arXiv:2606.03280v1 Announce Type: new Abstract: Recent work shows that language models can transmit behavioural traits through hidden signals in generated data during training. We ask whether a more direct and stricter channel is also viable: can one language model communicate useful intermediate reasoning state to another at inference time by translating and injecting hidden activations, rather than by passing natural-language text? We test this question in a controlled Pythia-160M to Pythia-410M multi-hop reasoning setting. A linear translation layer learns a strong normalized-space map betw
Source: arXiv cs.AI — read the full report at the original publisher.