LeCun 的自回归批评被"输出错误 token"训练回怼了吗？
有人拿"模型会喊 oops 然后改错"这个训练行为，去质疑 LeCun 那套自回归不能纠错的老论点，争议点很具体。
一条推文质疑 Yann LeCun 关于自回归模型无法纠错的核心论点：现代 LLM 在训练中会学到"输出错误 token 后承认并修正"的行为。
这种 oops 修正模式在 RLHF 与推理链训练中被显式强化。质疑者要求 LeCun 回应其论证是否已被此类训练数据削弱。
Did Yann ever answer why his argument isn't defeated by training the "oops, that was the wrong token!" behavior? https://t.co/pCLtdLqJOc
This report is categorized under AITOP's "Papers" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include Yann LeCun, 自回归, 推理模型, LLM.
Source attribution: curated from Teortaxes (@teortaxesTex).
