REVIEW 2 cited by
Cobra Effect in Reference-Free Image Captioning Metrics
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
Cobra Effect in Reference-Free Image Captioning Metrics
read the original abstract
Evaluating the compatibility between textual descriptions and corresponding images represents a core endeavor within multi-modal research. In recent years, a proliferation of reference-free methods, leveraging visual-language pre-trained models (VLMs), has emerged. Empirical evidence has substantiated that these innovative approaches exhibit a higher correlation with human judgment, marking a significant advancement in the field. However, does a higher correlation with human evaluations alone sufficiently denote the complete of a metric? In response to this question, in this paper, we study if there are any deficiencies in reference-free metrics. Specifically, inspired by the Cobra Effect, we utilize metric scores as rewards to direct the captioning model toward generating descriptions that closely align with the metric's criteria. If a certain metric has flaws, it will be exploited by the model and reflected in the generated sentences. Our findings reveal that descriptions guided by these metrics contain significant flaws, e.g. incoherent statements and excessive repetition. Subsequently, we propose a novel method termed Self-Improving to rectify the identified shortcomings within these metrics. We employ GPT-4V as an evaluative tool to assess generated sentences and the result reveals that our approach achieves state-of-the-art (SOTA) performance. In addition, we also introduce a challenging evaluation benchmark called Flaws Caption to evaluate reference-free image captioning metrics comprehensively. Our code is available at https://github.com/aaronma2020/robust_captioning_metric
Forward citations
Cited by 2 Pith papers
-
Do Image-Text Metrics Respect Semantic Invariances?
Five image-text metrics exhibit non-semantic sensitivities to spatial, object, and socio-linguistic perturbations, shifting scores by 6-9% on average and flipping rankings in up to 37% of cases, with a proposed post-h...
-
BEiTScore: Reference-free Image Captioning Evaluation with an Efficient Cross-Encoder Model
BEiTScore is a new efficient cross-encoder metric for reference-free image captioning evaluation that achieves state-of-the-art results on detailed caption benchmarks through VQA initialization and adversarial LLM aug...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.