Unverified50% confidenceFactExact time
Representation collapse is the most notorious failure mode of JEPA architectures, where the encoder learns to output constant embeddings independent of the input, reducing prediction loss while making the representations entirely uninformative.
1
Sources
50%
Confidence
Long-term
Relevance
8/16/2026
First Seen
Sources
复现TS-JEPA踩坑：从0%到74%控制精度的差距在哪
Related Claims
UnverifiedA reasonable prediction loss combined with 0% downstream control accuracy is highly consistent with representation collapse, meaning the encoder has fooled the self-supervised objective without learning any useful dynamic information.71% similarUnverifiedJEPA 若只在抽象空间中预测却无法解码回可操作的具体状态或观测，其在规划和控制中的价值会大打折扣69% similarUnverified级联失败是指单次工具调用错误会污染后续上下文，导致模型在错误的中间状态上继续推理，最终使整个任务偏轨68% similarUnverifiedThe core of the Itanium architecture's success or failure lay in betting on compiler technology, which never reached the desired level, especially in dealing with dynamic runtime issues such as branch prediction and memory access latency.68% similarUnverified表示坍缩指模型将所有输入映射到同一点或低维流形，使损失趋近于零但毫无意义，是自监督学习的核心挑战67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/756920API
curl https://kongchang.com/api/v1/knowledge/claims/756920MCP
get_claim(id=756920)