Peek-a-Boo Reasoning: Contrastive Region Masking in MLLMs
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866911326302896128 |
|---|---|
| author | Chaturvedi, Isha Nair, Anjana Li, Yushen Kumar, Adhitya Rajendra Zhu, Kevin Dev, Sunishchal Panda, Ashwinee Sharma, Vasu |
| author_facet | Chaturvedi, Isha Nair, Anjana Li, Yushen Kumar, Adhitya Rajendra Zhu, Kevin Dev, Sunishchal Panda, Ashwinee Sharma, Vasu |
| contents | We introduce Contrastive Region Masking (CRM), a training free diagnostic that reveals how multimodal large language models (MLLMs) depend on specific visual regions at each step of chain-of-thought (CoT) reasoning. Unlike prior approaches limited to final answers or attention maps, CRM provides causal, step-level attribution by systematically masking annotated regions and contrasting the resulting reasoning traces with unmasked baselines. Applied to datasets such as VisArgs, CRM reveals distinct failure modes: some models preserve reasoning structure, but hallucinate when evidence is missing, while others ground tightly to visual cues yet collapse under perturbations. By shifting the evaluation from correctness of answers to faithfulness of reasoning, CRM reframes visual benchmarks as diagnostic tools, highlighting the need for multimodal evaluation frameworks that measure not just performance, but also robustness and fidelity of reasoning. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_08976 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Peek-a-Boo Reasoning: Contrastive Region Masking in MLLMs Chaturvedi, Isha Nair, Anjana Li, Yushen Kumar, Adhitya Rajendra Zhu, Kevin Dev, Sunishchal Panda, Ashwinee Sharma, Vasu Machine Learning Artificial Intelligence We introduce Contrastive Region Masking (CRM), a training free diagnostic that reveals how multimodal large language models (MLLMs) depend on specific visual regions at each step of chain-of-thought (CoT) reasoning. Unlike prior approaches limited to final answers or attention maps, CRM provides causal, step-level attribution by systematically masking annotated regions and contrasting the resulting reasoning traces with unmasked baselines. Applied to datasets such as VisArgs, CRM reveals distinct failure modes: some models preserve reasoning structure, but hallucinate when evidence is missing, while others ground tightly to visual cues yet collapse under perturbations. By shifting the evaluation from correctness of answers to faithfulness of reasoning, CRM reframes visual benchmarks as diagnostic tools, highlighting the need for multimodal evaluation frameworks that measure not just performance, but also robustness and fidelity of reasoning. |
| title | Peek-a-Boo Reasoning: Contrastive Region Masking in MLLMs |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2512.08976 |