今日从 arXiv 订阅中筛选 8 篇论文。

Arxiv ID2604.12966幻觉翻译2604.12966

⚡ Boosting Visual Instruction Tuning with Self-Supervised Guidance

Boosting Visual Instruction Tuning with Self-Supervised Guidance
Arxiv ID2604.12425幻觉翻译2604.12425

⚡ Forecasting the Past Gradient-Based Distribution Shift Detection in Trajectory Prediction

Arxiv ID2604.12896幻觉翻译2604.12896

⚡ Don’t Show Pixels, Show Cues Unlocking Visual Tool Reasoning in Language Models via Perception Programs

Don't Show Pixels, Show Cues Unlocking Visual Tool Reasoning in Language Models via Perception Programs
Arxiv ID2604.12335幻觉翻译2604.12335

⚡ All in One A Unified Synthetic Data Pipeline for Multimodal Video Understanding

All in One A Unified Synthetic Data Pipeline for Multimodal Video Understanding
Arxiv ID2604.12346幻觉翻译2604.12346

⚡ Unlocking the Potential of Grounding DINO in Videos Parameter-Efficient Adaptation for Limited-Data Spatial-Temporal Loc

Unlocking the Potential of Grounding DINO in Videos Parameter-Efficient Adaptation for Limited-Data Spatial-Temporal Loc
Arxiv ID2604.13035幻觉翻译2604.13035

⚡ SceneCritic A Symbolic Evaluator for 3D Indoor Scene Synthesis

SceneCritic A Symbolic Evaluator for 3D Indoor Scene Synthesis
Arxiv ID2604.12630幻觉翻译2604.12630

⚡ GeoAlign Geometric Feature Realignment for MLLM Spatial Reasoning

GeoAlign Geometric Feature Realignment for MLLM Spatial Reasoning
Arxiv ID2604.12358幻觉翻译2604.12358

⚡ Why and When Visual Token Pruning Fails A Study on Relevant Visual Information Shift in MLLMs Decoding

Why and When Visual Token Pruning Fails A Study on Relevant Visual Information Shift in MLLMs Decoding

自动生成于 2026-04-16 · 基于 arXiv Daily Digest