Rexverse 2M: Multimodal Data for Object-Grounded Language Models

ChatRex is a Multimodal Large Language Model (MLLM) designed to integrate fine-grained object perception with language understanding. The dataset, created by IDEA-Research, supports this model with a decoupled, retrieval-based architecture for object detection using high-resolution visual inputs. It was last updated on January 23, 2025.

Use Cases

• Training retrieval-based object detection models based on the described decoupled architecture.
• Benchmarking MLLM performance on tasks requiring simultaneous perception and language generation.
• Developing models for visual question answering with fine-grained object grounding.
• Researching methods to integrate high-resolution visual inputs with language models.

Strengths

• Dataset is associated with a specific, described MLLM architecture (ChatRex).
• Model description explicitly mentions a retrieval-based approach for object detection.
• Model description highlights the use of high-resolution visual inputs.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset size are unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

IDEA-Research

Collection Method

Likely created to train and evaluate the ChatRex MLLM, but specific gathering method is not detailed.

Freshness

Last updated 2025-01-23 09:54:53; freshness should be verified.