing_policy/Cheap.txt
• prompts/optimizing_billing_policy/Expensive.txt
• prompts/optimizing_billing_policy/ISP.txt
• prompts/optimizing_billing_policy/Remote.txt
• prompts/optimizing_billing_policy/TinyInc.txt
• prompts/optimizing_billing_policy/final-report.txt
• topologies/celer_bridge.csv
• topologies/knowledge_plane.csv
• topologies/optimizing_billing_policy.csv