AI Paper Atlas

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

Learning Descriptive Image Captioning via Semipermeable Maximum Likelihood Estimation

Zihao Yue, Anwen Hu, Liang Zhang, Qin Jin

NeurIPS 2023 · Vision-language and multimodal › Human and embodied modeling · approach: Concept-guided image captioning · 1 citations (OpenAlex)

Read the paper

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
further train BLIP with SMILE instead of MLEBLIP · MSCOCO testcaption length: 10.0 -> 22.3Improved
further train BLIP with SMILEBLIP · MSCOCO testself-retrieval R@1: 6.7 -> 10.0Improved
further train BLIP with SMILEBLIP · MSCOCO testCLIPScore: 77.2 -> 75.0Hurt
further train BLIP with SMILEBLIP · Flickr30K testself-retrieval R@1: 25.4 -> 31.2Improved
train from scratch decoder with SMILE (no pre-training) (ablation)BLIP F.S. · MSCOCO testself-retrieval R@1: 1.9 -> 0.7Hurt
apply SMILE to pre-trained-only BLIP (ablation)BLIP PT · MSCOCO testself-retrieval R@1: 2.7 -> 8.5Improved
add extra MLE training steps instead of SMILE (ablation)BLIP · MSCOCO testself-retrieval R@1: 6.7 -> 6.5No clear effect
add initial context restriction (first-token MLE) to SMILE (ablation)BLIP-SMILE · MSCOCO testself-retrieval R@1: 6.8 -> 10.0Improved
first-token shifting instead of first-token MLE (ablation)BLIP-SMILE · MSCOCO testself-retrieval R@1: 10.0 -> 9.6Hurt
mix MLE into SMILE objective with lambda=0.01 (ablation)BLIP-SMILE · MSCOCO testself-retrieval R@5: 24.5 -> 25.1Improved
mix MLE into SMILE objective with lambda=0.01 (ablation)BLIP-SMILE · MSCOCO testCLIPScore: 75.0 -> 76.2Improved
use SMILE on MSR-VTT video captioningBLIP4video · MSR-VTTself-retrieval R@1: 19.0 -> 27.3Improved

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