Auto-Encoding Morph-Tokens for Multimodal LLM
This addresses a core challenge in multimodal AI for applications requiring both understanding and generation, though it appears incremental as it builds on existing token-based methods.
The paper tackles the conflicting objectives in multimodal LLMs between visual comprehension (textual output) and generation (visual output) by proposing morph-tokens that serve dual purposes, achieving a new state-of-the-art for both tasks simultaneously.
For multimodal LLMs, the synergy of visual comprehension (textual output) and generation (visual output) presents an ongoing challenge. This is due to a conflicting objective: for comprehension, an MLLM needs to abstract the visuals; for generation, it needs to preserve the visuals as much as possible. Thus, the objective is a dilemma for visual-tokens. To resolve the conflict, we propose encoding images into morph-tokens to serve a dual purpose: for comprehension, they act as visual prompts instructing MLLM to generate texts; for generation, they take on a different, non-conflicting role as complete visual-tokens for image reconstruction, where the missing visual cues are recovered by the MLLM. Extensive experiments show that morph-tokens can achieve a new SOTA for multimodal comprehension and generation simultaneously. Our project is available at https://github.com/DCDmllm/MorphTokens.