Unverified50% confidenceTradeoffExact time
When 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.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedExcessive safety fine-tuning can cause models to over-refuse, undermining their practical utility.70% similarUnverifiedOver-refusal strategies reduce risk but harm model utility, especially in scenarios such as professional development and security research that require touching sensitive boundaries.66% similarUnverifiedIf a low-cost model has insufficient reasoning capability, it leads to more retry rounds, and the hidden costs of accumulated failed retries may exceed those of a high-cost, high-efficiency solution (the paradox of "cheap models actually being more expensive").65% similarVerifiedSome models carry the risk of over-optimization toward test sets, causing leaderboard scores to deviate from real-world performance (benchmark saturation).65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/894894API
curl https://kongchang.com/api/v1/knowledge/claims/894894MCP
get_claim(id=894894)