Skip to content

Add claims domain and README for TauBench#135

Open
Risper8 wants to merge 2 commits intosierra-research:mainfrom
Risper8:add-claims-domain
Open

Add claims domain and README for TauBench#135
Risper8 wants to merge 2 commits intosierra-research:mainfrom
Risper8:add-claims-domain

Conversation

@Risper8
Copy link

@Risper8 Risper8 commented Jan 11, 2026

Added insurance claims domain and domain README for TauBench.

@Risper8 Risper8 closed this Jan 11, 2026
@Risper8 Risper8 reopened this Jan 11, 2026
@Risper8
Copy link
Author

Risper8 commented Jan 11, 2026

This PR adds a new Insurance Claims domain to TauBench.

Key contributions

  • Introduces a structured claims environment with policies, claims, users, and fraud handling
  • Defines a strict tool-usage and confirmation protocol aligned with MT-1 evaluation requirements
  • Adds a domain-specific policy (policy.md) and environment setup
  • Supports claim reporting, document verification, approvals/rejections, payments, and human escalation
  • Fully compatible with TauBench’s infrastructure

Motivation

This domain is designed to benchmark AI agent reliability in regulated, high-stakes workflows where:

  • Explicit user confirmation is required before state-changing actions
  • Tool sequencing and compliance are critical
  • Errors have real-world financial and legal implications

Testing

  • Executed simulations to validate environment setup and agent-tool interaction.

Feedback and suggestions are welcome.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant