Unverified50% confidenceFactExact time
Injecting "ties" (i.e., pairs of responses that are equally preferred) into preference data can weaken spurious correlations, because ties provide no discriminative signal and force the model to ignore irrelevant features.
1
Sources
50%
Confidence
Long-term
Relevance
10/27/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedEvaluations lacking a neutral control will misidentify spatial index bias as model susceptibility to persuasion, while matched controls can capture such confounding factors.79% similarUnverified验证集和测试集不应做任何数据增强，一旦施加变换会引入系统性偏差，破坏对模型泛化能力的无偏估计77% similarUnverifiedUnder domain mismatch conditions, relying solely on confidence-based filtering is insufficient to guarantee pseudo-label quality, because models may also exhibit high confidence on incorrect samples.74% similarUnverified端到端模型在代表性充足与代表性不足口音之间的WER差距并未按比例缩小，属于数据分布问题而非架构问题73% similarUnverifiedTwo widely cited direct-probing benchmarks exhibited saturation, as frontier models generally gave neutral responses to straightforward questions, causing the benchmarks to detect almost no bias.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1102723API
curl https://kongchang.com/api/v1/knowledge/claims/1102723MCP
get_claim(id=1102723)