Qiran Zhang
Qiran Zhang / Artificial Intelligence / Shanghai
I study how machines read signals, space, and tasks.
I am an undergraduate student majoring in Artificial Intelligence at Shanghai Jiao Tong University. My recent work moves between clinical EEG interpretation, programmatic spatial-temporal reasoning, and the evaluation of AI agents in realistic academic workflows.
Signal From EEG traces to fine-grained clinical language.
Space From programs to coherent temporal worlds.
Task From student challenges to agent evaluation.
Research Threads
Current questionsWhen signals become language
Vision-language models and instruction data for richer EEG interpretation beyond narrow labels.
When code has to keep its geometry
Benchmarks and diagnostics for models that generate runnable programs for animated spatial worlds.
When tasks refuse to be toy tasks
Agent evaluation on long-horizon academic workflows that require planning, tools, and execution.
Selected Work
All publicationsCerebraGloss: Instruction-Tuning a Large Vision-Language Model for Fine-Grained Clinical EEG Interpretation
An EEG-text instruction data engine, CerebraGloss-Bench, and a large vision-language model for generative clinical EEG interpretation.
PRISM: A Benchmark for Programmatic Spatial-Temporal Reasoning
A benchmark for testing whether language models can write runnable code that also preserves spatial and temporal coherence.
AcademiClaw: When Students Set Challenges for AI Agents
A bilingual benchmark of long-horizon academic tasks, sourced from real student workflows rather than simplified prompts.
Notes
CerebraGloss accepted to ICLR 2026.
PRISM and AcademiClaw released as public research benchmarks.
Contact
I am interested in conversations around multimodal model evaluation, code-generation benchmarks, and agent capabilities in realistic academic workflows.