The Art of Saying "Maybe": A Conformal Lens for Uncertainty Benchmarking in VLMs (2026.findings-eacl)
Copied to clipboard
| Challenge: | Recent advances in large vision-language models have led to remarkable progress in complex visual understanding across scientific and reasoning tasks. |
| Approach: | They evaluate 18 state-of-the-art vision-language models across 6 multimodal datasets with 3 distinct scoring functions and develop instruction-guided likelihood proxies for closed-source models lacking token-level logprob access. |
| Outcome: | The proposed model is able to achieve higher accuracy on multimodal benchmarks while performing poorer on reasoning tasks. |
Similar Papers
Seeing is Believing, but How Much? A Comprehensive Analysis of Verbalized Calibration in Vision-Language Models (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing approaches to quantify uncertainty are limited in vision-language models . however, current models display notable miscalibration across diverse tasks and settings . |
| Approach: | They evaluate verbalized confidence in vision-language models using visual reasoning . they propose a prompting strategy that improves confidence alignment in multimodal settings . |
| Outcome: | The proposed method improves confidence alignment across multimodal settings. |
Unveiling Uncertainty: A Deep Dive into Calibration and Performance of Multimodal Large Language Models (2025.coling-main)
Copied to clipboard
| Challenge: | Multimodal large language models combine visual and textual data for tasks like image captioning and visual question answering. |
| Approach: | They propose temperature scaling and iterative prompt optimization to calibrate MLLMs and enhance model reliability. |
| Outcome: | The proposed techniques improve MLLMs and improve model reliability. |
AlignMMBench: Evaluating Chinese Multimodal Alignment in Large Vision-Language Models (2025.acl-long)
Copied to clipboard
| Challenge: | Existing benchmarks focus on basic abilities using nonverbal methods, such as yes-no and multiple-choice questions. |
| Approach: | They propose a benchmark that provides more nuanced evaluations of alignment capabilities for large Vision-Language Models (VLMs) they use a rule-calibrated evaluator that exceeds GPT-4's evaluation ability and a “alignment score” to assess the robustness and stability of models across diverse prompts. |
| Outcome: | The proposed benchmark covers 13 tasks across three categories and includes both single-turn and multi-turn dialogue scenarios. |
Can MLLMs Reason Beyond Language? VisReason: A Comprehensive Benchmark for Vision-Centric Reasoning (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent advances in multimodal large language models demonstrate strong performance on visual reasoning benchmarks. |
| Approach: | They propose a benchmark for vision-centric reasoning that integrates visual and textual information for non-trivial reasoning. |
| Outcome: | The proposed benchmark exposes gaps between humans and current MLLMs and reveals limited benefits from test-time reasoning strategies. |
Exploring Response Uncertainty in MLLMs: An Empirical Evaluation under Misleading Scenarios (2025.emnlp-main)
Copied to clipboard
Yunkai Dang, Mengxi Gao, Yibo Yan, Xin Zou, Yanggan Gu, Jungang Li, Jingyu Wang, Peijie Jiang, Aiwei Liu, Jia Liu, Xuming Hu
| Challenge: | Existing studies have focused mainly on visual–textual misalignment, leaving largely unexplored the MLLMs’ ability to preserve an original correct answer when confronted with misleading information. |
| Approach: | They propose a two-stage evaluation pipeline to quantify the response uncertainty phenomenon by eliciting each model’s original response on unperturbed inputs and injecting explicit (false-answer hints) and implicit (contextual contradictions) misleading instructions. |
| Outcome: | The proposed model overturns a correct answer in 65% of cases after receiving a single deceptive cue. |
Tiny Scales, Great Challenges: The Limits of Multimodal LLMs in Scale Recognition (2026.acl-long)
Copied to clipboard
| Challenge: | Existing benchmarks focus on a single type of quantity or a specific format, lacking a comprehensive evaluation of scale recognition capabilities. |
| Approach: | They propose a visual scale recognition benchmark built using images from COCO, Open Images, and Flickr to evaluate scale recognition capabilities of multimodal large language models. |
| Outcome: | The proposed model achieves 42.60% accuracy, lower than the 97.40% of humans. |
Mind’s Eye: A Benchmark of Visual Abstraction, Transformation and Composition for Multimodal LLMs (2026.acl-long)
Copied to clipboard
| Challenge: | Existing evaluations of multimodal large language models (MLLMs) have demonstrated compelling visual understanding in recent years. |
| Approach: | They propose a multimodal large language model with eight visuo-cognitive tasks inspired by classic human intelligence tests organized under a novel A–R–T taxonomy: Abstraction, Relation, and Transformation. |
| Outcome: | The proposed frameworks are based on eight visuo-cognitive tasks inspired by human intelligence tests and organized under a novel A–R–T taxonomy: Abstraction, Relation, and Transformation. |
MathSight: A Benchmark Exploring Have Vision-Language Models Really Seen in University-Level Mathematical Reasoning? (2026.acl-long)
Copied to clipboard
| Challenge: | Existing benchmarks rarely isolate how much visual information contributes to reasoning . a growing collection of benchmarks has catalyzed rapid progress in multimodal reasoning - but how much it contributes remains unclear . |
| Approach: | They propose a university-level multimodal mathematical reasoning benchmark to quantify the effect of visual input. |
| Outcome: | The proposed benchmark disentangles and quantifies the effect of visual input on multimodal reasoning models. |
From Behavioral Performance to Internal Competence: Interpreting Vision-Language Models with VLM-Lens (2025.emnlp-demos)
Copied to clipboard
Hala Sheta, Eric Haoran Huang, Shuyu Wu, Ilia Alenabi, Jiajun Hong, Ryker Lin, Ruoxi Ning, Daniel Wei, Jialin Yang, Jiawei Zhou, Ziqiao Ma, Freda Shi
| Challenge: | Existing vision-language models are based on exactmatch based accuracy and its derivations to evaluate performance. |
| Approach: | They propose a toolkit that supports systematic benchmarking, analysis, and interpretation of vision-language models by extracting intermediate outputs from any layer during the forward pass of open-source VLMs. |
| Outcome: | The proposed toolkit supports 16 state-of-the-art base VLMs and their over 30 variants and is extensible to accommodate new models without changing the core logic. |
An Examination of the Compositionality of Large Generative Vision-Language Models (2024.naacl-long)
Copied to clipboard
| Challenge: | Recent studies have focused on the compositionality of vision-language models (VLMs) however, the performance of GVLMs in multimodal compositional reasoning remains under-explored. |
| Approach: | They propose a syntactical bias score to quantify GVLMs' syntaktical bias . they propose 'SADE' task to assess GVLs's robustness against inclination toward syntical correctness. |
| Outcome: | The proposed benchmarks are based on evaluation metrics and current benchmarks. |