Towards On-Policy Data Evolution for Visual-Native Multimodal Deep Search Agents
cs.CL
Submitted: 2026-05-11
Updated: 2026-11-04
Comments: NeurIPS 2026
Code: https://github.com/JoeYing1019/ODE
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Back to Basics: Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMs
- Qwen3-VL Technical Report
- Humanity's Last Exam
- Facilitating Multi-turn Function Calling for LLMs via Compositional Instruction Tuning
- OpenSearch-VL: An Open Recipe for Frontier Multimodal Search Agents
- Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities
- Vision-DeepResearch: Incentivizing DeepResearch Capability in Multimodal Large Language Models
- MMSearch: Benchmarking the Potential of Large Models as Multi-modal Search Engines
- MM-BrowseComp: A Comprehensive Benchmark for Multimodal Browsing Agents
- ToolACE: Winning the Points of LLM Function Calling
- Visual Agentic Reinforcement Fine-Tuning
- AgentInstruct: Toward Generative Teaching with Agentic Flows
- DeepMMSearch-R1: Empowering Multimodal LLMs in Multimodal Web Search
- APIGen-MT: Agentic Pipeline for Multi-Turn Data Generation via Simulated Agent-Human Interplay
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- OpenAI GPT-5 System Card
- Thinking with Images for Multimodal Reasoning: Foundations, Methods, and Future Frontiers
- AgentVista: Evaluating Multimodal Agents in Ultra-Challenging Realistic Visual Scenarios
- Tongyi DeepResearch Technical Report
- FVQA: Fact-based Visual Question Answering
Related papers
- ALARM: Audio-Language Alignment for Reasoning Models
- ThinkSum: Probabilistic reasoning over sets using large language models
- Coherence boosting: When your pretrained language model is not paying enough attention
- From Facts to Insights: A Persona-Driven Dual Memory Framework and Dataset for Role-Playing Agents
- Improving General Role-Playing Agents via Psychology-Grounded Reasoning and Role-Aware Policy Optimization
- The Art of Asking: Multilingual Prompt Optimization for Synthetic Data