demon-bench.

Multi Image Reasoning

Multi Image Reasoning is the reported evaluation metric for demon-bench. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Muted rows were not state of the art when published — an earlier or same-year result already scored better.

Grounded Qa

Grounded Qa is the reported evaluation metric for demon-bench. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Muted rows were not state of the art when published — an earlier or same-year result already scored better.

Knowledge Images Qa

Knowledge Images Qa is the reported evaluation metric for demon-bench. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Muted rows were not state of the art when published — an earlier or same-year result already scored better.

Multimodal Dialogue

Multimodal Dialogue is the reported evaluation metric for demon-bench. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Muted rows were not state of the art when published — an earlier or same-year result already scored better.

Accuracy

Accuracy is the reported evaluation metric for demon-bench. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Muted rows were not state of the art when published — an earlier or same-year result already scored better.

Visual Inference

Visual Inference is the reported evaluation metric for demon-bench. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Muted rows were not state of the art when published — an earlier or same-year result already scored better.

Relation Cloze

Relation Cloze is the reported evaluation metric for demon-bench. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Muted rows were not state of the art when published — an earlier or same-year result already scored better.

Storytelling

Storytelling is the reported evaluation metric for demon-bench. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Muted rows were not state of the art when published — an earlier or same-year result already scored better.