AI Paper Atlas

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

Fusing Pre-Trained Language Models With Multimodal Prompts Through Reinforcement Learning

Youngjae Yu, Jiwan Chung, Heeseung Yun, Jack Hessel, Jae Sung Park, Ximing Lu, Rowan Zellers, Prithviraj Ammanabrolu, Ronan Le Bras, Gunhee Kim, Yejin Choi

CVPR 2023 · Vision-language and multimodal › Human and embodied modeling · approach: Vision-language pretraining · 15 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
Use CLIP-reward RL alignment with an unpaired caption-domain language modelESPER-CaptionLM · COCO test splitCIDEr: 63.6 -> 78.2Improved
Use CLIP-reward RL alignment with an unpaired caption-domain language modelESPER-CaptionLM · COCO test splitinference time: 65s -> 0.65sImproved
Initialize supervised fine-tuning with an encoder trained by ESPER RL (ablation)ESPERInit-Full · COCO Captions test splitCIDEr: 108.4 -> 111.1Improved
Initialize encoder-only supervised fine-tuning with an encoder trained by ESPER RL (ablation)ESPERInit-MLP · COCO Captions test splitCIDEr: 94.4 -> 103.1Improved
Apply ESPER RL multimodal alignment to a text-only commonsense modelESPER-VCGLM · Visual Commonsense Graph validation splitBLEU-4: 9.9 -> 13.0Improved
Apply ESPER RL multimodal alignment to COMETESPER-COMET · COCO Captions + COMET validation splitBLEU-4: 17.8 -> 28.3Improved
Apply ESPER RL alignment to a news-domain language modelESPER-NewsLM · VisualNews test splitCIDEr: 1.3 -> 4.6Improved
Apply ESPER RL alignment to a dialogue-domain language modelESPER-DialogLM · VisDial validation splitR@1: 5.7 -> 14.6Improved
Use an audio reward model and unpaired audio data for multimodal alignmentESPER-CaptionLM-Audio · AudioCaps test splitCIDEr: 6.21 -> 7.92Improved
Use an unpaired domain-specific caption language model rather than general GPT-2ESPER · COCO test splitCIDEr: 29.1 -> 78.2Improved
Use an unpaired domain-specific language model rather than general GPT-2ESPER · AudioCaps test splitCIDEr: 2.84 -> 7.92Improved
Use zero-shot ESPER-NewsLM rather than ESPERInit-MLPESPER · VisualNews test splitCIDEr: 15.7 -> 4.6Hurt

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