Share the Symbol, Not the Vector: Training-Free Fact Communication Across Language Models

Hugging Face Daily Papers Papers
language-models multi-agent-systems hidden-states training-free knowledge-transfer cross-model-communication

Summary

The paper introduces SYNC (Symbol-Indexed Neural Communication), a training-free, word-indexed interface that lets heterogeneous language models share selective object facts via word indices and receiver-local hidden-state directions, achieving strong four-answer success across nine receivers from five model families.

Multi-agent systems allow language models to solve tasks collaboratively by sharing complementary information. These systems typically communicate through text, while recent approaches also explore exchanging internal model states. However, sharing these states across architectures often requires learned translators or assumptions about model compatibility. We introduce SYNC (Symbol-Indexed Neural Communication), a word-indexed interface that connects heterogeneous models without fine-tuning or a learned translator. From descriptions only it can see, the sender selects words from a shared list and sends their indices for each object, without seeing the questions. The receiver writes the described facts into its hidden states using its own word directions, vectors fitted once from its activations. We test selective use by swapping one attribute between two objects and requiring correct answers about both changed and unchanged attributes before and after the swap, four answers in total. We choose an early write layer on four development receivers and then fix it. With three different senders and nine newer receivers from five families, four-answer success on new scenes reaches 72.0-98.3%, up to 96 percentage points above middle-layer writing, and the two largest receivers match plain text. A task-free probe predicted in advance which depths each new receiver can use (45 of 45 cases). Without retuning, messages listing every fact keep 89.6-100% success under new layouts, and sender messages reach 23.9-64.0% on a landmark task with multi-token symbols, below text but from 0% without a message. These results show that shared word indices and receiver-local directions can support selective fact use across model families without additional training.
Original Article
View Cached Full Text

Cached at: 10/15/26, 10:59 AM

Paper page - Share the Symbol, Not the Vector: Training-Free Fact Communication Across Language Models

Source: https://huggingface.co/papers/2610.19696 Published on Oct 14

·

Submitted byhttps://huggingface.co/XINLI1997

XinLion Oct 15

Abstract

SYNC enables heterogeneous language models to share selective object facts via word indices and receiver-local hidden-state directions without fine-tuning or translators.

Multi-agent systemsallow language models to solve tasks collaboratively by sharing complementary information. These systems typically communicate through text, while recent approaches also explore exchanginginternal model states. However, sharing these states across architectures often requires learned translators or assumptions about model compatibility. We introduceSYNC(Symbol-Indexed Neural Communication), aword-indexed interfacethat connects heterogeneous models without fine-tuning or a learned translator. From descriptions only it can see, the sender selects words from a shared list and sends their indices for each object, without seeing the questions. The receiver writes the described facts into itshidden statesusing its ownword directions, vectors fitted once from its activations. We testselective useby swapping one attribute between two objects and requiring correct answers about both changed and unchanged attributes before and after the swap, four answers in total. We choose an early write layer on four development receivers and then fix it. With three different senders and nine newer receivers from five families, four-answer success on new scenes reaches 72.0-98.3%, up to 96 percentage points above middle-layer writing, and the two largest receivers match plain text. Atask-free probepredicted in advance which depths each new receiver can use (45 of 45 cases). Without retuning, messages listing every fact keep 89.6-100% success under new layouts, and sender messages reach 23.9-64.0% on a landmark task withmulti-token symbols, below text but from 0% without a message. These results show that shared word indices and receiver-local directions can support selective fact use across model families without additional training.

View arXiv pageView PDFProject pageGitHub0Add to collection

Get this paper in your agent:

hf papers read 2610\.19696

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper1

#### XINLI1997/SYNC-directions Updatedabout 9 hours ago

Datasets citing this paper1

#### XINLI1997/SYNC-data Preview• Updatedabout 9 hours ago • 37

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2610.19696 in a Space README.md to link it from this page.

Collections including this paper1

Similar Articles

Supportive Token Revealing for Fast Diffusion Language Model Decoding

arXiv cs.CL

This paper proposes AXON, a training-free module that improves the quality-latency trade-off of discrete diffusion language model decoding by intelligently selecting 'anchor' tokens to reveal first, using attention, uncertainty, and confidence signals to support subsequent denoising steps. Experiments on reasoning and code-generation benchmarks show AXON reduces function evaluations while maintaining or improving accuracy.