Zhijian Xu

From Samples to Scenarios: A New Paradigm for Probabilistic Forecasting

Reasoning Scaffolding: Distilling the Flow of Thought from LLMs

Reinforcement Learning with Verifiable Rewards Implicitly Incentivizes Correct Reasoning in Base LLMs

Fidel-TS: A High-Fidelity Benchmark for Multimodal Time Series Forecasting

Solve-Detect-Verify: Inference-Time Scaling with Flexible Generative Verifier

Making Slow Thinking Faster: Compressing LLM Chain-of-Thought via Step Entropy

Influence-Aware Forecasting: Breaking the Self-Stimulation Barrier in Time Series

On the Necessity of Step-Specific Representation Learning for Multi-Step Forecasting

Dependency Matters: Enhancing LLM Reasoning with Explicit Knowledge Grounding

Beyond Trend and Periodicity: Guide Time Series Forecasting with Textual Cues