Search papers, labs, and topics across Lattice.
Heriot-Watt University
2
0
4
LLMs' factual recall falters when fine-tuned on new information, and this can be traced to specific latent directions in the residual stream.
Hybrid Transformer-SSM architectures can match or exceed pure Transformers in data-efficient in-context retrieval, but only if the task doesn't require precise positional reasoning.
