Search papers, labs, and topics across Lattice.
SimpleSearch-VL introduces a novel framework for multimodal agentic search that enhances the search-and-verification process without relying on extensive data or complex model components. By employing Factorized Adaptive Rollout (FAR) for improved sampling efficiency and integrating evidence-verified reasoning through chain-of-thought verification, the framework effectively assesses the relevance of visual and textual cues. The approach demonstrates significant performance gains, with a 15.8 and 16.0 point improvement over existing Qwen3-VL baselines for the 8B and 30B-A3B variants, respectively, and competes with the advanced Gemini-3-Pro model.
SimpleSearch-VL achieves a remarkable 15.8-point boost in agentic search performance with minimal data, challenging the need for larger models or extensive training.
We present SimpleSearch-VL, an efficient, reliable, and practical framework for multimodal agentic search. Its core idea is to improve the agent's own search-and-verification process rather than scaling data, tools, or auxiliary model components. For efficiency, Factorized Adaptive Rollout (FAR) improves sampling efficiency by forming more informative training groups while using redundant samples to mitigate long-tail latency and expose hard samples. For reliability, SimpleSearch-VL performs evidence-verified reasoning, explicitly using chain-of-thought verification to assess the relevance of retrieved visual and textual cues to the original context. For practicality, SimpleSearch-VL keeps a lightweight tool interface and performs webpage self-summary within the agent, requiring no additional external dependencies. With only 5K supervised tool-interleaved trajectories and 2K RL data, SimpleSearch-VL improves Qwen3-VL agentic baselines by 15.8 and 16.0 average points for the 8B and 30B-A3B variants, respectively. The SimpleSearch-VL-30B-A3B model further achieves performance competitive with agentic Gemini-3-Pro.
