Agent算法实习生（J103894）
Core
Researching methods to improve LLM agents on difficult tasks by building benchmarks, exploring inference-time search scaling, and developing learning-from-trajectory techniques.
Role type
Research intern (LLM Agents & Inference-Time Scaling)
Builds
Reproducible benchmark systems, agent harnesses, and validation pipelines for industrial decision scenarios.
Domain
Artificial Intelligence / Machine Learning / Operations Research
Deliverable
production ML models
Required skills
Python engineering, LLM Agent harness design, Inference-time search, Reinforcement learning, Benchmark construction, Self-improvement algorithms
Preferred skills
Research publications, Competitive programming, Open-source contributions
Technologies
Python, LLMs, RLHF, AlphaEvolve, FunSearch
Responsibilities
Systematically construct and evaluate task families that cause model failure; Explore scaling paths for inference-time search mechanisms; Research learning from high-quality trajectories; Design and implement agent harness and evaluation systems; Validate methods in real industrial decision scenarios.
Seniority
Intern
