REVIEW 6 cited by
VL-RewardBench: A Challenging Benchmark for Vision-Language Generative Reward Models
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
abstract
Vision-language generative reward models (VL-GenRMs) play a crucial role in aligning and evaluating multimodal AI systems, yet their own evaluation remains under-explored. Current assessment methods primarily rely on AI-annotated preference labels from traditional VL tasks, which can introduce biases and often fail to effectively challenge state-of-the-art models. To address these limitations, we introduce VL-RewardBench, a comprehensive benchmark spanning general multimodal queries, visual hallucination detection, and complex reasoning tasks. Through our AI-assisted annotation pipeline that combines sample selection with human verification, we curate 1,250 high-quality examples specifically designed to probe VL-GenRMs limitations. Comprehensive evaluation across 16 leading large vision-language models demonstrates VL-RewardBench's effectiveness as a challenging testbed, where even GPT-4o achieves only 65.4% accuracy, and state-of-the-art open-source models such as Qwen2-VL-72B, struggle to surpass random-guessing. Importantly, performance on VL-RewardBench strongly correlates (Pearson's r $>$ 0.9) with MMMU-Pro accuracy using Best-of-N sampling with VL-GenRMs. Analysis experiments uncover three critical insights for improving VL-GenRMs: (i) models predominantly fail at basic visual perception tasks rather than reasoning tasks; (ii) inference-time scaling benefits vary dramatically by model capacity; and (iii) training VL-GenRMs to learn to judge substantially boosts judgment capability (+14.7% accuracy for a 7B VL-GenRM). We believe VL-RewardBench along with the experimental insights will become a valuable resource for advancing VL-GenRMs.
Forward citations
Cited by 6 Pith papers
-
LLaVA-Critic-R1: Your Critic Model is Secretly a Strong Policy Model
RL training on preference-labeled critic data transforms a 7B vision-language model into both a stronger critic and a stronger generative policy, improving average benchmark accuracy by 5.7% and enabling self-critique...
-
VL-GenRM: Enhancing Vision-Language Verification via Vision Experts and Iterative Training
A vision-expert-filtered, chain-of-thought-guided, iteratively fine-tuned reward model boosts a compact 7B model's ability to judge vision-language responses, especially detecting hallucinations.
-
MM-RLHF: The Next Step Forward in Multimodal LLM Alignment
A human-annotated multimodal preference dataset plus critique-based reward modeling and reward-margin-weighted DPO improves MLLM performance across many benchmarks.
-
Agent-RewardBench: Towards a Unified Benchmark for Reward Modeling across Perception, Planning, and Safety in Real-World Multimodal Agents
A new step-level reward-modeling benchmark for multimodal agents shows current MLLMs reach at most 61.6 percent accuracy, and benchmark score correlates strongly (r=0.981 across five models) with downstream A* search ...
-
MiMo-VL Technical Report
MiMo-VL-7B-RL, a 7B open-source vision-language model, reports state-of-the-art results on 35 of 40 benchmarks and a 59.4 OlympiadBench score, with the report crediting long-CoT pretraining data and mixed on-policy RL.
-
VF-Eval: Evaluating Multimodal LLMs for Generating Feedback on AIGC Videos
A new benchmark, VF-Eval, measures how well multimodal LLMs check, detect, and reason about errors in AI-generated videos, and shows frontier models remain far below human performance.
Discussion (0). Continue with ORCID to comment.