研究者：期刊桌面拒稿率高，会议人人可获实审是重大优势
avt_im · x · 2026-10-17
In a debate on NeurIPS review quality, researcher avtim argues desk rejects are common at journals — papers of his desk-rejected by JRSS-B easily made it into JMLR. He concludes that having every paper receive an actual review is a major advantage of the conference system, pushing back on using LLMs to adjudicate review decisions.
「研究」频道最新
- Vals AI 科学指数：OpenAI 三款新模型分榜领跑，无一全科称王 — BenBlaiszik · 2026-10-17
- NeurIPS 2026 论文：把 AI 文本检测当作持续学习问题求解 — lucy3_li · 2026-10-17
- Regents Labs 周报:精选 17 篇 AI/agent 科学论文附 AI 精读 — seanwbren · 2026-10-17
- 开发者构建「前数字」AI 数学系统，用关系与张力训练图案 — debreuil · 2026-10-17
- 开源本地 PDF/电子书检索工具 research-rag:CPU 可跑、支持 MCP — kishkisan · 2026-10-17
- Epoch AI 千局杀戮尖塔实测：模型胜率超过人类作者但仍每幕犯错 — Jsevillamol · 2026-10-17