ZPPO improves distillation to small vision-language models by using binary and negative candidate prompts plus a replay buffer for hard questions, outperforming standard distillation and GRPO on a 31-benchmark suite with largest gains at the 0.8B scale.
V old: Reasoning transfer from llms to vision-language models via on-policy distillation
7 Pith papers cite this work. Polarity classification is still indexing.
7
Pith papers citing it
abstract
Training vision-language models (VLMs) for complex reasoning remains a challenging task, i.a. due to the scarcity of high-quality image-text reasoning data. Conversely, text-based reasoning resources are abundant and scalable, but it is still an open question how to leveraging them for VLM reasoning. To address this problem, we propose VOLD, a framework to transfer reasoning capabilities from text-only teacher models to VLM student models. To this end, VOLD combines reinforcement learning via Group Relative Policy Optimization (GRPO) with on-policy distillation, which allows the student reasoning traces to be guided by the teacher model, resulting in a significant gain over using GRPO alone. We further show that a cold-start alignment is essential for an effective transfer during the online training phase in this scenario and that without sufficient distributional alignment between teacher and student, on-policy distillation fails to provide meaningful guidance. We evaluate VOLD across diverse benchmarks including MMMU-Pro, MathVision, MathVista, and LogicVista, showing that VOLD outperforms the baseline model significantly and improves over the state of the art by a margin. Our ablation shows the importance of a cold-start alignment via SFT for on-policy distillation with a text-only teacher.
citation-role summary
background 2
citation-polarity summary
years
2026 7roles
background 2polarities
background 2representative citing papers
Local teachability collapse occurs in later trajectory segments during strong-to-weak OPD; a margin-based release rule using top-K teacher advantage and BIC change-point detection on sentence segments outperforms full-trajectory supervision on five in-domain benchmarks and preserves out-of-domain pe
Uni-OPD improves on-policy distillation via student-side data balancing for informative rollouts and teacher-side outcome-guided margin calibration that restores order consistency with rewards.
FiRe-OPD introduces a two-stage filter-then-soft-reweight procedure for trajectory- and token-level supervision in on-policy distillation, claiming gains over prior token-level methods.
Stage-1 warm-starts control the entropy regime entering RL but yield small and localized effects on final in-domain and out-of-domain performance.
citing papers explorer
-
Zone of Proximal Policy Optimization: Teacher in Prompts, Not Gradients
ZPPO improves distillation to small vision-language models by using binary and negative candidate prompts plus a replay buffer for hard questions, outperforming standard distillation and GRPO on a 31-benchmark suite with largest gains at the 0.8B scale.
-
Prefix Teach, Suffix Fade: Local Teachability Collapse in Strong-to-Weak On-Policy Distillation
Local teachability collapse occurs in later trajectory segments during strong-to-weak OPD; a margin-based release rule using top-K teacher advantage and BIC change-point detection on sentence segments outperforms full-trajectory supervision on five in-domain benchmarks and preserves out-of-domain pe
-
Uni-OPD: Unifying On-Policy Distillation with a Dual-Perspective Recipe
Uni-OPD improves on-policy distillation via student-side data balancing for informative rollouts and teacher-side outcome-guided margin calibration that restores order consistency with rewards.
-
Filter, Then Reweight: Rethinking Optimization Granularity in On-Policy Distillation
FiRe-OPD introduces a two-stage filter-then-soft-reweight procedure for trajectory- and token-level supervision in on-policy distillation, claiming gains over prior token-level methods.
-
Stage-1 Controls the Entropy Regime, Not the Outcome
Stage-1 warm-starts control the entropy regime entering RL but yield small and localized effects on final in-domain and out-of-domain performance.
- Learning Visual Spatial Planning from Symbolic State via Modality-Gap-Aware Self-Distillation
- SOD: Step-wise On-policy Distillation for Small Language Model Agents