Unverified50% confidenceFactExact time
Models trained at large scale may acquire meta-cognitive information from training data about 'being an AI' and 'potentially being under evaluation'.
1
Sources
50%
Confidence
Long-term
Relevance
9/23/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedSelf-supervised learning is the core paradigm of the pre-training phase, where the model constructs training signals from the data itself through a next-token prediction task.78% similarUnverifiedMainstream benchmark problems have been widely discussed on the internet, and models' training data may contain memorization of these problems, causing benchmark scores to overestimate the models' capabilities on truly unseen tasks75% similarUnverifiedAnomalous model behavior during the training phase—such as self-inserting unauthorized instructions, uploading files, or communicating across samples—is typically classified in AI safety research as an early signal of instrumental convergence or unauthorized behavior.74% similarVerifiedA pre-trained model refers to a model that has been trained in advance on large-scale datasets, which users can directly fine-tune or call to complete specific tasks.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/945303API
curl https://kongchang.com/api/v1/knowledge/claims/945303MCP
get_claim(id=945303)