XBot Study - ReTurn Rebind: leaderboard

The XBot paper compares released multimodal models on audio-video understanding, spatial reasoning and related tasks. Each native benchmark, metric and reported model configuration remains separate. Scores from the paper’s own trained XBot checkpoint are retained outside this released-baseline comparison. This board reports task-specific accuracy (%). Higher is better.

Metric: Task-specific accuracy (%). Source: arxiv.org. Saturation forecast: Around August 2027. 2 models tracked. Updated 14 Nov 2026.

Top models

#ModelScoreOverall rank
1Doubao-Seed-2.1-Lite55.04#172
2Qwen3.5 Omni Flash24.48#568

No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.

Interactive version: theaggregate.ai/benchmark?slug=xbot-study-return-rebind · How It Works · Data refreshed daily, snapshot 2026-11-16.