Each Judge Its Own Yardstick: Discovering Per-VLM Taxonomies for Physical Video Evaluation

• URL: http://arxiv.org/abs/2606.22918v1
• Date: Mon, 22 Jun 2026 06:58:39 GMT
• Title: Each Judge Its Own Yardstick: Discovering Per-VLM Taxonomies for Physical Video Evaluation
• Abstract summary: We propose an iterative refinement procedure that discovers a per-VLM evaluation taxonomy.<n>We calibrate the VLM's per-dimension scores to human physical-commonsense ratings, diagnose which dimensions it scores unreliably or redundantly, and prompt an LLM to repair them.<n>The refined taxonomy outperforms the global-schema baseline on held-out videos for every VLM tested, with a mean relative improvement of approximately 32%.
• Score: 73.85089704122315
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Maintaining physical consistency in video generators and world models increasingly relies on vision-language models (VLMs) as automated judges that provide reward signals, ranking decisions, and data-filtering criteria. Yet VLMs differ substantially in training data and architecture, encoding physical phenomena through distinct internal representations. A single global evaluation schema therefore gives every VLM the same axes of competence, regardless of what each can actually perceive. We propose JudgeFit, an iterative refinement procedure that discovers a per-VLM evaluation taxonomy. An initial taxonomy is constructed by prompting the target VLM to enumerate physics errors on a small set of videos and clustering the resulting descriptions. The taxonomy is then refined through a diagnostic step: we calibrate the VLM's per-dimension scores to human physical-commonsense ratings, diagnose which dimensions it scores unreliably or redundantly, and prompt an LLM to repair them, iterating until convergence. We further instantiate this procedure as a benchmark and apply it to 16 VLMs spanning eight model families. The refined taxonomy outperforms the global-schema baseline on held-out videos for every VLM tested, with a mean relative improvement of approximately 32%. Beyond aggregate accuracy, the per-VLM profiles expose model-specific blind spots that overall rankings cannot anticipate, with reliability patterns differing markedly across model families.

Related papers

• ReVSI: Rebuilding Visual Spatial Intelligence Evaluation for Accurate Assessment of VLM 3D Reasoning [59.558706734431276]
  Current evaluations of spatial intelligence can be systematically invalid under modern vision-language model (VLM) settings.<n>We introduce ReVSI, a benchmark and protocol that ensures each QA pair is answerable and correct under the model's actual inputs.
  arXiv  Detail & Related papers  (2026-04-27T10:45:51Z)
• Play Favorites: A Statistical Method to Measure Self-Bias in LLM-as-a-Judge [17.40713507922006]
  Large language models (LLMs) can serve as judges that offer rapid and reliable assessments of other outputs.<n>LLMs may systematically assign overly favorable ratings to their own outputs, a phenomenon known as self-bias.<n>We present a statistical framework that explicitly formalizes assumptions under which self-bias can be identified and estimated.
  arXiv  Detail & Related papers  (2025-08-08T21:22:12Z)
• Do Vision-Language Models Have Internal World Models? Towards an Atomic Evaluation [54.3628937181904]
  Internal world models (WMs) enable agents to understand the world's state and predict transitions.<n>Recent large Vision-Language Models (VLMs), such as OpenAI o3, GPT-4o and Gemini, exhibit potential as general-purpose WMs.
  arXiv  Detail & Related papers  (2025-06-27T03:24:29Z)
• SPARC: Score Prompting and Adaptive Fusion for Zero-Shot Multi-Label Recognition in Vision-Language Models [74.40683913645731]
  Zero-shot multi-label recognition (MLR) with Vision-Language Models (VLMs) faces significant challenges without training data, model tuning, or architectural modifications.<n>Our work proposes a novel solution treating VLMs as black boxes, leveraging scores without training data or ground truth.<n>Analysis of these prompt scores reveals VLM biases and AND''/OR' signal ambiguities, notably that maximum scores are surprisingly suboptimal compared to second-highest scores.
  arXiv  Detail & Related papers  (2025-02-24T07:15:05Z)
• A Benchmark for Crime Surveillance Video Analysis with Large Models [22.683394427744616]
  Anomaly analysis in surveillance videos is a crucial topic in computer vision.<n>In recent years, multimodal large language models (MLLMs) have outperformed task-specific models in various domains.<n>We propose a benchmark for crime surveillance video analysis with large models denoted as UCVL.
  arXiv  Detail & Related papers  (2025-02-13T13:38:17Z)
• FIOVA: A Multi-Annotator Benchmark for Human-Aligned Video Captioning [15.363132825156477]
  We introduce FIOVA, a human-centric benchmark tailored for evaluation of large vision-language models (LVLMs)<n>It comprises 3,002 real-world videos (about 33.6s each), each annotated independently by five annotators.<n>We propose FIOVA-DQ, an event-level evaluation metric that incorporates cognitive weights derived from annotator consensus.
  arXiv  Detail & Related papers  (2024-10-20T03:59:54Z)
• VHELM: A Holistic Evaluation of Vision Language Models [75.88987277686914]
  We present the Holistic Evaluation of Vision Language Models (VHELM) VHELM aggregates various datasets to cover one or more of the 9 aspects: visual perception, knowledge, reasoning, bias, fairness, multilinguality, robustness, toxicity, and safety. Our framework is designed to be lightweight and automatic so that evaluation runs are cheap and fast.
  arXiv  Detail & Related papers  (2024-10-09T17:46:34Z)
• Large Language Models as Automated Aligners for benchmarking Vision-Language Models [48.4367174400306]
  Vision-Language Models (VLMs) have reached a new level of sophistication, showing notable competence in executing intricate cognition and reasoning tasks. Existing evaluation benchmarks, primarily relying on rigid, hand-crafted datasets, face significant limitations in assessing the alignment of these increasingly anthropomorphic models with human intelligence. In this work, we address the limitations via Auto-Bench, which delves into exploring LLMs as proficient curation, measuring the alignment betweenVLMs and human intelligence and value through automatic data curation and assessment.
  arXiv  Detail & Related papers  (2023-11-24T16:12:05Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.