Expired50% confidenceEventExact time
That developer fine-tuned the Gemma series of models (both base and instruction-tuned versions), and neither was able to surpass the performance of directly prompting the base model on the self-built evaluation set.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
9/12/2026
First Seen
Valid until: 9/26/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedFine-tuning can deeply alter a model's behavioral patterns, whereas prompt engineering can only operate within the boundaries of the model's existing capabilities72% similarUnverifiedThe prompts submitted to the models did not include question-type hints, requiring the models to determine the question type on their own.72% similarUnverified提示注入利用的是模型无法可靠区分「数据」与「指令」的天然缺陷，这一缺陷根植于LLM的训练机制本身70% similarUnverifiedPrompt技巧无法与特定模型版本解耦，需要从业者持续跟进模型迭代69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/910244API
curl https://kongchang.com/api/v1/knowledge/claims/910244MCP
get_claim(id=910244)