Search papers, labs, and topics across Lattice.
3
0
5
2
This work finds that simple difference of means vectors coherently represent reward hacking in Kimi K3, GLM 5.2, and Qwen 3.8 Max, and provides evidence that simple, white-box methods can be used to scalably study and monitor reward hacking behaviors in frontier open source models.
Internal representations can serve as powerful lie detectors, revealing discrepancies in LLM forecasts that CoT reasoning may obscure.
Sparse autoencoders, despite their popularity for extracting interpretable features, often fail to capture the underlying manifold structure of concepts, instead fragmenting them across multiple, diluted features.
