Taurus harness cuts CT annotation costs by half with LLM review
Researchers at an unnamed lab tested whether an LLM could label CT scans more efficiently when guided by a structured harness (Taurus) versus free-form coding. They compared Taurus to a free-form agent using Claude Opus 5.5 across six public datasets (939 volumes total). The harness’s first version improved results on only two datasets by a negligible margin. However, refining the automated pass…
1 source primary source
Key points
- Taurus harness reduces CT annotation cost to **$1.22 per case**, half of Claude Opus 5.5’s **$2.71**
- Final version matches free-form agent accuracy on six public datasets (939 volumes total)
- Most efficiency gains stem from automation, not the LLM’s review step
Model page: Claude Opus 5.5 →
Read the original at arXiv cs.AI · by Bing Hu, Rishabh Sharma, Jatin Mehta primary sourceOpen source ↗
Topics · follow one to build your own front page
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Researchers introduce ResMem to turn retrieved knowledge into parametric memory · 1 src
- Circuit Oracle automates attribution graph analysis · 1 src
- arXiv study finds AI humanizers alter 35–65% of fact units in documents · 1 src
- Researchers propose gavagai to improve Vision-Language model alignment · 1 src
- Gemini 3.6 Flash shows language-dependent negotiation behavior in study · 1 src
Comments
via GitHub Discussions