What 'hard prompts' look like in vision: rare concept combos like armadillos in berets
gowthami_s · x · 2026-10-28
The thread author offers vision equivalents of hard prompts: rare concept combos like "3 armadillos with one wearing a red beret and the other two wearing gold crowns sitting on ottomans," people in weird dance poses, or long-text posters where each line needs its own color and font—prompts where whole groups fail, exactly where extra compute should go.
Related event: RL group-sampling trick: small K plus persistence on hard prompts(3 posts)→
More from Research
- Eric Topol highlights immunome paper combining multi-assay data with AI to model immune aging — EricTopol · 2026-10-28
- Google Research's 'sleep framework' lets LLMs keep learning, retaining up to 10M tokens — behrouz_ali · 2026-10-28
- CATCH: ~200 examples of adversarial training make LLMs robustly honest across benchmarks — irinarish · 2026-10-28
- Andrew Critch's dubious credences: P≠NP at 80%, Riemann hypothesis at just 60% — AndrewCritchPhD · 2026-10-28
- AI-driven Indian biotech PopVax, backed by $20M from Vitalik's Balvi fund, enters human trials with fridge-stable broad-spectrum Covid vaccine — chaitjo · 2026-10-28
- Cognition launches FrontierCode 2.0, a coding benchmark testing judgment beyond the spec — DevinAI · 2026-10-28