VLM Post-Training

VLM: Vision-Language Model

Follow this topic Ask about this topic

Latest papers

11 papers4 papers in the last four weeks, against 1 the four weeks before. 0.0% of all new papers.

Do As I Do, Not As I Say: On-Policy Distillation Inherits the Teacher in Vision-Language Models
Oct 14, 2026Frank F. Yang, Kefei Yan, Yiren Wang +1VLM Distillation On-Policy Distillation
Georgia Institute of Technology, Atlanta, GA, USA
Rethinking Contrastive Loss in CLIP Post-training: A Complementary Framework with Frozen Text Encoder
Oct 8, 2026Zidan Wang, Yaqian Li, Xiaokai Zhang +3VLM Post-Training VLM Adaptation
Huazhong University of Science and Technology · Li Auto Inc.
From Reward Signal to Visual Utility: A Controlled Audit of Medical VLM Post-Training
Sep 25, 2026Wang JingxinVLM Post-Training Vision-Language Models
Institute of Neuroscience, CAS
VisionPsy-Nano: Improving Accuracy, Efficiency, and Reliability in On-Device Vision-Language Models
Sep 23, 2026Khurram Azeem Hashmi, Mohammadreza Zolfaghari, Changdae Park +6VLM Post-Training Vision-Language Models
Tether AI Research
Post-Training VLMs for Video Mistake Detection
Aug 28, 2026Federico Spurio, Olga Zatsarynna, Lars Doorenbos +3VLM Post-Training Vision-Language Models
University of Bonn Bonn, Germany · Lamarr Institute for Machine Learning and Artificial Intelligence Bonn, Germany · Toyota Motor Europe Belgium
From Recovery to Drop-off: How Action Post-training Reduces a VLM's Late-Layer Depth Decodability
Aug 9, 2026Alexander Hackett, Arnaud Denis-Remillard, Axel CassouVLM Interpretability Vision-Language-Action Models
New York University · Reflex · Université de Montréal +1
Self-Boosting Vision-Language Models with Noisy Student On-Policy Self-Distillation
Jul 25, 2026Shuai Wang, Daoan Zhang, Zhe Tang +2VLM Adaptation On-Policy Self-Distillation
The Hong Kong University of Science and Technology (Guangzhou) · University of Rochester · Zhejiang University of Technology +1
From Seeing to Thinking: Decoupling Perception and Reasoning Improves Post-Training of Vision-Language Models
May 19, 2026Juncheng Wu, Hardy Chen, Haoqin Tu +6Vision-Language Models VLM Reasoning
Amazon · UC Santa Cruz · University of Waterloo +1
Watch Before You Answer: Learning from Visually Grounded Post-Training
Apr 6, 2026Yuxuan Zhang, EunJeong Hwang, Huaisong Zhang +8VLM Post-Training Vision-Language Models
University of British Columbia · Vector Institute · Etude AI +4
VLM-AutoDrive: Post-Training Vision-Language Models for Safety-Critical Autonomous Driving Events
Mar 18, 2026Mohammad Qazim Bhat, Yufan Huang, Niket Agarwal +7VLM Post-Training VLMs for Autonomous Driving
NVIDIA, California, USA
Learning Transferable Temporal Primitives for Video Reasoning via Synthetic Videos
Mar 18, 2026Songtao Jiang, Sibo Song, Chenyi Zhou +12VLM Post-Training Temporal Video Understanding
Zhejiang University · Qwen Team, Alibaba Group · Shanghai Jiao Tong University +1

Oct 14, 2026J/K move · Enter open · S save

Do As I Do, Not As I Say: On-Policy Distillation Inherits the Teacher in Vision-Language Models

Frank F. Yang, Kefei Yan, Yiren Wang, Yiyang Wang

Georgia Institute of Technology, Atlanta, GA, USA

VLM Distillation On-Policy Distillation VLM Post-Training VLM Reasoning

Open paper AskPDF arXiv

Improving the reasoning capabilities of vision-language models (VLMs) often relies on post-training with supervision from stronger models. Two common strategies are supervised fine-tuning (SFT), which trains on teacher-generated demonstrations, and on-policy distillation (OPD), which provides teacher feedback on trajectories generated by the student itself. While OPD can be more data-efficient by supervising student-generated trajectories, it is unclear how the teacher's supervision at each state affects the student's learned behavior. To study this question, we conduct a controlled comparison of OPD and filtered SFT across data budgets and tasks where the teacher exhibits different levels of reliability, and analyze how these differences shape student accuracy and completion behavior. Across the two tasks, we find that OPD is effective only when the teacher remains reliable on student-generated trajectories. On ChartQA, where the teacher completes 99% of its solutions, OPD trained on only 10% of the data comes within 1.4 accuracy points of full-data SFT. In contrast, on Geometry3K, where the teacher fails to complete 22% of its solutions, OPD underperforms SFT by 7.9 points, with most of the gap explained by incomplete responses. More strikingly, starting OPD from an SFT student reduces its completion rate from 76% to 62%, effectively erasing the completion behavior learned through SFT. Filtering teacher demonstrations for completeness alone recovers much of SFT's advantage. These results show that on-policy supervision can transfer not only teacher knowledge but also persistent failure modes on student-visited states. Code is available at https://github.com/ffy208/opd-inherits-teacher-vlm.