Search papers, labs, and topics across Lattice.
Imperial College London, University of Edinburgh, Nanyang Technological University, MBZUAI
5
0
9
Implicit self-description can achieve comparable or superior debiasing results to traditional methods that rely on sensitive attributes.
Pythagoras-Prover achieves state-of-the-art performance in formal proving with dramatically fewer parameters, challenging the notion that bigger models always yield better results.
LLMs' factual recall falters when fine-tuned on new information, and this can be traced to specific latent directions in the residual stream.
Steer LLM attention without the memory bottleneck: SEKA unlocks prompt highlighting and other control capabilities even with FlashAttention.
Diffusion language models get a nearly 20% boost on MBPP by strategically planning the order in which they fill in the blanks.
