Bo Zhang

Researcher@AI for Science, Shanghai AI Laboratory·中国·OpenReview

研究方向

Scientific Discovery Agents · VLMs in Complex Problem Solving · Long-horizon Reasoning · 3D Domain Continuous Learning; 3D Pre-training · Domain Adaptation; Continuous Learning · Few-shot Learning; Object Detection; Remote Sensing Images Processing · Channel Pruning; Model Compression

Bo Zhang

GeoBench: Rethinking Multimodal Geometric Problem-Solving via Hierarchical Evaluation

InternAgent-MLE: Navigating Fine-Grained Optimization for Coding Agent

GDI-Bench: A Benchmark for General Document Intelligence with Vision and Reasoning Decoupling

MME-Reasoning: A Comprehensive Benchmark for Logical Reasoning in MLLMs

A Benchmark for Self-Evolving Agents via Experience-Driven Lifelong Learning

PointARSim: Point-cloud–enhanced Generative Auto-Regressive Simulation for Closed-loop End-to-end Autonomous Driving Evaluation

RewardCode: Training Generalist Code Reward Model via Pairwise Reinforcement Learning

PhysicsMinions: Winning Gold Medals in the Latest Physics Olympiads with a Coevolutionary Multimodal Multi-Agent System

MCP-R1: Generalized Real-World Task Agent Mastering Dozens of Tools

Self-Tooling Agent: Dynamically Extending Agent Capabilities through Scientific Tool Synthesis and Invocation

ML-Agent: Reinforcing LLM Agents for Autonomous Machine Learning Engineering

OmniQuality-R: Advancing Reward Models through All-Encompassing Quality Assessment

DualResearch: Entropy-Gated Dual-Graph Retrieval for Answer Reconstruction

InternAgent-DR: Advancing deep research with dynamic structured knowledge flow

VideoShield: A Unified Framework for Multimodal Risk Detection and Control in Video Generative Models

TREND: Unsupervised 3D Representation Learning via Temporal Forecasting for LiDAR Perception

OmniCorpus: A Unified Multimodal Corpus of 10 Billion-Level Images Interleaved with Text

GeoX: Geometric Problem Solving Through Unified Formalized Vision-Language Pre-training

Breaking the Compression Ceiling: Data-Free Pipeline for Ultra-Efficient Delta Compression

DocGenome: A Large Benchmark for Multi-Modal Language Models in Real-World Academic Document Understanding

MME-CoT: Benchmarking Chain-of-Thought in Large Multimodal Models for Reasoning Quality, Robustness, and Efficiency

HyperDet: Generalizable Detection of Synthesized Images by Generating and Merging A Mixture of Hyper LoRAs