Unverified50% confidenceFactExact time
The alignment tax is not uniformly distributed but shows significant task selectivity: structured extraction is almost unaffected, JSON generation declines slightly, and the routing task degrades severely.
1
Sources
50%
Confidence
Long-term
Relevance
11/5/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedRouting benefits depend heavily on request composition: if the share of Agent intermediate steps is low and the share of reasoning steps is high, routing benefits shrink significantly.65% similarUnverifiedWhen Meta open-sourced the Llama series of models, it found that excessive safety alignment significantly reduces the model's usefulness on normal tasks, creating the alignment tax dilemma.63% similarUnverifiedThe alignment tax refers to the phenomenon where, when a model is overtrained to refuse harmful requests, its refusal boundary may generalize to normal harmless scenarios, resulting in excessive caution.63% similarUnverifiedThe positive correlation between base task performance and generalization ability weakens under significant distribution shift.62% similarUnverifiedUnder extremely resource-constrained scenarios, a greedy cost-minimization strategy concentrates requests onto the lowest-scored instance, actually performing worse than blind dispersion.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1138293API
curl https://kongchang.com/api/v1/knowledge/claims/1138293MCP
get_claim(id=1138293)