Reproduce “Closed-Loop Physics-Based Manipulation Evaluation at a Commodity Price: What 4.4 L4 GPU hours at $3.50 Buys You” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Rohun Kshirsagar, Diana Rypkema
The authors build a serverless harness for physics-based robot manipulation evaluations and use it to test action chunking on a pick-and-place task. Eight shards complete a 200-rollout replication for $3.49 in 37.7 minutes. After excluding five initial-condition successes, single-step behavior-cloning checkpoints achieved 2.6–11.8% success, while chunked checkpoints scored zero; this result is limited to the tested task and data scale.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.