Entity · paper

MODUS: Decoder-Only Any-to-Any Modeling of Diverse Modalities

paperactiveprovisionalmodus-decoder-only-any-to-any-modeling-of-diverse-modalities-ac71ba9e·1 events·first seen 46h ago

Aliases: MODUS: Decoder-Only Any-to-Any Modeling of Diverse Modalities

Co-occurring entities

MODUS EPFL

More like this (12)

Co-Learning for Missing Arbitrary Modalities in Multi-modal Classification Latent World Recovery for Multimodal Learning with Missing Modalities Hierarchical Acoustic-Semantic Modeling: Modality Separation and Semantic Coherence for Full-Duplex SLMs Unified Multimodal Models (UMMs)Cognitive-structured Multimodal Agent Local Modality Substitution Modality-Informed Reciprocal Reasoning Optimization Modeling Complex Behaviors: Multi-Personality Composition and Dynamic Switching in Vision-Language Models multimodal embedding Multimodal Learning multimodal classification models Multimodal Augmented Generation via Multimodal Retrieval Workshop

Recent events (1)

6arXiv · cs.AI·46h ago·source ↗

MODUS: Decoder-only architecture for any-to-any multimodal modeling without modality-specific heads

Researchers at EPFL introduce Modus, a decoder-only model that treats all modalities symmetrically for any-to-any prediction without modality-specific heads, losses, or task pipelines. Unlike prior any-to-any models that use encoder-decoder or diffusion architectures trained from scratch, Modus leverages pre-trained decoder-only models as a prior. The model supports chained generation through intermediate modalities and cross-modal self-verification, achieving competitive performance with specialist and multitask baselines across benchmarks. All materials are open-sourced.

Frontier Model Releases Multimodal Progress MODUS: Decoder-Only Any-to-Any Modeling of Diverse Modalities MODUS EPFL