Direct Semantic Communication Between Large Language Models via Vector Translation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Yang, Fu-Chun, Eshraghian, Jason
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909889569226752
author Yang, Fu-Chun
Eshraghian, Jason
author_facet Yang, Fu-Chun
Eshraghian, Jason
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2511_03945
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Direct Semantic Communication Between Large Language Models via Vector Translation
Yang, Fu-Chun
Eshraghian, Jason
Computation and Language
Artificial Intelligence
I.2.7
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.
title Direct Semantic Communication Between Large Language Models via Vector Translation
topic Computation and Language
Artificial Intelligence
I.2.7
url https://arxiv.org/abs/2511.03945