Direct Semantic Communication Between Large Language Models via Vector Translation
This work addresses the challenge of constrained information transfer and computational overhead in collaborative AI systems, though it appears incremental as it builds on existing representation learning techniques.
The paper tackles the problem of inefficient token-based communication between large language models in multi-agent settings by introducing a method for direct semantic exchange via vector translations, achieving an average cosine alignment of 0.538 and enabling stable generation with a 2.01:1 transfer asymmetry.
In multi-agent settings, such as debate, reflection, or tool-calling, large language models (LLMs) pass messages as plain tokens, discarding most latent semantics. This constrains information transfer and adds unnecessary computational overhead. We form a latent bridge via vector translations, which use learned mappings that enable direct semantic exchange between representation spaces. A dual-encoder translator trained between Llama-2-7B and Mistral-7B-Instruct attains an average cosine alignment of 0.538. Injecting the translated vectors at 30 percent blending strength steers the target model's generation without destabilizing logits. Bidirectional evaluation shows a 2.01:1 transfer asymmetry, indicating that general-purpose models yield more transferable representations than instruction-tuned variants. This conservative injection preserves computational stability while demonstrating that cross-model latent communication is feasible, enabling collaborative AI systems that share meaning rather than tokens.