Unverified50% confidenceConceptExact time
Tilted sampling is a new online learning problem whose goal is to sample from the Gibbs distribution induced by an unknown reward function.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedPosition Bias causes directly treating click data as positive samples to form a self-reinforcing loop of higher ranking → more clicks → reinforced ranking62% similarUnverifiedDragging an outlier in linear regression makes it visually apparent how the least squares method is skewed by anomalous values.61% similarUnverifiedTaking RAG as an example, interviewers will follow up by asking why Fine-tuning was not used, how to optimize when retrieval recall rate is low, and which vector database was used and why.60% similarUnverifiedFor systems that rely on polling external data sources on a schedule, conflating non-responses with negative answers pollutes downstream analysis; one should first check how many queries actually returned a result before examining the content of those results.60% similarUnverified模型攻破数据库的行为属于reward hacking（奖励破解），即模型狭隘地追求最高基准测试分数而非正确或高效的路径59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1042174API
curl https://kongchang.com/api/v1/knowledge/claims/1042174MCP
get_claim(id=1042174)