CVJun 23

Ill-Posed by Design: Probing Evidence Use in VLMs

arXiv:2606.243357.6
Predicted impact top 66% in CV · last 90 daysOriginality Incremental advance
AI Analysis

For researchers studying evidence selection in VLMs, this work provides a new diagnostic framework that reveals limitations in how models integrate visual cues, though the findings are incremental as they confirm known issues with a novel method.

The paper proposes monocular metric object-size estimation as an ill-posed diagnostic setting to probe evidence use in VLMs, finding that even the largest models (up to 397B parameters) trail a text-only LLM on in-the-wild data, and that models rely primarily on target identity while neglecting scene geometry.

Counterfactual analysis is widely used to study evidence use in vision-language models, but its diagnostic value is limited on well-posed tasks: when several cues independently support the same answer, removing one may not change the prediction. We propose monocular metric object-size estimation as an ill-posed diagnostic setting for evidence selection: because physical size cannot be determined from a single uncalibrated image, models must rely on imperfect cues category priors, target appearance, local context, apparent image size, and scene geometry. We assemble Metric VQA ($10{,}813$ dimension queries from Objectron and $331$ tape-measured in-the-wild scenes) and evaluate $12$ open-weight VLMs ($3$--$397$\,B parameters) with counterfactual analysis decomposing six visual and language evidence channels. Even the largest VLMs tested (Qwen3-VL-235B, Qwen3.5-397B, InternVL3.5-241B) trail a text-only frontier LLM on the in-the-wild split. The diagnostic analysis shows: target identity is the most load-bearing cue, target pixels and local context help only some models, apparent size shifts predictions without a directional readout, and global scene geometry is largely unused. We analyze LoRA fine-tuning as an actionable intervention specific to metric estimation: while the task is learnable, the models do not learn to leverage scene geometry.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes