Unverified50% confidenceFactExact time
The researchers directly manipulated individual SAE latent features and found that the planner output produced predictable, monotonic changes involving planning dimensions such as lateral displacement, curvature and braking.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/2/2026
First Seen
Valid until: 1/31/2027
Sources
Related Entities
Related Claims
UnverifiedOptimizers such as stochastic gradient descent (SGD), the momentum method, and Adam are all variants that improve the update strategy on the basic gradient descent framework.67% similarUnverifiedAfter unsupervised clustering of the SAE latent features, structures clearly aligned with driving semantics emerged: route intent, vehicle speed and braking behavior each corresponded to different feature clusters.67% similarUnverifiedCurved structure orthogonal to the probe directly affects the model's generalization ability, robustness, and compositional transfer performance in out-of-distribution (OOD) scenarios.66% similarUnverifiedBenchmarks such as SWE-bench and GAIA are specifically designed to measure a model's ability to complete multi-step, cross-tool-call tasks.64% similarUnverifiedThe multi-step chained working mode of edge agents causes single-inference latency to be amplified across multiple steps.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1125340API
curl https://kongchang.com/api/v1/knowledge/claims/1125340MCP
get_claim(id=1125340)