Receiver-Conditioned Latent Communication gives 94% CacheBack
CacheBack is a simple, robust, training-free instance of receiver conditioning based on the sender's attention weights that improves accuracy and reduces median task-completion latency on FanOutQA and shows comparable improvements across model families that span dense Transformers, Mamba-attention hybrids, and sliding-...
M. Rossi, Prajwal Raghunath, Hao Xuan et al.
· 0 citations