In safety-critical traffic scenarios, answering complex questions relies on minute, localized visual cues. However, standard Multimodal Large Language Models (MLLMs) tend to over-attend to backgrounds, overwhelming crucial small objects during visual-language alignment, a failure mode we term 'critical evidence dilution.' Furthermore, existing visual question answering (VQA) datasets rarely expose this flaw, as they lack large-scale, distractor-heavy evaluations that require pinpointing local evidence. To bridge this evaluation and architecture gap, we introduce the Fine-Grained Traffic Reasoning Benchmark (FGTR-Bench) and the Text-Guided Small-Object Reasoning MLLM (TSR-MLLM). FGTR-Bench comprises 40,236 single-image Multiple-Choice Questions (MCQs) created via multi-agent generation, consistency checks, and expert audits, alongside a disjoint 4,947-sample blind test split. To resolve evidence dilution, TSR-MLLM, built on Qwen3-VL-4B, uses a query-conditioned Text-Guided Small-Object Focus (TG-SOF) map. Applied once at the decoder boundary, the map adds sparse Top-K gated residuals to the most question-relevant vision slots while leaving text tokens unchanged. Together with lightweight decoder adaptation, TSR-MLLM preserves single-pass inference without external detectors or image re-encoding. Under matched settings, TSR-MLLM outperforms the strongest 4B baseline by 2.1 points on FGTR-Bench (74.1% overall), with larger gains on evidence-local tracks. Furthermore, it remains competitive on DriveQA-V (CARLA Signs) under greedy decoding without task-specific fine-tuning.
Multimodal large language models (MLLMs) fail at fine-grained visual questions less because they cannot reason than because they never see the evidence: high-resolution images are downsampled before encoding, so the model answers from linguistic priors. The standard remedies are expensive: annotated answers (SFT), hand-engineered verifiers (RLVR), or a large external teacher (on-policy distillation). We ask whether the visual evidence itself can supply the signal for free. We formalize the contrastive evidence gap, the per-token log-likelihood ratio that a model assigns to its own output when conditioned on a question-relevant region versus an irrelevant one, and study it across Qwen2.5-VL-7B, Qwen3-VL-8B, and Qwen3-VL-30B-A3B on V*Bench. Our main positive result is training-free: selecting the candidate crop under which the model's answer distribution is most peaked, using a single-view, label-free criterion, discovers the answer-bearing region with no bounding boxes, training, or labels. It localizes the target 4.4 to 5.1 times better than chance and raises fine-grained accuracy from 70 percent to 85 percent at inference. We further show that the gap is complementary to the model's own confidence. Combining them predicts correctness better than either alone, with AUC up to 0.99, and flags confidently wrong answers, with AUC ranging from 0.97 to 1.00 within the high-confidence subset. All effects concentrate on perception-bottleneck questions and vanish on a global-context control. Finally, we report an honest negative result: converting the same signal into a training method, gated self-distillation (SEG-Distill), does not outperform the base model at pilot scale across three gate designs, while more aggressive gating degrades accuracy. The signal is real, but converting it into training gains remains an open problem.
High benchmark accuracy does not guarantee genuine use of visual evidence. We study this problem in traffic accident Video Question Answering (VideoQA), where correct answers should depend on scene-specific visual evidence but may instead be inferred from textual shortcuts. Through an audit of four public benchmarks, we find that several recent open-weight Vision-Language Models (VLMs) perform competitively, and sometimes better, without video input. On the MM-AU benchmark, removing video consistently improves accuracy, and adding more frames further degrades performance. To quantify visual dependence, we introduce two dataset-level diagnostics: Blind Gap, measuring above-chance text-only performance, and Visual Gain, measuring the marginal benefit of adding video. We further propose an instance-level Shortcut Score that combines text-only confidence with visual necessity signals, enabling continuous, training-free filtering of shortcut-prone questions. The resulting subsets reduce shortcut bias and improve visual grounding. Our findings reveal large differences in grounding quality across benchmarks and show that visually grounded evaluation, not just high accuracy, is essential in safety-critical VideoQA.
Sena Korkut, María Alejandra Bravo Sarmiento, Sanghwan Kim +1
Test-time scaling (TTS) reliably improves reasoning in large language models, but whether it transfers to small open vision-language models remains unclear. We examine this on EXAMS-V, a multilingual visual multiple-choice benchmark, comparing self-consistency, describe-then-reason with PRM-guided beam search, and two post-hoc selectors across Qwen2.5-VL-7B-Instruct and Qwen3.5-4B. What matters is the conditions under which TTS runs, not the search or verification machinery. The largest factor is parseability: an early prompt format left many chains reasoning correctly yet never committing to an answer letter, which a standard answer cue and a guided repair step largely remove. A larger decoding budget removes the rest: raising the per-chain token limit from 1k to 2k recovers 3.7 pp, whereas sampling more chains (8 to 16) adds only 0.15 pp. Once chains have room to finish, elaborate methods contribute little: PRM-guided beam search trails plain self-consistency by 0.39 pp at over eight times the cost, and neither a training-free generative critic nor a trained multimodal PRM beats majority vote across both policies. The largest gain comes instead from the policy model itself (+11.4 pp). Our best configuration reaches 84.1% on the held-out ImageCLEF 2026 test split, ranking first on the Visual MCQ leaderboard.