Unverified50% confidenceFactExact time
Self-supervised forgetting refers to the gradual loss of general capabilities acquired during pretraining when a model adapts online using only self-generated labels.
1
Sources
50%
Confidence
Long-term
Relevance
11/9/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedPersona conditioning is widely used for user simulation, but its effectiveness depends on whether the model can truly "forget" or "isolate" its own pretrained knowledge; the knowledge leakage study shows that this isolation is incomplete.77% 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 tasks76% similarUnverifiedRecursive training makes extensive use of task completion rates and code execution results as reward signals, lacking explicit supervision for narrative consistency, which may lead to catastrophic forgetting of general language modeling capabilities, and this is more pronounced in models with small parameter counts.76% similarUnverifiedAlignment training typically includes large amounts of demonstration data such as "refusing harmful requests" and "stating one's own limitations," which makes models inclined to produce safety-compliant statements at the text level, but this text-level alignment does not automatically transfer to the action level.75% similarUnverifiedMethods such as freezing LLM backbone parameters during the visual alignment training phase, staged training, data ratio replay, and LoRA can mitigate catastrophic forgetting in multimodal training.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1150288API
curl https://kongchang.com/api/v1/knowledge/claims/1150288MCP
get_claim(id=1150288)