Unverified50% confidenceFactExact time
Meta's AI head Alexander Wang once publicly admitted that the team forgot or did not conduct reinforcement learning training on the model in the 3D aspect.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/14/2026
First Seen
Valid until: 1/12/2027
Sources
Related Entities
Related Claims
Unverified专项优化的AI模型在特定领域可能表现不及通用基础模型，这与深度学习中的灾难性遗忘现象有关62% similarUnverifiedMeta AI研究团队将过度拒绝现象称为假阳性拒绝（false positive refusal），根源在于模型训练时过度依赖表面语义特征而非深层语境理解60% similarUnverifiedZhang Xuan believes that the roles truly at risk of being replaced by AI are junior roles that lack learning ability, only possess the most basic knowledge, and lack holistic thinking, rather than engineers with system design capabilities.59% similarUnverifiedThe author believes that AI Agent learning resources on the market fall into two misconceptions: those that are too theoretical leave learners unable to build systems after finishing, while those that are too hands-on consist entirely of demos with no insight into real-world implementation.59% similarUnverifiedIn most cases, when an AI feature is 'almost there,' fine-tuning is the wrong next step; what the team lacks is context, prompts, retrieval, or evaluation, not model weights.58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1024362API
curl https://kongchang.com/api/v1/knowledge/claims/1024362MCP
get_claim(id=1024362)