Unverified50% confidenceOpinionExact time
TJ stated that he had barely given any additional prompts, yet the AI automatically referenced specific content they had discussed in their previous podcast episodes, leading him to speculate that his podcast content may have already entered the model's training data.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/29/2026
First Seen
Valid until: 12/28/2026
Sources
Related Entities
Related Claims
UnverifiedThe experiment comes from a single social media source and lacks complete reproducible test details, making it impossible to confirm whether the parameter and task settings for Laya and Jev were fair.62% similarUnverifiedDirectly importing raw chat logs into a fine-tuning dataset yields poor training results, because raw chat logs lack emotional and contextual annotations, leaving the model only able to learn surface character patterns rather than the decision logic of when to use a certain style.62% similarUnverified此前的AI模型在处理医患对话中'先浅后深'的对话模式时经常出错，要么将两次讨论视为不同话题，要么无法正确关联前后信息62% similarUnverifiedThe performance of evaluation methods validated only on synthetic data is questionable when applied to real conversations.61% similarUnverifiedThe United States handles AI training data under the 'fair use' doctrine, in a relatively ambiguous manner.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/958301API
curl https://kongchang.com/api/v1/knowledge/claims/958301MCP
get_claim(id=958301)