The paper investigates how multimodal language models can generate perceptual representations without dedicated encoders. It shows that the shared transformer can internally create these representations in its early-to-middle layers, a structure termed a Virtual Encoder. Experiments with linear probing, similarity metrics, and causal analysis reveal that this encoder-like computation emerges even when models receive only perceptual tokens, indicating that perception and language processing can be decoupled within a single architecture.
arXiv:2607. 26596v1 Announce Type: cross Abstract: Multimodal large language models (MLLMs) have demonstrated remarkable capabilities by integrating visual and textual understanding within a unified transformer architecture.
arXiv:2606. 20077v1 Announce Type: cross Abstract: Visual tokens enter Large Language Models (LLMs) as raw, foreign signals.
arXiv:2503. 06211v3 Announce Type: replace-cross Abstract: Text-pretrained language models (LMs) encode rich world knowledge, but adapting them to process and generate perceptual modalities such as audio and images while effectively leveraging that knowledge remains challenging.
arXiv:2606. 10147v1 Announce Type: new Abstract: Multimodal Large Language Models (MLLMs) can listen and see, but how do audio and visual signals actually travel through the network to shape an answer?
arXiv:2607. 04423v1 Announce Type: cross Abstract: Unified Multimodal Models (UMMs) integrate image understanding and generation within a single architecture, yet how the two tasks interact remains understudied.
arXiv:2606. 03871v1 Announce Type: cross Abstract: Visual instruction tuning effectively adapts a pre-trained Large Language Model (LLM) to process image information alongside text.
arXiv:2606. 00959v1 Announce Type: new Abstract: Understanding modality interaction in multimodal large language models (MLLMs) is central to reliable deployment.
arXiv:2605. 18160v2 Announce Type: replace-cross Abstract: In recent years, multimodal large language models (MLLMs) have achieved remarkable progress, primarily attributed to effective paradigms for integrating visual and textual information.
arXiv:2602. 14134v2 Announce Type: replace-cross Abstract: Multimodal Large Language Models (MLLMs) have demonstrated exceptional capabilities in high-level visual understanding.
NeoMME is a family of 260M and 800M‑parameter multimodal‑native multilingual encoders that process text and raw image patches in a single bidirectional Transformer. Trained from scratch with a masked discrete‑diffusion objective conditioned on visible image patches, NeoMME supports a 16,384‑token context, enabling encoding of up to two 4K UHD images. In downstream tests, NeoMME‑Retriever models outperform all sub‑800M‑parameter baselines on the ViDoRe v3 benchmark and achieve twice the throughput of ColModernVBERT on an NVIDIA L40S, while hierarchical token pooling and asymmetric quantization compress embeddings 255× with minimal loss in retrieval performance.
arXiv:2510. 25013v2 Announce Type: replace-cross Abstract: Mechanistic interpretability aims to reverse-engineer large language models (LLMs) into human-understandable computational circuits.
Multimodal models often build on architectures designed for generative vision-language modeling, typically combining separately pretrained vision encoders with causal language models. Visual document...