20+ ICLR 2027 Submissions Found With Fabricated Results, Sparking Reproducibility Critique
srchvrs · x · 2026-10-23
Gautam Kamath highlights Nihar B. Shah's blog post documenting 20+ ICLR 2027 submissions with clear signs of fabricated results — some effectively exposed by their own figure captions.
In the replies, srchvrs adds a related pathology: worse than releasing no code is releasing "code" that doesn't work, doesn't match the paper, and can't be fixed even by two experienced ML engineers — with authors refusing to provide fixes or clarifications.
Related event: Scholar Finds 20+ ICLR Submissions Exposing Fake AI-Generated Results(3 posts)→
More from Research
- Tohoku team launches Digital Hydrogen Platform with 30,000+ entries from 4,000 papers — BenBlaiszik · 2026-10-23
- Ropedia releases DONIC: one policy controls two dexterous robot hands, 86.79% success on unseen trajectories — liuziwei7 · 2026-10-23
- Tencent releases WeVisOne-VL: one VLM for detection, segmentation, OCR and matting — _akhaliq · 2026-10-23
- Tencent releases Hy-Nomas-Bench: 280 cases testing video world model controllability — _akhaliq · 2026-10-23
- Datology's Zephon data curation tool ships v0.2 two weeks after launch — josh_wills · 2026-10-23
- Pleias releases TD Bench, a controlled multilingual benchmark of fluency judge models across 25 languages — antoine_chaffin · 2026-10-23