Blind test of 90 page pairs: named style references beat adjectives 68% across Opus, Sol, Kimi
maxsloef · x · 2026-09-27
maxsloef built a small style-elicitation eval: Opus 5.5, Sol 6, and Kimi K3 each generated two landing pages for a fake startup — one from an adjective set (dreamy, gentle, luminous), one from a named style reference (e.g., James Turrell) — and he blind-rated 90 pairs.
- Named references won 68% of the time: Sol 73%, Kimi 67%, Opus 63%.
- The effect holds even when the model picks the reference itself.
- He released the eval as Namedrop and hopes labs will hillclimb on it.
Related event: Named reference beats adjectives for style prompting in blind test(2 posts)→
More from Models
- Running DeepSeek 4.1 locally on an $800 rig with a used RTX 3090 at 2-3 tok/s — LankyGuitar6528 · 2026-11-11
- Google open-sources Harness², an agent that rewrites its own setup per task — Arindam_1729 · 2026-11-11
- Dev dumps Codex desktop app over quality issues, falls back to codex-cli — tekbog · 2026-11-11
- Why do old models get dumber before launches? Lab party, says joking X post — glcst · 2026-11-11
- Grok Bot now detects hidden prompt injections, warns users and refuses to follow them — XFreeze · 2026-11-11
- Shanghai Institute Open-Sources Nex-N2.6, a 321B Multimodal Agentic Model — victormustar · 2026-11-11