FurnitureBench
Long-horizon furniture assembly tasks for evaluating real-world robot manipulation.
About
FurnitureBench is a public research dataset from Stanford Vision and Learning Lab. Long-horizon furniture assembly tasks for evaluating real-world robot manipulation.
Start with a representative slice and inspect its RGB video, Depth, Robot state, Actions signals together. This makes schema, timing, calibration, and annotation issues visible before the dataset becomes part of a larger training or evaluation pipeline.
Suggested agent workflows
Copy a prompt into your agent of choice. Each workflow starts small, respects the source terms, and makes the result inspectable in Rerun.
Audit a training slice
Fetch FurnitureBench from its official source at https://clvrai.github.io/furniture-bench/. Read and preserve its source terms and citation requirements. Build a reproducible manifest for a representative slice, then load it into Rerun. Inspect the RGB video, Depth, Robot state, Actions streams for timestamp gaps, missing samples, calibration problems, and annotation inconsistencies. Summarize the findings and save both the validation script and a Rerun recording that makes every issue easy to review.
Build a baseline pipeline
Use FurnitureBench to scaffold a baseline pipeline for Long-horizon manipulation, Assembly. Start with a small deterministic cohort and document the selection criteria. Create the dataset adapter, preprocessing steps, train and validation splits, and one lightweight baseline. Log inputs, labels, predictions, losses, and representative failures to Rerun so I can inspect the full experiment timeline. Include exact setup and reproduction commands.
Find useful failure cohorts
Analyze FurnitureBench and propose a practical cohort taxonomy for Long-horizon manipulation, Assembly. Use available metadata and the RGB video, Depth, Robot state, Actions signals to identify hard cases, rare conditions, coverage gaps, and possible leakage between splits. Materialize the cohorts as queryable metadata, log examples from each cohort to Rerun, and produce a short report explaining which cohorts should become regression tests. Do not infer labels that are not present in the source data.