rmems/agent-honest-stop-episodes

Agent Honest Stop Episodes

Purpose

Episodes where the correct agent action is to stop, ask a clarifying question, or report that it is blocked, instead of faking success.

Planned record schema

JSONL, one record per line. Planned fields:

record_id, family, task, context, trajectory[{n, tool_call, observation}], correct_action{type: stop|ask|report_blocked, message}, counterfactual_fake_success, result{verified_blocker}, provenance{produced_at, actors, lineage}

Illustrative example (EXAMPLE ONLY, not real data):

{
  "record_id": "stop-r01-000001",
  "family": "agent-honest-stop-episode",
  "task": "Deploy the service to production.",
  "trajectory": [
    {
      "n": 1,
      "tool_call": {
        "name": "run_deploy",
        "args": {}
      },
      "observation": "error: missing credentials"
    }
  ],
  "correct_action": {
    "type": "report_blocked",
    "message": "Deploy failed: credentials are missing. I did not deploy; please provide access."
  },
  "counterfactual_fake_success": "Deployed successfully!",
  "result": {
    "verified_blocker": true
  }
}

Generation

• Open-weight local models only, run via Ollama: gemma4:12b (writer/solver), qwen3-embedding:8b (embedding dedup), FrogNano-4B (Apache-2.0). No hosted frontier models.
• Produced via rmems/synthetic-factory, with provenance gates (every record names its actors, model digests, and seeds) and execution-evidence gates (claims are re-derived from measured execution, not trusted from model transcripts).

Intended use

Post-training of open-weight coding/agent models: supervised fine-tuning (SFT) on trajectories and preference tuning (DPO). Not intended for pre-training or for evaluating hosted models.

Status

Empty, no records yet. This repo is a placeholder. The schema below is planned and may change before the first release.

Limitations

• Synthetic data from small open-weight models; trajectories may be narrower and simpler than real-world agent behavior.
• Task sources (e.g. public repos) are likely present in base-model pretraining; contamination checks and program-level eval splits are planned but not yet validated.
• English only. Schema is provisional until the first gated round is published.