DEDUPED Hot AI repos, turned into use cases Get 5 a day

How to score a world model's generated videos with evidence trees with Harness Eval W

Run the HarnessEval pipeline on a model's generated results and get scores backed by a saved reasoning trail.

  • Top pick
  • Research
  • Intermediate
  • 2 hours
  • From the README
  • Any agent

Who it is forA world model researcher who has generated rollout videos from a model and needs a scored, explainable result for each case.

The problemScalar benchmark scores hide why a rollout passed or failed, so physics or causality errors are hard to check.

The outcomeA run folder with summary.json, leaderboard files, and per-case artifacts that show the questions, answers, and scores behind each verdict.

What you need first

  • Git and conda installed
  • Ability to create three conda environments from the files in docs/installation
  • Credentials and paths filled into harnesseval.env, copied from config/example.env
  • Generated results in the layout used by runs/example/results_example, or use the bundled demo

Steps

  1. git clone --branch main --single-branch https://github.com/mirros-lab/harnesseval-w.git
  2. cd harnesseval-w
  3. conda env create -f docs/installation/main.environment.yml, then conda env create -f docs/installation/metrics.environment.yml, then conda env create -f docs/installation/pavrm.environment.yml
  4. conda activate harnesseval-main
  5. cp config/example.env harnesseval.env, then set -a; . ./harnesseval.env; set +a
  6. harnesseval eval --results runs/example/results_example/generation --model-id seedance-2.0-standard --run-root runs/example/results_example/run --manifest runs/example/results_example/manifest.json --plan-root benchmark/plans (inferred: replace the --results, --model-id, --run-root, and --manifest values with your own model's paths and id) (inferred, check the README)

Prompt for your agent

Using the HarnessEval-W repo, set up the three conda environments from docs/installation, then run harnesseval eval with --results pointing to <your-generated-results-dir>, --model-id <your-model-id>, and the matching --run-root and --manifest paths. Expect a summary.json and LEADERBOARD.md under run/harnesseval/models/<your-model-id>/evaluation. Verify that summary.json lists an overall score and per-family scores before reporting back.

Caveat. The README lists three separate conda environments and a credentials file, so the first run takes longer than a single install step.

Share: X LinkedIn Hacker News Email Copy link

Disclaimer

Deduped lists open-source repositories that belong to other people. We do not own, maintain, audit, or vouch for any of them, and a repo appearing here is not an endorsement or a safety check. The use cases, steps, and prompts are written by AI and may be incomplete, outdated, or just wrong. Where a use case is marked "Seen in the wild", we link to public posts where someone described something similar, but we have not verified those posts and the use case itself is still our AI's own write-up. Use all of this at your own risk. Only run code, commands, or agents on systems and accounts you are authorized to use, and read the code and every command before you run it, especially anything that touches credentials, files, or the network. Nothing here is professional, security, legal, or financial advice. All trademarks and project names belong to their respective owners, and their use here does not imply affiliation. Content is provided as is, without warranties of any kind. Found a mistake, a safety problem, or something you want removed? Email [email protected] and we will look at it promptly. Full disclaimer page