Qwen Look Again: Guiding Vision-Language Reasoning Models to Re-attention Visual Information
read the original abstract
Inference time scaling drives extended reasoning to enhance the performance of Vision-Language Models (VLMs), thus forming powerful Vision-Language Reasoning Models (VLRMs). However, long reasoning dilutes visual tokens, causing visual information to receive less attention and may trigger hallucinations. Although introducing text-only reflection processes shows promise in language models, we demonstrate that it is insufficient to suppress hallucinations in VLMs. To address this issue, we introduce Qwen-LookAgain (Qwen-LA), a novel VLRM designed to mitigate hallucinations by incorporating a vision-text reflection process that guides the model to re-attention visual information during reasoning. We first propose a reinforcement learning method Balanced Reflective Policy Optimization (BRPO), which guides the model to decide when to generate vision-text reflection on its own and balance the number and length of reflections. Then, we formally prove that VLRMs lose attention to visual tokens as reasoning progresses, and demonstrate that supplementing visual information during reflection enhances visual attention. Therefore, during training and inference, Visual Token COPY and Visual Token ROUTE are introduced to force the model to re-attention visual information at the visual level, addressing the limitations of text-only reflection. Experiments on multiple visual QA datasets and hallucination metrics indicate that Qwen-LA achieves leading accuracy performance while reducing hallucinations. Our code is available at: https://github.com/Liar406/Look_Again
This paper has not been read by Pith yet.
Forward citations
Cited by 11 Pith papers
-
DeepLatent: Think with Images via Parallel Latent Visual Reasoning
DeepLatent introduces a parallel latent visual reasoning framework with learnable 2D tokens and continuous RL, trained via distillation then RL, plus a new 180K dataset, claiming SOTA benchmark results.
-
Towards Order Fairness: Mitigating LLMs Order Sensitivity through Dual Group Advantage Optimization
DGAO uses reinforcement learning to optimize LLMs for both accuracy and order stability by balancing intra-group accuracy advantages and inter-group stability advantages.
-
Reflection Anchors for Propagation-Aware Visual Retention in Long-Chain Multimodal Reasoning
RAPO uses an information-theoretic lower bound on visual gain to select high-entropy reflection anchors and optimizes a chain-masked KL surrogate, delivering gains over baselines on reasoning benchmarks across LVLM backbones.
-
VIDA: A dataset for Visually Dependent Ambiguity in Multimodal Machine Translation
VIDA provides 2,500 visually-dependent ambiguous translation examples and span-level disambiguation metrics; CoT-SFT on LVLMs improves out-of-distribution performance over standard SFT.
-
Vision Inference Former: Sustaining Visual Consistency in Multimodal Large Language Models
Vision Inference Former adds a direct visual-to-output bridge that continuously injects visual semantics during MLLM decoding to sustain consistency and reduce modality imbalance.
-
Vision Inference Former: Sustaining Visual Consistency in Multimodal Large Language Models
VIF is a new inference-time module that maintains visual grounding in MLLMs by directly bridging pure visual representations to the output space throughout generation.
-
PDCR: Perception-Decomposed Confidence Reward for Vision-Language Reasoning
PDCR improves vision-language reasoning by computing separate normalized confidence advantages for perception steps and reasoning steps after unsupervised decomposition.
-
VIDA: A dataset for Visually Dependent Ambiguity in Multimodal Machine Translation
VIDA provides 2,500 visually-dependent ambiguous MT instances and LLM-judge metrics; chain-of-thought SFT improves disambiguation accuracy over standard SFT, especially out-of-distribution.
-
The Landscape of Agentic Reinforcement Learning for LLMs: A Survey
Survey that defines agentic RL for LLMs via POMDPs, introduces a taxonomy of planning/tool-use/memory/reasoning capabilities and domains, and compiles open environments from over 500 papers.
-
Trust Region On-Policy Distillation
TrOPD stabilizes on-policy distillation for LLMs with trust-region learning, outlier estimation, and off-policy guidance, outperforming prior OPD methods on reasoning and code benchmarks.
-
A Survey of Reinforcement Learning for Large Reasoning Models
A survey compiling RL methods, challenges, data resources, and applications for enhancing reasoning in large language models and large reasoning models since DeepSeek-R1.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.