Unverified50% confidenceFactExact time
The dedicated player model (behaviorally cloning a model of "me") has a cross-entropy loss of 2.44 vs 4.20 nats/decision and movement-key accuracy of 90.6% vs 80.6%.
1
Sources
50%
Confidence
Long-term
Relevance
11/6/2026
First Seen
Sources
Related Entities
Related Claims
Unverified当不动动作占据80%以上帧时,模型通过最小化交叉熵损失可在不学习有意义特征的情况下达到80%准确率,这称为多数类捷径73% similarVerifiedThe cross-entropy loss in classification tasks is essentially the negative of the log-likelihood function, and minimizing cross-entropy is equivalent to maximizing the log-likelihood of the model on the training data.69% similarUnverified任务的输出熵越低(答案越可预测),推测性解码的加速效益越显著;GSM8K接受率从位置0的0.92衰减至0.53,MTBench从0.78急跌至0.1366% similarUnverified当所有商家评分集中在4.5-4.9的狭窄区间时,评分系统的信息熵急剧下降,用户从评分中获得的有效决策信息趋近于零65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1144443API
curl https://kongchang.com/api/v1/knowledge/claims/1144443MCP
get_claim(id=1144443)