Unverified50% confidenceEventExact time
In the feature elimination exercise, after removing number_of_courses_viewed accuracy dropped by only 0.002, whereas after removing interaction_count accuracy dropped by 0.044, a difference of more than twenty times.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAfter applying DA, Gemma-4-31B reduced attention tokens by 52.0%, with an accuracy drop of only 1.27 percentage points.68% similarUnverified研究表明在INT4/INT8精度下多数推理和生成任务的输出质量下降幅度在1-3%以内67% similarUnverified4-bit quantization compared to FP16 typically results in a 0.5–2% drop in benchmark accuracy.67% similarUnverifiedLID accuracy at switching points is typically 10–20 percentage points lower than LID accuracy at non-switching points.66% similarUnverifiedIn the final round of experiments, the composite evaluation score improved by 0.0903, and zero-score conversations decreased by 40%.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1011817API
curl https://kongchang.com/api/v1/knowledge/claims/1011817MCP
get_claim(id=1011817)