CoVeR: Coverage-Based Token Pruning for Multi-View 3D Reasoning in VLMs

Hugging Face Daily Papers Papers
token-pruning vlm 3d-reasoning coverage-based training-free multi-view visual-tokens

Summary

CoVeR is a training-free spatial token selector that improves 3D reasoning in Vision-Language Models by enforcing exact token budgets and full scene coverage, outperforming prior state-of-the-art methods.

Representing a 3D scene as multi-view images allows 2D VLMs to reason in 3D by reusing priors from pre-training, sidestepping the scarcity of annotated 3D data. However, it produces thousands of redundant visual tokens whose cost grows with every view. Existing visual token pruners fall into two families, each limited in the 3D multi-view setting. Learned importance methods rank tokens by attention or encoder features; because redundancy here is fundamentally spatial, they keep near-duplicate tokens from a few prominent regions and leave most of the scene unrepresented. Voxelization methods improve spatial coverage but cannot enforce an exact token budget and saturate as multi-view observations overlap in 3D, capping retention well below the target. We show that spatial coverage is associated with 3D reasoning performance and introduce CoVeR, a deterministic, training-free selector that uses only token coordinates, with no learned signals. CoVeR selects tokens that collectively cover every region of the scene, and solves the limitations of both families: it enforces an exact per-scene budget, breaks the voxelization saturation plateau, and avoids the near-duplicate selections of learned importance. Extensive experiments show CoVeR outperforms prior SOTAs on all three 3D reasoning benchmarks and generalizes as a plug-and-play module tested across four VLMs. Notably, with only approx8% of visual tokens, it preserves 93.5% of full-token performance, surpassing SOTA by 3.9 percentage points on average across benchmarks.
Original Article
View Cached Full Text

Cached at: 09/09/26, 08:31 AM

Paper page - CoVeR: Coverage-Based Token Pruning for Multi-View 3D Reasoning in VLMs

Source: https://huggingface.co/papers/2609.08345

Abstract

CoVeR is a training-free spatial token selector that preserves 3D reasoning performance by enforcing exact budgets and full scene coverage across multi-view visual tokens.

Representing a 3D scene as multi-view images allows2D VLMsto reason in 3D by reusing priors from pre-training, sidestepping the scarcity of annotated 3D data. However, it produces thousands of redundant visual tokens whose cost grows with every view. Existing visual token pruners fall into two families, each limited in the 3D multi-view setting.Learned importancemethods rank tokens by attention or encoder features; because redundancy here is fundamentally spatial, they keep near-duplicate tokens from a few prominent regions and leave most of the scene unrepresented.Voxelizationmethods improvespatial coveragebut cannot enforce an exacttoken budgetand saturate as multi-view observations overlap in 3D, capping retention well below the target. We show thatspatial coverageis associated with 3D reasoning performance and introduceCoVeR, a deterministic, training-free selector that uses only token coordinates, with no learned signals.CoVeRselects tokens that collectivelycoverevery region of the scene, and solves the limitations of both families: it enforces an exact per-scene budget, breaks thevoxelizationsaturation plateau, and avoids the near-duplicate selections oflearned importance. Extensive experiments showCoVeRoutperforms prior SOTAs on all three 3D reasoning benchmarks and generalizes as a plug-and-play module tested across four VLMs. Notably, with only approx8% of visual tokens, it preserves 93.5% of full-token performance, surpassing SOTA by 3.9 percentage points on average across benchmarks.

View arXiv pageView PDFProject pageGitHub2Add to collection

Get this paper in your agent:

hf papers read 2609\.08345

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2609.08345 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2609.08345 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2609.08345 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

CaVe-VLM-CoT: An Interpretable Vision-Language Model Framework

arXiv cs.AI

CaVe-VLM-CoT is a modular reflection-based agentic-RAG framework for vision-language models that enforces evidence-grounded reasoning through a five-stage pipeline, achieving 87.1% accuracy on ScienceQA and proposing a suite of 23 metrics for evaluation.

OneVL: One-Step Latent Reasoning and Planning with Vision-Language Explanation

Hugging Face Daily Papers

OneVL is a unified vision-language-action framework that compresses chain-of-thought reasoning into latent tokens supervised by both language and visual world model decoders, achieving state-of-the-art trajectory prediction accuracy for autonomous driving at answer-only inference latency. It is the first latent CoT method to surpass explicit CoT across four benchmarks.