{"total":11,"items":[{"citing_arxiv_id":"2607.00461","ref_index":22,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Multimodal Continuous Reasoning via Asymmetric Mutual Variational Learning","primary_cat":"cs.CV","submitted_at":"2026-07-01T05:29:02+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"AMVL applies bidirectional KL calibration to align answer-agnostic prior with answer-conditioned posterior in variational multimodal reasoning, reducing leakage and yielding +10.83 average gain on BLINK benchmark.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.30168","ref_index":2,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Latent Noise Mask for Reducing Visual Redundancy in Multimodal Large Language Models","primary_cat":"cs.CV","submitted_at":"2026-06-29T11:44:09+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Lens purifies visual evidence in MLLMs via question-conditioned latent noise masking with a LET token, yielding 2.4-6.4 point gains on VQA and grounding tasks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.24233","ref_index":5,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Latent Visual States for Efficient Multimodal Reasoning","primary_cat":"cs.CV","submitted_at":"2026-06-23T07:22:22+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"EVA generates adaptive Latent_slot tokens as internal visual thoughts, trained end-to-end with text tokens via D-GSPO on the EVA-230K dataset, claiming performance gains and better inference efficiency.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.17888","ref_index":110,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"MathVis-Fine: Aligning Visual Supervision with Necessity via Progressive Dependency-Guided Training for Multimodal Mathematical Reasoning","primary_cat":"cs.AI","submitted_at":"2026-06-16T13:09:32+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"MathVis-Fine proposes a dataset with fine-grained visual annotations and dependency ratings plus a progressive two-stage training paradigm to align visual supervision with sample-specific necessity in multimodal mathematical reasoning.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.11740","ref_index":135,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"UniReason-Med: A Shared Grounded Reasoning Interface for 2D-to-3D Transfer in Medical VQA","primary_cat":"cs.CV","submitted_at":"2026-06-10T07:16:27+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"UniReason-Med introduces a unified framework for 2D and 3D medical VQA with shared grounded reasoning, trained on a 220K dataset, claiming that joint 2D+3D supervision improves 3D performance over 3D-only training.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.09585","ref_index":20,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Optical Reasoning: Rethinking Images as an Expressive Reasoning Medium Beyond Text","primary_cat":"cs.AI","submitted_at":"2026-06-08T14:58:59+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Optical reasoning encodes rationales in images rather than text, matching or exceeding text-based performance on math, science, and multimodal benchmarks while cutting tokens by 28.57% on language tasks and 16% on multimodal tasks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.08728","ref_index":74,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Artificial Intelligence for Mathematical Reasoning: An Integrated Survey of Language Models, Neuro-symbolic Systems, and Verified Discovery","primary_cat":"cs.AI","submitted_at":"2026-06-07T16:50:07+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":3.0,"formal_verification":"none","one_line_summary":"An integrated survey organizing AI mathematical reasoning into informal, formal, discovery, and technique axes while cataloging benchmarks and assessing failure modes.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.27959","ref_index":10,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"ROVER: Routing Object-Centric Visual Evidence for Grounded Multi-Image Reasoning","primary_cat":"cs.CV","submitted_at":"2026-05-27T04:52:42+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"ROVER introduces a learnable routing plugin for object-centric visual evidence in MLLMs via token triplets and differential attention, reporting gains on MM-GCoT and VideoEspresso when integrated into Qwen2.5-VL-7B.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.25842","ref_index":7,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"MuCRASP: Multimodal Chain-of-thought Reasoning aware Structured Pruning","primary_cat":"cs.AI","submitted_at":"2026-05-25T13:36:46+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"MuCRASP prunes VLMs in a CoT-aware manner, outperforming baselines by preserving reasoning quality at 30-50% compression rates on models like Qwen2.5-VL-7B.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2601.09536","ref_index":5,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Omni-R1: Towards the Unified Generative Paradigm for Multimodal Reasoning","primary_cat":"cs.AI","submitted_at":"2026-01-14T14:57:33+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Omni-R1 unifies multimodal reasoning by generating intermediate images during the process in a SFT-plus-RL framework, with an Omni-R1-Zero variant that matches or exceeds it using only text data.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2510.04978","ref_index":20,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Aligning Perception, Reasoning, Modeling and Interaction: A Survey on Physical AI","primary_cat":"cs.AI","submitted_at":"2025-10-06T16:16:03+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"A survey of physical AI that distinguishes theoretical physics reasoning from applied understanding and synthesizes advances in symbolic reasoning, embodied systems, and generative models to advocate for physics-grounded world models.","context_count":1,"top_context_role":"method","top_context_polarity":"use_method","context_text":"ing procedures, while transformer architectures [17], [18] increasingly incorporate physics-aware atten- tion mechanisms and constraints. Beyond archi- tecture,training methodologyplays a crucial role. Researchers have developed physics-guided loss functions that explicitly incorporate physical con- straints [19], curriculum learning strategies that pro- gressively introduce physical complexity [20], and reinforcement learning frameworks with physics- informed reward structures [21], [22], [23]. These training paradigms aim to instill physical intuition during learning rather than relying solely on pat- tern recognition. Attest-time inference, approaches explicitly integrate physical laws and symbolic rea- soning into causal modeling [24], [25], [26], while"}],"limit":50,"offset":0}