Follow-up: the shortest peer review in the AI-reviewer scandal was 43% AI-generated
nsaphra · x · 2026-11-06
nsaphra shares more detail on her paper rejected with apparently AI-written reviews: the shortest review tested at 43% AI-generated. She says that would be acceptable if any human-written reviews existed — but reviewers are giving up on reading amid a mountain of AI slop, and review quality has collapsed.
More from Research
- Hugging Face spent 800k GPU hours to publish a free Ultimate MoE Training Handbook — gordic_aleksa · 2026-11-07
- AI2 Releases ToyBlocks, an Open Simulated Tabletop for Robot Learning Experiments — allen_ai · 2026-11-07
- No LoRA learning rate can emulate full fine-tuning, finds agentic-experiments deep dive — thebasepoint · 2026-11-07
- Viral TDV paper under fire: only compared against nerfed iBOT and DINO, doesn't scale — gabriberton · 2026-11-07
- Why alpha shouldn't scale with rank: LoRA effective learning rate dynamics — thebasepoint · 2026-11-07
- LoRA ranks move in lockstep early in training; higher rank means higher effective LR — thebasepoint · 2026-11-07