REVIEW 15 cited by
SPA-VL: A Comprehensive Safety Preference Alignment Dataset for Vision Language Model
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
The emergence of Vision Language Models (VLMs) has brought unprecedented advances in understanding multimodal information. The combination of textual and visual semantics in VLMs is highly complex and diverse, making the safety alignment of these models challenging. Furthermore, due to the limited study on the safety alignment of VLMs, there is a lack of large-scale, high-quality datasets. To address these limitations, we propose a Safety Preference Alignment dataset for Vision Language Models named SPA-VL. In terms of breadth, SPA-VL covers 6 harmfulness domains, 13 categories, and 53 subcategories, and contains 100,788 samples of the quadruple (question, image, chosen response, rejected response). In terms of depth, the responses are collected from 12 open-source (e.g., QwenVL) and closed-source (e.g., Gemini) VLMs to ensure diversity. The construction of preference data is fully automated, and the experimental results indicate that models trained with alignment techniques on the SPA-VL dataset exhibit substantial improvements in harmlessness and helpfulness while maintaining core capabilities. SPA-VL, as a large-scale, high-quality, and diverse dataset, represents a significant milestone in ensuring that VLMs achieve both harmlessness and helpfulness.
Forward citations
Cited by 15 Pith papers
-
MDIT-Bench: Evaluating the Dual-Implicit Toxicity in Large Multimodal Models
The authors release a large multimodal benchmark showing that current LMMs struggle to detect toxicity that emerges only from combining image and text, and that many-shot toxic demonstrations further reduce their accuracy.
-
Multimodal Reward Hacking in Reinforcement Learning
Imperfect multimodal RL rewards systematically create new failures (NRFR > RHR); scaling and answer-aware rewards help but do not eliminate hacking, and unreliable visual verifiers actively increase it.
-
When Are Reasoning-Based Guardrails Not Efficient? ResponseGuard: A Fast Vision-Language Guard for Real-Time Moderation
A label-only 2B guard outperforms a 3B reasoning-based vision-language guard on response-harmfulness detection at about 150x lower latency, with the remaining image gap attributed to frozen encoders.
-
GrAInS: Gradient-based Attribution for Inference-Time Steering of LLMs and VLMs
GrAInS uses Integrated Gradients to identify the most influential tokens, then builds layer-wise steering vectors that improve truthfulness, reduce hallucination, and preserve general capabilities in LLMs and VLMs.
-
Bridging the Gap in Vision Language Models in Identifying Unsafe Concepts Across Modalities
Vision-language models consistently recognize unsafe content better from text than from images, and a simplified reinforcement learning fine-tune narrows that gap.
-
The Man Behind the Sound: Demystifying Audio Private Attribute Profiling via Multimodal Large Language Model Agents
A multi-agent audio-language model framework can automatically profile private attributes, such as age, health, and income, directly from general audio recordings.
-
GThinker: Towards General Multimodal Reasoning via Cue-Guided Rethinking
A cue-tagging and rethinking training recipe lifts a 7B multimodal model to 81.5% on M3CoT, though the gain may reflect training on the same benchmark.
-
Bootstrapping LLM Robustness for VLM Safety via Reducing the Pretraining Modality Gap
Reducing the modality gap between image and text embeddings during LVLM pretraining reduces unsafe response rates by up to 16.3% across models and benchmarks.
-
USB: A Comprehensive and Unified Safety Evaluation Benchmark for Multimodal Large Language Models
USB-SafeBench is a unified MLLM safety benchmark with 61 risk categories, 4 modality combinations, and dual-language vulnerability and oversensitivity tests.
-
ShieldVLM: Safeguarding the Multimodal Implicit Toxicity via Deliberative Reasoning with LVLMs
ShieldVLM detects multimodal implicit toxicity through deliberate cross-modal reasoning, outperforming existing moderation APIs and models on the new MMIT benchmark.
-
A Survey on Evaluating Quality and Trustworthiness in LLM-Generated Data
A metric-oriented survey that classifies intrinsic quality and trustworthiness metrics for LLM-generated data across six modalities and documents systematic evaluation gaps in the current literature.
-
Seeing the Threat: Vulnerabilities in Vision-Language Models to Adversarial Attack
A two-stage evaluation framework and token-projection analysis show that LVLMs encode harmful semantic cues from images even without OCR, while remaining vulnerable to cross-modal attacks.
-
VSCBench: Bridging the Gap in Vision-Language Model Safety Calibration
A new benchmark, VSCBench, measures oversafety and undersafety in vision-language models and shows that most models, including proprietary ones, are miscalibrated on at least one safety dimension.
-
A Survey of Safety on Large Vision-Language Models: Attacks, Defenses and Evaluations
A survey of LVLM safety that adds a lifecycle taxonomy and new benchmark results showing Janus-Pro-7B has weaker safety than several open-source LVLMs.
-
Aligning Large Vision-Language Models by Deep Reinforcement Learning and Direct Preference Optimization
A survey categorizing deep reinforcement learning and direct preference optimization methods for aligning large vision-language models, with tables of studies and datasets and no new experimental result.
Discussion (0). Sign in to comment.