analiza aiqbench
TechPulse take: This research introduces GSM8K-V, a novel benchmark that exposes a critical gap in Vision Language Models' ability to perform mathematical reasoning based on implicit visual cues rather than explicit text. The study reveals that current models struggle significantly with this modality shift, identifying "Implicit Visual Inference Error" as the primary bottleneck preventing human-level performance. This work is essential for understanding the distinct capabilities required for VLMs to navigate complex, image-based problem-solving environments.
GSM8K-V: Can Vision Language Models Solve Grade School Math Word Problems in Visual Contexts
arXiv:2509.25160v2 Announce Type: replace-cross Abstract: Mathematical reasoning is a key capability for vision-language models (VLMs), yet current benchmarks mainly evaluate text-based or explicitly symbolic visual inputs. It remains unclear whether VLMs can reason mathematically when information must be perceived and inferred from images rather than read from explicit symbols. We introduce GSM8K-V, a benchmark transforming GSM8K into multi-image sequences with semantic equivalence preserved. By mapping text-based problems into visual form via an automated pipeline and human verification, we curate 1,319 high-quality samples. In GSM8K-V, quantities must be extracted through visual perception, and reasoning chains must be reconstructed by integrating implicit cues across scenes. Evaluation of 34 VLMs reveals a striking modality gap: while most models exceed 90\% on text, the best model achieves only 59\% on GSM8K-V, far below the 91\% human accuracy. Notably, models enhanced for visual math reasoning show no improvement on GSM8K-V despite large gains on existing benchmarks, confirming that it evaluates a distinct capability. Error analysis shows that the primary bottleneck lies in Implicit Visual Inference Error (IVIE), where models fail to recover visual semantics that are implied rather than explicitly stated. Our code and data are released at https://github.com/ZJU-REAL/GSM8K-V.
Abstract:Mathematical reasoning is a key capability for vision-language models (VLMs), yet current benchmarks mainly evaluate text-based or explicitly symbolic visual inputs. It remains unclear whether VLMs can reason mathematically when information must be perceived and inferred from images rather than read from explicit symbols. We introduce GSM8K-V, a benchmark transforming GSM8K into multi-image sequences with semantic equivalence preserved. By mapping text-based problems into visual form via an automated pipeline and human verification, we curate 1,319 high-quality samples. In GSM8K-V, quantities must be extracted through visual perception, and reasoning chains must be reconstructed by integrating implicit cues across scenes. Evaluation of 34 VLMs reveals a striking modality gap: while most models exceed 90\% on text, the best model achieves only 59\% on GSM8K-V, far below the 91\% human accuracy. Notably, models enhanced for visual math reasoning show no improvement on GSM8K-V despite large gains on existing benchmarks, confirming that it evaluates a distinct capability. Error analysis shows that the primary bottleneck lies in Implicit Visual Inference Error (IVIE), where models fail to recover visual semantics that are implied rather than explicitly stated. Our code and data are released at this https URL.
59 pages, 7 figures, Project Page: this https URL Code: this https URL Datasets: this https URL Accepted at EMNLP 2026 Main Conference. Updated to the camera-ready version with additional experiments, analyses, and revisions
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
arXiv:2509.25160 [cs.CV]
(or arXiv:2509.25160v2 [cs.CV] for this version)
Text integral sub licența deschisă a publicației · 246 cuvinte
Subiecte conexe
[AINews] Meta Muse Wonder & Wonder Full: MSL’s Claude 5.5-tier Frontier Model
Latent Space · 13 oct. 2026
Artificial Intelligence for Mathematical Reasoning: An Integrated Survey of Language Models, Neuro-symbolic Systems, and Verified Discovery
arXiv · 9 iun. 2026
[AINews] High Taste Testers Ep. 1
Latent Space · 17 oct. 2026
Academia is for Ambition — Alex Zhang, MIT
Latent Space · 2 oct. 2026