Skip to content

V2 P0: CortexBench baseline — evals scaffold + signature BLOCK scenario #72

Description

@cursor

Background

V2 Definition of Done requires measurable reliability, not retrieval-only demos.

Problem Statement

No eval harness, scenarios, or baselines comparing Cortex to keyword/vector/graph retrieval.

Proposed Solution

  • Scaffold evals/ (datasets, scenarios, baselines, metrics, reports)
  • First scenarios: temporal change, superseded decision, contradiction, poisoned memory, missing knowledge, irreversible risk
  • Baselines A–C: keyword, vector-only, graph+vector
  • docs/EVALS.md methodology stub + runnable smoke
  • Signature scenario: payments restart → BLOCK

Acceptance Criteria

  • evals/ runnable in CI or documented make target
  • Signature BLOCK scenario passes against gate
  • EVALS.md lists methodology + known limitations

Depends on

Reliability Gate (for full signature scenario); scaffold can start earlier

Plan ref

todo cortexbench-baseline — V2 Enhancement 8

Activity

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

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions