AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

Uni-X: Mitigating Modality Conflict with a Two-End-Separated Architecture for Unified Multimodal Models

Jitai Hao, Hao Liu, Xinyan Xiao, Qiang Huang, Jun Yu

ICLR 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Vision-language pretraining

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
separate shallow and deep layers by modality (Uni-X 9:5) vs fully shared transformerShared Transformer · identical training (28B tokens), Qwen2.5-1.5Baverage of MMLU/GenEval/MMB: 38.0 -> 41.6Improved
Uni-X vs hard-routed vision MoEHardMoE · identical training (28B tokens), Qwen2.5-1.5Baverage: 41.3 -> 41.6Improved
Uni-X vs MoT and UniFork baselinesMoT, UniFork · identical training (28B tokens), Qwen2.5-1.5Baverage: MoT 34.6, UniFork 29.5 -> 41.6Improved
add loss masking on instruction tokens in pretraining (ablation)Uni-X (9:5) · identical training (28B tokens), Qwen2.5-1.5BGenEval: 34.8 -> 43.3Improved
increase total separated layers from 6 (3:3) to 14 (7:7) (ablation)Uni-X · Qwen2.5-1.5Baverage: 38.9 -> 40.1Improved
increase separated layers to 22 (11:11) (ablation)Uni-X · Qwen2.5-1.5Baverage: 40.1 -> 39.8Hurt
use more shallow than deep separated layers (9:5 vs 5:9) (ablation)Uni-X · Qwen2.5-1.5Baverage: 39.1 -> 41.6Improved
use 11:3 split (too few deep separated layers) (ablation)Uni-X · Qwen2.5-1.5BGenEval: 43.3 -> 25.1Hurt
increase vision-specific separated layers (14:8 to 14:20) (ablation)Uni-X · Qwen2.5-1.5Baverage: 37.4 -> 41.2Improved
scale Uni-X from 140B to 240B training tokensUni-X · Qwen2.5-3BGenEval: 82 -> 83Improved
use Uni-X architecture vs shared transformer at inferenceUni-X · H800 inferencetokens/s: 975.2 -> 910.2Mixed

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy