Papers
arxiv:2506.15681

GenRecal: Generation after Recalibration from Large to Small Vision-Language Models

Published on Jun 18
· Submitted by BK-Lee on Jun 19
#3 Paper of the day
Authors:
,

Abstract

GenRecal, a novel distillation framework, improves performance of vision-language models by aligning feature representations across different architectures.

AI-generated summary

Recent advancements in vision-language models (VLMs) have leveraged large language models (LLMs) to achieve performance on par with closed-source systems like GPT-4V. However, deploying these models in real-world scenarios, particularly on resource-constrained devices, remains challenging due to their substantial computational demands. This has spurred interest in distilling knowledge from large VLMs into smaller, more efficient counterparts. A key challenge arises here from the diversity of VLM architectures, which are built on different LLMs and employ varying token types-differing in vocabulary size, token splits, and token index ordering. To address this challenge of limitation to a specific VLM type, we present Generation after Recalibration (GenRecal), a novel, general-purpose distillation framework for VLMs. GenRecal incorporates a Recalibrator that aligns and adapts feature representations between heterogeneous VLMs, enabling effective knowledge transfer across different types of VLMs. Through extensive experiments on multiple challenging benchmarks, we demonstrate that GenRecal significantly improves baseline performances, eventually outperforming large-scale open- and closed-source VLMs.

Community

Paper author Paper submitter
  • Project page: https://byungkwanlee.github.io/GenRecal-page/
  • Authors: Byung-Kwan Lee (1,2*), Ryo Hachiuma (1), Yong Man Ro (2), Yu-Chiang Frank Wang (1,3), Yueh-Hua Wu (1)
  • 1: NVIDIA, 2: KAIST, 3: National Taiwan University
  • *: Work Done during Internship

Sign up or log in to comment

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2506.15681 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2506.15681 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2506.15681 in a Space README.md to link it from this page.

Collections including this paper 2