Unverified60% confidenceFactExact time
Current mainstream large language models enter the deployment phase after pre-training is complete, with weights essentially frozen and unable to continue learning through interaction.
2
Sources
60%
Confidence
Long-term
Relevance
7/14/2026
First Seen
Sources
持续学习：通往AGI路上被忽视的核心瓶颈
Related Claims
UnverifiedAfter large language models exceed a critical scale, new capabilities that were not explicitly taught during the training phase emerge, such as multi-step reasoning, cross-lingual analogy, and tool-use planning.78% similarVerifiedLarge language models have two inherent weaknesses: their knowledge is cut off at training time and they cannot obtain the latest information, and they can only generate text and cannot directly perform actions.77% similarUnverifiedDiffusion language models are still at an early stage and continue to lag behind mature autoregressive models in generation quality, long-text coherence, and training stability.76% similarUnverifiedA language model in the pre-training stage only learns to continue text and does not understand the concept of answering questions or following instructions74% similarVerifiedModern multilingual large models establish language-agnostic semantic representations during the pre-training phase, building cross-lingual semantic alignment in high-dimensional vector space through joint training on massive multilingual corpora.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/510076API
curl https://kongchang.com/api/v1/knowledge/claims/510076MCP
get_claim(id=510076)