Strong, reproducible multimodal safety result — but paper not ready: submit or not?
mariyaivasileva · x · 2026-11-16
Mariya Ivasileva poses a real research dilemma: a result that reproduces across several domains and frontier models, targeting a gap in multimodal safety evaluation — but the authors aren't happy with paper readiness, want more human QA, and wish they had fixed obvious data-quality issues in existing benchmarks. Submit or not? The question highlights the tension between filling an evaluation gap and rigor/data quality in safety research.
More from Safety
- Fortune fires back at 'Bring Back Jensen' chorus over Nvidia governance — GaryMarcus · 2026-11-16
- Open-source advocates: models are a right, not a permission — don't accept the 'allowed' framing — nptacek · 2026-11-16
- SemiAnalysis: renting US compute to China yields Nvidia ~$7M profit per megawatt — xiaosun86 · 2026-11-16
- Reuters: Nvidia chips pile up in Hong Kong as China withholds import approval — _weiping · 2026-11-16
- AI exam cheating surges as spy camera and smart glasses rentals sell out days before national exam — ffff111222333 · 2026-11-16
- CANTraj: CAN Signals Recover 88% of Driving Routes Even Without GPS Data — MOTIF-Lab · 2026-11-16