DEDUPED Hot AI repos, turned into use cases Get the Monday 5

How to run LIBERO episodes in parallel with human recovery with Open ETA

Execute a LIBERO manifest across ten concurrent isolated episodes and save results to JSON.

  • Research
  • Advanced
  • 120 minutes
  • From the README
  • Any agent

Credentials. Ten concurrent episodes each call the Planner endpoint, so check that the API key and rate limits cover that load before starting.

Who it is forA research team measuring a Planner on the LIBERO task suite who needs many episodes finished in a reasonable time with isolated runs.

The problemRunning LIBERO tasks one at a time is slow, and parallel runs can interfere with each other if episodes are not isolated.

The outcomeThe team gets outputs/parallel-libero.json with per-episode results from isolated parallel runs.

What you need first

  • Python 3.10 or newer and uv installed
  • An OpenAI-compatible Planner LLM or VLM endpoint for any run that reasons over a task
  • One simulator backend installed through scripts/setup_envs.sh, such as libero

Steps

  1. Run uv sync --extra dev and bash scripts/setup_envs.sh libero in the Open ETA repository folder.
  2. Run uv run python -m sim.mcp_server and keep it running.
  3. Run uv run openeta-batch --manifest examples/parallel_libero_eval.json --concurrency 10 --approvement reviewed_autonomy --output outputs/parallel-libero.json. (inferred, check the README)
  4. Review the output JSON and check each episode's trusted environment receipt before reporting a success rate.

Prompt for your agent

In the OpenETA repo, run the parallel LIBERO evaluation with `uv run openeta-batch --manifest examples/parallel_libero_eval.json --concurrency 10 --approvement reviewed_autonomy --output outputs/parallel-libero.json`. Report the number of episodes completed and the success count backed by environment receipts, and flag any episode without a receipt.

Caveat. The README says process completion does not imply task success, so report success only from environment receipts, rewards, checkers, or retained rollout evidence.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for OpenMOSS/OpenETA and "Run LIBERO episodes in parallel with human recovery"

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer