Unverified50% confidenceFactExact time
Domain fine-tuning improves performance in the target domain while weakening the model's performance on general tasks.
1
Sources
50%
Confidence
Long-term
Relevance
11/2/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedVarying the stakes of a scenario causes severe but localized failures in models, causing them to change their responses even when the underlying professional obligation remains unchanged.79% similarVerifiedThe limitation of specialized small models is that their generalization ability rapidly degrades when task boundaries are ambiguous, cross-domain reasoning is required, or unseen edge cases need to be handled.79% similarUnverifiedThe positive correlation between base task performance and generalization ability weakens under significant distribution shift.79% similarVerifiedIrrelevant information occupying context space (context pollution) can significantly degrade the model's performance on core tasks.76% similarUnverified相关研究指出基础模型通常能从约束中受益，但指令微调过的模型在开放式生成任务上质量会明显下降76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1125127API
curl https://kongchang.com/api/v1/knowledge/claims/1125127MCP
get_claim(id=1125127)