Through the Lens of Contrast: Self-Improving Visual Reasoning in VLMs
Zhiyu Pan, Yizheng Wu, Jiashen Hua, Junyi Feng, Shaotian Yan, Bing Deng, Zhiguo Cao, Jieping Ye
Abstract
Reasoning has emerged as a key capability of large language models. In linguistic tasks, this capability can be enhanced by self-improving techniques that refine reasoning paths for subsequent fine-tuning. However, extending these language-based self-improving approaches to vision language models (VLMs) presents a unique challenge: visual hallucinations in reasoning paths cannot be effectively verified or rectified. Our solution starts with a key observation about visual contrast: when presented with a contrastive VQA pair, i.e., two visually similar images with synonymous questions, VLMs identify relevant visual cues more precisely compared with when given a single VQA sample. Motivated by this observation, we propose Visual Contrastive Self-Taught Reasoner (VC-STaR), a novel self-improving framework that leverages visual contrast to mitigate hallucinations in model-generated rationales. We collect a diverse suite of VQA datasets, curate contrastive pairs according to multi-modal similarity, and generate rationales using VC-STaR. Consequently, we obtain a new visual reasoning dataset, VisCoR-$55$K, which is then used to boost the reasoning capability of various VLMs through supervised finetuning. Extensive experiments show that VC-STaR not only outperforms existing self-improving approaches but also surpasses models finetuned on the SoTA visual reasoning datasets, demonstrating that the inherent contrastive ability of VLMs can bootstrap their own visual reasoning. The code, dataset and trained models will be released upon acceptance.
VC-STaR mitigates visual hallucinations through contrastive VQA pairs for self-improving visual reasoning.
- Observation that VLMs identify visual cues more precisely when presented with contrastive VQA pairs
- Self-improving framework leveraging visual contrast to mitigate hallucinations in model-generated rationales
- VisCoR-55K dataset curated with contrastive pairs according to multi-modal similarity
- Outperforms self-improving baselines and models trained on SoTA visual reasoning datasets
- Self-improving learning
- Contrastive visual reasoning
- VQA pairs
- Supervised fine-tuning
- VisCoR-55K
Authors did not state explicit limitations.
Authors did not state explicit future directions.
Author keywords
- Reasoning
- Vision-Language Models
- Contrasting
Related orals
Benchmarking Empirical Privacy Protection for Adaptations of Large Language Models
Benchmarks practical privacy risks in differential privacy-adapted LLMs, revealing distribution shifts and model choice impact effectiveness.
Half-order Fine-Tuning for Diffusion Model: A Recursive Likelihood Ratio Optimizer
Proposes Recursive Likelihood Ratio optimizer for efficient fine-tuning of diffusion models with lower variance gradient estimation.
Invisible Safety Threat: Malicious Finetuning for LLM via Steganography
Demonstrates LLMs can be finetuned to generate harmful steganographically-hidden outputs while appearing benign to safety systems.
Reducing Belief Deviation in Reinforcement Learning for Active Reasoning of LLM Agents
Proposes T3 algorithm to detect belief deviation in LLM agents and truncate trajectories for improved reinforcement learning in active reasoning tasks.
RefineStat: Efficient Exploration for Probabilistic Program Synthesis
RefineStat enforces semantic constraints and applies diagnostic-aware refinement for synthesizing valid probabilistic programs from smaller language models.