Look, Recite, Then Answer: Enhancing VLM Performance via Self-Generated Knowledge Hints

Fuente: arXiv
Saved in:
Bibliographic Details
Main Author: Feng, Xisheng
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909941050114048
author Feng, Xisheng
author_facet Feng, Xisheng
contents Vision-Language Models (VLMs) exhibit significant performance plateaus in specialized domains like precision agriculture, primarily due to "Reasoning-Driven Hallucination" where linguistic priors override visual perception. A key bottleneck is the "Modality Gap": visual embeddings fail to reliably activate the fine-grained expert knowledge already encoded in model parameters. We propose "Look, Recite, Then Answer," a parameter-efficient framework that enhances VLMs via self-generated knowledge hints while keeping backbone models frozen. The framework decouples inference into three stages: (1) Look generates objective visual descriptions and candidate sets; (2) Recite employs a lightweight 1.7B router to transform visual cues into targeted queries that trigger candidate-specific parametric knowledge; (3) Answer performs parallel evidence alignment between descriptions and recited knowledge to select the most consistent label. On AgroBench, our method achieves state-of-the-art results, improving Weed Identification accuracy by 23.52% over Qwen2-VL-72B and surpassing GPT-4o without external search overhead. This modular design mitigates hallucinations by transforming passive perception into active, controllable knowledge retrieval
format Preprint
id arxiv_https___arxiv_org_abs_2512_00882
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Look, Recite, Then Answer: Enhancing VLM Performance via Self-Generated Knowledge Hints
Feng, Xisheng
Computer Vision and Pattern Recognition
Artificial Intelligence
Vision-Language Models (VLMs) exhibit significant performance plateaus in specialized domains like precision agriculture, primarily due to "Reasoning-Driven Hallucination" where linguistic priors override visual perception. A key bottleneck is the "Modality Gap": visual embeddings fail to reliably activate the fine-grained expert knowledge already encoded in model parameters. We propose "Look, Recite, Then Answer," a parameter-efficient framework that enhances VLMs via self-generated knowledge hints while keeping backbone models frozen. The framework decouples inference into three stages: (1) Look generates objective visual descriptions and candidate sets; (2) Recite employs a lightweight 1.7B router to transform visual cues into targeted queries that trigger candidate-specific parametric knowledge; (3) Answer performs parallel evidence alignment between descriptions and recited knowledge to select the most consistent label. On AgroBench, our method achieves state-of-the-art results, improving Weed Identification accuracy by 23.52% over Qwen2-VL-72B and surpassing GPT-4o without external search overhead. This modular design mitigates hallucinations by transforming passive perception into active, controllable knowledge retrieval
title Look, Recite, Then Answer: Enhancing VLM Performance via Self-Generated Knowledge Hints
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2512.00882