AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

TokLIP: Marry Visual Tokens to CLIP for Multimodal Comprehension and Generation

Haokun Lin, Teng Wang, Yixiao Ge, Yuying Ge, Zhichao Lu, Ying Wei, Qingfu Zhang, Zhenan Sun, Ying Shan

IJCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Training-free token reduction for MLLMs · 56 citations (OpenAlex)

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
add [CLS] token distillation from teacher (ablation)TokLIP with contrastive loss · CC3M zero-shot ImageNetTop-1 accuracy: 32.93 -> 52.46Improved
initialize vision encoder with pretrained weights (ablation)TokLIP · CC3M zero-shot ImageNetTop-1 accuracy: 15.73 -> 32.93Improved
use MLP projection instead of codebook mapping (ablation)TokLIP · CC3M zero-shot ImageNetTop-1 accuracy: 35.17 -> 52.46Improved
use MLP instead of soft MLP projection (ablation)TokLIP · CC3M zero-shot ImageNetTop-1 accuracy: 51.53 -> 52.46Improved
distill all tokens instead of [CLS] only (ablation)causal TokLIP · CC3M zero-shot ImageNetTop-1 accuracy: 52.46 -> 51.04Hurt
distill all tokens instead of single token (ablation)bidirectional TokLIP · CC3M zero-shot ImageNetIN Top1: 56.32 -> 59.47Improved
use bidirectional instead of causal attention (ablation)TokLIP · CC3M comprehensionIN Top1: 52.46 -> 56.32Improved
use bidirectional instead of causal attention (ablation)TokLIP · ImageNet generation (LlamaGen-B)FID: 15.05 -> 214.36Hurt
fuse VQGAN and semantic features by concatenation instead of sum (ablation)TokLIP generation · ImageNet generation (LlamaGen-B, 384, 10K samples)FID: 16.37 -> 14.93Improved
add semantic features with learnable weighted sum (ablation)VQGAN generation · ImageNet generation (LlamaGen-B, 384, 10K samples)FID: 17.31 -> 15.46Improved
replace VQGAN tokenizer with TokLIPLlamaGen GPT-B 384 · ImageNet generation (50 epochs)FID: 14.48 -> 12.37Improved
replace VQGAN tokenizer with TokLIPLlamaGen GPT-B 256 · ImageNet generation (50 epochs)FID: 13.21 -> 13.12No clear effect

Later papers that compared with this method

1 later accepted paper reports TokLIP as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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